Sadly, leaving crossbar and wamp is imminent

#1

I’m sad to say that we have determined we need to migrate off of crossbar and the wamp protocol.

The number 1 reason is that crossbar does not scale well and it doesn’t appear that it ever will.

It’s currently our linch pin, when crossbar crashes it takes our entire site with it.

There is some promise with the nexus router https://github.com/gammazero/nexus/issues/144

but there isn’t enough movement on this to stay our fears or fix our current issues…

I’ve considered deploying duplicate stack clusters and add load balancing to our clients but this is a costly and potentially problematic solution.

We are seriously looking into gRPC as a replacement, Im wondering if anyone else is having scaling issues?

Has anyone made the change from crossbar to gRPC?

0 Likes

#2

Have you tried Wiola? I didn’t with not too great success, but YMMV. The architecture looks promising though.

···

On Wednesday, 10 October 2018 15:02:20 UTC+11, Greg Keys wrote:

I’m sad to say that we have determined we need to migrate off of crossbar and the wamp protocol.

The number 1 reason is that crossbar does not scale well and it doesn’t appear that it ever will.

It’s currently our linch pin, when crossbar crashes it takes our entire site with it.

There is some promise with the nexus router https://github.com/gammazero/nexus/issues/144

but there isn’t enough movement on this to stay our fears or fix our current issues…

I’ve considered deploying duplicate stack clusters and add load balancing to our clients but this is a costly and potentially problematic solution.

We are seriously looking into gRPC as a replacement, Im wondering if anyone else is having scaling issues?

Has anyone made the change from crossbar to gRPC?

0 Likes

#3

Hi Greg,

As a long-term Crossbar user, I feel your pain and indeed posted a similar question here three years ago.

That said, there is a lot happening on Crossbar that isn’t really visible at the moment, maybe I could tempt you to hold-fire for a short while?

Scaling is now very much on the roadmap from a Crossbar perspective, and I believe there is “currently” work happening on the WAMP spec to facilitate this.
(I think there a developer discussion later today to cover some of the WAMP spec stuff)

In terms of Crossbar itself, there’s a lot of new (maybe hidden?) functionality that is becoming available, some will be in the OS version and some will
only be available in the ‘Enterprise’ version, I’m still waiting to hear exactly how this will be split.

So, this is a screenshot from the new Crossbar UI showing;

  • A connection to a Crossbar Fabric Controller
  • Three management realms (currently selected “MyRealm”)
  • Two connected Crossbar Nodes (Node1/Node2)
  • … etc
  • Screen is looking at the docker instances it can see (split over the two nodes) [Crossbar FX has docker management built-in]
  • Tab to the right lists instances and allows containers to be created
  • Current screen allows stop, start etc
  • Current screen is showing an in-UI console pane connecting to the docker console of the docker container called “Demo1” running on “Node1” … over WAMP. (and the container is running a standard Crossbar cookiecutter example app)
    i.e. [ UI -> pane within UI (xterm.js) -> WAMP -> Crossbar Fabric Controller -> WAMP -> Node1 -> Dockerd -> Container -> Cookie Cutter App ]
  • ‘Start shell’ option on the console window lets you start a new tab in the frame that runs a shell in the same container as the application so you can debug (and alter the code) of the running application inside the running docker container - in real time.
    UI is written using Vue.js and is fully reactive to the extent that someone else creates a container or “pulls” a docker image on a connected node, it’ll popup in the listing in real-time.
    So … stuff is moving behind the scenes … any chance you could document your specific needs / requirement within the context of scaling, maybe I can compare it to the capabilities of stuff currently under development?

console.png

Video demo here; https://www.youtube.com/watch?v=RFTRIWQovjw&t=4s

···

On Wednesday, 10 October 2018 05:02:20 UTC+1, Greg Keys wrote:

I’m sad to say that we have determined we need to migrate off of crossbar and the wamp protocol.

The number 1 reason is that crossbar does not scale well and it doesn’t appear that it ever will.

It’s currently our linch pin, when crossbar crashes it takes our entire site with it.

There is some promise with the nexus router https://github.com/gammazero/nexus/issues/144

but there isn’t enough movement on this to stay our fears or fix our current issues…

I’ve considered deploying duplicate stack clusters and add load balancing to our clients but this is a costly and potentially problematic solution.

We are seriously looking into gRPC as a replacement, Im wondering if anyone else is having scaling issues?

Has anyone made the change from crossbar to gRPC?

0 Likes

#4

The main features we use from crossbar are the websocket connections, rpc and pub sub.
We don’t really use realms or any other wamp/crossbar features.

We are not terribly dependent on the actual wamp protocol so much as we are the router, because of that I can’t really see us using crossbar fabric

Our services are deployed and orchestrated into kubernetes clusters, we deploy a stack (the attached image represents a stack) per environment into their own namespaces

e.g. demo-dev.example.com goes into a demo-dev namespace, demo-qa.example.com goes into demo-qa etc…

we typically run 5 environments per customer. A customer called demo would get the following environments: dev, qa, uat, blue, green

Each environment has its own crossbar router, the only environments that need HA are blue and green, when the router dies so does the entire stack (currently about 50 services).

We have been able to mitigate restarts by increasing resource limits but that can only realistically go so far on any given machine in our clusters.

we really need to be able to run multiple instances of the router each with their own resource limits so that they can spread out in the cluster.

We’ve considered a lot of different options and technologies and have come to the conclusion that we could replicate our current crossbar usage with gRPC and envoy,

envoy scales infinitely, we’d basically run envoy sidecar to all of our services which register to the envoy cluster, which provides the connections to our clients, this would allow

us to scale every aspect of our stack and eliminate the single point of failure.

If crossbar could do something similar we would not need to migrate away from it, but given it’s current path im not sure it fits our needs anymore, it seems that crossbar is positioning itself

on more of a commercial path (Congratulations on that, I understand everyone needs to pay the bills) but that doesnt really work for us because the cost would very likely be prohibitive.

···

On Wednesday, October 10, 2018 at 3:08:53 AM UTC-7, Gareth Bult wrote:

Hi Greg,

As a long-term Crossbar user, I feel your pain and indeed posted a similar question here three years ago.

That said, there is a lot happening on Crossbar that isn’t really visible at the moment, maybe I could tempt you to hold-fire for a short while?

Scaling is now very much on the roadmap from a Crossbar perspective, and I believe there is “currently” work happening on the WAMP spec to facilitate this.
(I think there a developer discussion later today to cover some of the WAMP spec stuff)

In terms of Crossbar itself, there’s a lot of new (maybe hidden?) functionality that is becoming available, some will be in the OS version and some will
only be available in the ‘Enterprise’ version, I’m still waiting to hear exactly how this will be split.

So, this is a screenshot from the new Crossbar UI showing;

  • A connection to a Crossbar Fabric Controller
  • Three management realms (currently selected “MyRealm”)
  • Two connected Crossbar Nodes (Node1/Node2)
  • … etc
  • Screen is looking at the docker instances it can see (split over the two nodes) [Crossbar FX has docker management built-in]
  • Tab to the right lists instances and allows containers to be created
  • Current screen allows stop, start etc
  • Current screen is showing an in-UI console pane connecting to the docker console of the docker container called “Demo1” running on “Node1” … over WAMP. (and the container is running a standard Crossbar cookiecutter example app)
    i.e. [ UI -> pane within UI (xterm.js) -> WAMP -> Crossbar Fabric Controller -> WAMP -> Node1 -> Dockerd -> Container -> Cookie Cutter App ]
  • ‘Start shell’ option on the console window lets you start a new tab in the frame that runs a shell in the same container as the application so you can debug (and alter the code) of the running application inside the running docker container - in real time.
    UI is written using Vue.js and is fully reactive to the extent that someone else creates a container or “pulls” a docker image on a connected node, it’ll popup in the listing in real-time.
    So … stuff is moving behind the scenes … any chance you could document your specific needs / requirement within the context of scaling, maybe I can compare it to the capabilities of stuff currently under development?

console.png

Video demo here; https://www.youtube.com/watch?v=RFTRIWQovjw&t=4s

On Wednesday, 10 October 2018 05:02:20 UTC+1, Greg Keys wrote:

I’m sad to say that we have determined we need to migrate off of crossbar and the wamp protocol.

The number 1 reason is that crossbar does not scale well and it doesn’t appear that it ever will.

It’s currently our linch pin, when crossbar crashes it takes our entire site with it.

There is some promise with the nexus router https://github.com/gammazero/nexus/issues/144

but there isn’t enough movement on this to stay our fears or fix our current issues…

I’ve considered deploying duplicate stack clusters and add load balancing to our clients but this is a costly and potentially problematic solution.

We are seriously looking into gRPC as a replacement, Im wondering if anyone else is having scaling issues?

Has anyone made the change from crossbar to gRPC?

0 Likes

#5

Ok, well firstly thanks for taking the time to detail your setup, sounds like you have something fairly substantial there. Secondly I’m going to add a caveat with regards
to the next bit; the specifics of licensing have yet to be set in stone, so what I’m about to say contains a degree of personal opinion and ‘ideas so far’.

If you look back in these groups, you’ll see I’ve been a user for many users and started asking questions about scalability maybe three years ago (!) so for me, it’s
right there at the top of my ‘must have’ list. Given my background (as an OS user) , having this as an Enterprise only feature makes no sense at all because in order

for people to see that the product ‘does’ scale, you need to be able to see it in action. To this end I’m expecting there to be a ‘free tier’ which includes HA features.

i.e. the product hasn’t suddenly gone ‘all commercial’ :slight_smile:

(which I must admit, some of the documentation might imply)

Whereas the UI is designed to be an Enterprise feature, the same sort of thing applies, if nobody can see it, getting people to pay for it is going to be difficult, so again
there will need to be some sort of free tier in order to promote adoption. If you want to give FX a try with the current public instance of the UI, let me know and I’ll see
if I can sort you out with copies of stuff. (CrossbarFX is currently available as a public download)

The other thing that I had initially envisaged was “Enterprise version == prohibitively expensive” and historically I had ignored all postings relating to the Enterprise
version for just that reason. However, my understanding is that we’re talking (in terms of scale) “Google Docs pricing” rather than “Enterprise Exchange Server pricing”,

i.e. on a small scale it’s probably free, but if you’re working on a larger scale, you pay a very modest fee for a commercially supported product. So if I were to go back
to the OS community, I’m relatively convinced that I would continue to develop HA applications on the free tier, on the basis that I would probably be happy to pay for
(and be able to pay for) the enterprise stuff should I scale that far.

Just going back to your diagram, I’d not seen a use for nor used realms prior to May this year … however ( :slight_smile: ) … after discovering what they do and how easy they are
to use (specifically with the UI) I now use them (literally) for everything. In the context of your configuration for example, typically I would make 1 realm == 1 environment,

which means you get 5 routers / processes instead of 1, which immediately gives 5x the routing capacity vs using a single realm (?!) (not to mention ‘hard’ partitioning
between environments)

(the Rabbit bit is interesting, I kind of came to the conclusion queues could be done in Crossbar which makes for a single messaging library, I’d be interested to hear

about the use-case if it’s something you could share?)

With regards to gRPC, I’ve been using this over the last month as an alternative interface to HTTP when talking to a service that supports both … and in terms of performance
it’s much better than http, maybe between 3 and 6 times faster in real terms. However … (again in terms of performance) it doesn’t seem to compare fare all that well against

WAMP / Crossbar … if you get round to doing any benchmarking and come up with a comparison, again I’d be interested in seeing any results?

(also, I found gRPC relatively painful to develop, maybe I’m too used to Autobahn … :wink: )

With regards to a commercial path … on the one hand there is no future in developing a product like this just for a few corporate users, but on the other hand it’s not possible
to develop the product with zero income … ideally someone would pay us lots of money to develop a completely free Enterprise version … that may still happen … :wink:

Let me know if I can help or clarify …

Regards,
Gareth.

···

On Thursday, 11 October 2018 00:12:47 UTC+1, Greg Keys wrote:

The main features we use from crossbar are the websocket connections, rpc and pub sub.
We don’t really use realms or any other wamp/crossbar features.

We are not terribly dependent on the actual wamp protocol so much as we are the router, because of that I can’t really see us using crossbar fabric

Our services are deployed and orchestrated into kubernetes clusters, we deploy a stack (the attached image represents a stack) per environment into their own namespaces

e.g. demo-dev.example.com goes into a demo-dev namespace, demo-qa.example.com goes into demo-qa etc…

we typically run 5 environments per customer. A customer called demo would get the following environments: dev, qa, uat, blue, green

Each environment has its own crossbar router, the only environments that need HA are blue and green, when the router dies so does the entire stack (currently about 50 services).

We have been able to mitigate restarts by increasing resource limits but that can only realistically go so far on any given machine in our clusters.

we really need to be able to run multiple instances of the router each with their own resource limits so that they can spread out in the cluster.

We’ve considered a lot of different options and technologies and have come to the conclusion that we could replicate our current crossbar usage with gRPC and envoy,

envoy scales infinitely, we’d basically run envoy sidecar to all of our services which register to the envoy cluster, which provides the connections to our clients, this would allow

us to scale every aspect of our stack and eliminate the single point of failure.

If crossbar could do something similar we would not need to migrate away from it, but given it’s current path im not sure it fits our needs anymore, it seems that crossbar is positioning itself

on more of a commercial path (Congratulations on that, I understand everyone needs to pay the bills) but that doesnt really work for us because the cost would very likely be prohibitive.

On Wednesday, October 10, 2018 at 3:08:53 AM UTC-7, Gareth Bult wrote:

Hi Greg,

As a long-term Crossbar user, I feel your pain and indeed posted a similar question here three years ago.

That said, there is a lot happening on Crossbar that isn’t really visible at the moment, maybe I could tempt you to hold-fire for a short while?

Scaling is now very much on the roadmap from a Crossbar perspective, and I believe there is “currently” work happening on the WAMP spec to facilitate this.
(I think there a developer discussion later today to cover some of the WAMP spec stuff)

In terms of Crossbar itself, there’s a lot of new (maybe hidden?) functionality that is becoming available, some will be in the OS version and some will
only be available in the ‘Enterprise’ version, I’m still waiting to hear exactly how this will be split.

So, this is a screenshot from the new Crossbar UI showing;

  • A connection to a Crossbar Fabric Controller
  • Three management realms (currently selected “MyRealm”)
  • Two connected Crossbar Nodes (Node1/Node2)
  • … etc
  • Screen is looking at the docker instances it can see (split over the two nodes) [Crossbar FX has docker management built-in]
  • Tab to the right lists instances and allows containers to be created
  • Current screen allows stop, start etc
  • Current screen is showing an in-UI console pane connecting to the docker console of the docker container called “Demo1” running on “Node1” … over WAMP. (and the container is running a standard Crossbar cookiecutter example app)
    i.e. [ UI -> pane within UI (xterm.js) -> WAMP -> Crossbar Fabric Controller -> WAMP -> Node1 -> Dockerd -> Container -> Cookie Cutter App ]
  • ‘Start shell’ option on the console window lets you start a new tab in the frame that runs a shell in the same container as the application so you can debug (and alter the code) of the running application inside the running docker container - in real time.
    UI is written using Vue.js and is fully reactive to the extent that someone else creates a container or “pulls” a docker image on a connected node, it’ll popup in the listing in real-time.
    So … stuff is moving behind the scenes … any chance you could document your specific needs / requirement within the context of scaling, maybe I can compare it to the capabilities of stuff currently under development?

console.png

Video demo here; https://www.youtube.com/watch?v=RFTRIWQovjw&t=4s

On Wednesday, 10 October 2018 05:02:20 UTC+1, Greg Keys wrote:

I’m sad to say that we have determined we need to migrate off of crossbar and the wamp protocol.

The number 1 reason is that crossbar does not scale well and it doesn’t appear that it ever will.

It’s currently our linch pin, when crossbar crashes it takes our entire site with it.

There is some promise with the nexus router https://github.com/gammazero/nexus/issues/144

but there isn’t enough movement on this to stay our fears or fix our current issues…

I’ve considered deploying duplicate stack clusters and add load balancing to our clients but this is a costly and potentially problematic solution.

We are seriously looking into gRPC as a replacement, Im wondering if anyone else is having scaling issues?

Has anyone made the change from crossbar to gRPC?

0 Likes

#6

(the Rabbit bit is interesting, I kind of came to the conclusion queues could be done in Crossbar which makes for a single messaging library, I’d be interested to hear

about the use-case if it’s something you could share?)

Gareth, can you implement persistent queues with crossbar? (those where subscriber can pick up messages he missed while being down)

···

Zaar

0 Likes

#7

Hi Zaar,

Ok, bear in mind “came to the conclusion …” so I’ve not done this myself, but previously when looking at infrastructure options, that was my thinking.

Firstly you can obviously do this yourself ‘relatively’ easily depending on your architecture with a simple persistent store and a “get_since” call, however
crossbar has a message history option;

https://crossbar.io/docs/Event-History/

Again depends on your architecture / requirement … but I figured from the available options, “another” messaging library wasn’t necessarily a requirement …

… one of the issues we’re seeing atm is documentation and it’s organisation … there are lots of lesser-known features kicking around in Crossbar
which aren’t maybe as easy to find as they could be … :slight_smile:

hth

Gareth

···

On Thursday, 11 October 2018 12:09:26 UTC+1, Zaar Hai wrote:

On Thu, 11 Oct 2018 at 21:27, Gareth Bult gare...@gmail.com wrote:

(the Rabbit bit is interesting, I kind of came to the conclusion queues could be done in Crossbar which makes for a single messaging library, I’d be interested to hear

about the use-case if it’s something you could share?)

Gareth, can you implement persistent queues with crossbar? (those where subscriber can pick up messages he missed while being down)

On Thursday, 11 October 2018 00:12:47 UTC+1, Greg Keys wrote:

The main features we use from crossbar are the websocket connections, rpc and pub sub.
We don’t really use realms or any other wamp/crossbar features.

We are not terribly dependent on the actual wamp protocol so much as we are the router, because of that I can’t really see us using crossbar fabric

Our services are deployed and orchestrated into kubernetes clusters, we deploy a stack (the attached image represents a stack) per environment into their own namespaces

e.g. demo-dev.example.com goes into a demo-dev namespace, demo-qa.example.com goes into demo-qa etc…

we typically run 5 environments per customer. A customer called demo would get the following environments: dev, qa, uat, blue, green

Each environment has its own crossbar router, the only environments that need HA are blue and green, when the router dies so does the entire stack (currently about 50 services).

We have been able to mitigate restarts by increasing resource limits but that can only realistically go so far on any given machine in our clusters.

we really need to be able to run multiple instances of the router each with their own resource limits so that they can spread out in the cluster.

We’ve considered a lot of different options and technologies and have come to the conclusion that we could replicate our current crossbar usage with gRPC and envoy,

envoy scales infinitely, we’d basically run envoy sidecar to all of our services which register to the envoy cluster, which provides the connections to our clients, this would allow

us to scale every aspect of our stack and eliminate the single point of failure.

If crossbar could do something similar we would not need to migrate away from it, but given it’s current path im not sure it fits our needs anymore, it seems that crossbar is positioning itself

on more of a commercial path (Congratulations on that, I understand everyone needs to pay the bills) but that doesnt really work for us because the cost would very likely be prohibitive.

On Wednesday, October 10, 2018 at 3:08:53 AM UTC-7, Gareth Bult wrote:

Hi Greg,

As a long-term Crossbar user, I feel your pain and indeed posted a similar question here three years ago.

That said, there is a lot happening on Crossbar that isn’t really visible at the moment, maybe I could tempt you to hold-fire for a short while?

Scaling is now very much on the roadmap from a Crossbar perspective, and I believe there is “currently” work happening on the WAMP spec to facilitate this.
(I think there a developer discussion later today to cover some of the WAMP spec stuff)

In terms of Crossbar itself, there’s a lot of new (maybe hidden?) functionality that is becoming available, some will be in the OS version and some will
only be available in the ‘Enterprise’ version, I’m still waiting to hear exactly how this will be split.

So, this is a screenshot from the new Crossbar UI showing;

  • A connection to a Crossbar Fabric Controller
  • Three management realms (currently selected “MyRealm”)
  • Two connected Crossbar Nodes (Node1/Node2)
  • … etc
  • Screen is looking at the docker instances it can see (split over the two nodes) [Crossbar FX has docker management built-in]
  • Tab to the right lists instances and allows containers to be created
  • Current screen allows stop, start etc
  • Current screen is showing an in-UI console pane connecting to the docker console of the docker container called “Demo1” running on “Node1” … over WAMP. (and the container is running a standard Crossbar cookiecutter example app)
    i.e. [ UI -> pane within UI (xterm.js) -> WAMP -> Crossbar Fabric Controller -> WAMP -> Node1 -> Dockerd -> Container -> Cookie Cutter App ]
  • ‘Start shell’ option on the console window lets you start a new tab in the frame that runs a shell in the same container as the application so you can debug (and alter the code) of the running application inside the running docker container - in real time.
    UI is written using Vue.js and is fully reactive to the extent that someone else creates a container or “pulls” a docker image on a connected node, it’ll popup in the listing in real-time.
    So … stuff is moving behind the scenes … any chance you could document your specific needs / requirement within the context of scaling, maybe I can compare it to the capabilities of stuff currently under development?

console.png

Video demo here; https://www.youtube.com/watch?v=RFTRIWQovjw&t=4s

On Wednesday, 10 October 2018 05:02:20 UTC+1, Greg Keys wrote:

I’m sad to say that we have determined we need to migrate off of crossbar and the wamp protocol.

The number 1 reason is that crossbar does not scale well and it doesn’t appear that it ever will.

It’s currently our linch pin, when crossbar crashes it takes our entire site with it.

There is some promise with the nexus router https://github.com/gammazero/nexus/issues/144

but there isn’t enough movement on this to stay our fears or fix our current issues…

I’ve considered deploying duplicate stack clusters and add load balancing to our clients but this is a costly and potentially problematic solution.

We are seriously looking into gRPC as a replacement, Im wondering if anyone else is having scaling issues?

Has anyone made the change from crossbar to gRPC?

You received this message because you are subscribed to a topic in the Google Groups “Crossbar” group.

To unsubscribe from this topic, visit https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to crossbario+...@googlegroups.com.

To post to this group, send email to cros...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/crossbario/f7200fb0-500c-4e83-a4b4-76708c9830a5%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.


Zaar

0 Likes

#8

Ok, well firstly thanks for taking the time to detail your setup, sounds like you have something fairly substantial there. Secondly I’m going to add a caveat with regards
to the next bit; the specifics of licensing have yet to be set in stone, so what I’m about to say contains a degree of personal opinion and ‘ideas so far’.

If you look back in these groups, you’ll see I’ve been a user for many users and started asking questions about scalability maybe three years ago (!) so for me, it’s
right there at the top of my ‘must have’ list. Given my background (as an OS user) , having this as an Enterprise only feature makes no sense at all because in order

for people to see that the product ‘does’ scale, you need to be able to see it in action. To this end I’m expecting there to be a ‘free tier’ which includes HA features.

i.e. the product hasn’t suddenly gone ‘all commercial’ :slight_smile:

(which I must admit, some of the documentation might imply)

Glad to hear its hasn’t gone all commercial, that is the sense I got when reading the 11,000 per node in the documentation

Whereas the UI is designed to be an Enterprise feature, the same sort of thing applies, if nobody can see it, getting people to pay for it is going to be difficult, so again
there will need to be some sort of free tier in order to promote adoption. If you want to give FX a try with the current public instance of the UI, let me know and I’ll see
if I can sort you out with copies of stuff. (CrossbarFX is currently available as a public download)

I keep wanting to try FX but havent really had a chance, it’s hard to justify investing more time in something when

it’s future is so unclear. Any free time I have has been spent and defining our needs and looking for things to meet them.

The other thing that I had initially envisaged was “Enterprise version == prohibitively expensive” and historically I had ignored all postings relating to the Enterprise
version for just that reason. However, my understanding is that we’re talking (in terms of scale) “Google Docs pricing” rather than “Enterprise Exchange Server pricing”,

i.e. on a small scale it’s probably free, but if you’re working on a larger scale, you pay a very modest fee for a commercially supported product. So if I were to go back
to the OS community, I’m relatively convinced that I would continue to develop HA applications on the free tier, on the basis that I would probably be happy to pay for
(and be able to pay for) the enterprise stuff should I scale that far.

Just going back to your diagram, I’d not seen a use for nor used realms prior to May this year … however ( :slight_smile: ) … after discovering what they do and how easy they are
to use (specifically with the UI) I now use them (literally) for everything. In the context of your configuration for example, typically I would make 1 realm == 1 environment,

which means you get 5 routers / processes instead of 1, which immediately gives 5x the routing capacity vs using a single realm (?!) (not to mention ‘hard’ partitioning
between environments)

I tried really hard at the beginning to use realms but they only ever complicated things for us. especially now that

we are using namespaces in kubernetes clusters. We’d have to stop deploying crossbar in the namespaces and give it it’s

own namespace, this would make it a single point of failre for 5 environments versus 1 which only compounds the problem

(the Rabbit bit is interesting, I kind of came to the conclusion queues could be done in Crossbar which makes for a single messaging library, I’d be interested to hear

about the use-case if it’s something you could share?)

we are actually working on eliminating rabbitmq ourselves, it’s a huge pain, although we will still use a queue system;

we are currently looking at amazon sqs.

We try to only use the bare minimum features in crossbar, my thinking on that comes from several recommendations from micro service experts to keep your pipelines dumb and build the smarts into

your services, this is a good read https://medium.com/@nathankpeck/microservice-principles-smart-endpoints-and-dumb-pipes-5691d410700f

With regards to gRPC, I’ve been using this over the last month as an alternative interface to HTTP when talking to a service that supports both … and in terms of performance
it’s much better than http, maybe between 3 and 6 times faster in real terms. However … (again in terms of performance) it doesn’t seem to compare fare all that well against

WAMP / Crossbar … if you get round to doing any benchmarking and come up with a comparison, again I’d be interested in seeing any results?

(also, I found gRPC relatively painful to develop, maybe I’m too used to Autobahn … :wink: )

Regarding gRPC, the most attractive thing to me about that is actually envoy, if crossbar were to become more like envoy

that would probably solve all our problems. things like hot restarts without loosing connections, and drop dead simple

clustering, where we could add new instances and it would just work.

With regards to a commercial path … on the one hand there is no future in developing a product like this just for a few corporate users, but on the other hand it’s not possible
to develop the product with zero income … ideally someone would pay us lots of money to develop a completely free Enterprise version … that may still happen … :wink:

Agreed, Im always curious how so many companies give their product away and still make money. I guess its because what they give away isn’t really their primary product?

0 Likes

#9

Hi Greg,

Glad to hear its hasn’t gone all commercial, that is the sense I got when reading the 11,000 per node in the documentation

Erm, this page?

https://crossbario.com/products/enterprise-support/

This is the Enterprise support package … (there are some very large Crossbar users about) … whereas I rather suspect that may need to be updated, I believe it’s quoting the price for support, as opposed to the product.
(but I take your point, some more definitive pricing would be good)

I keep wanting to try FX but havent really had a chance, it’s hard to justify investing more time in something when it’s future is so unclear. Any free time I have has been spent and defining our needs and looking for things to meet them.

Mmm, I’m getting a distinct sense of Deja vu … I think I may’ve said the same thing myself …

We’d have to stop deploying crossbar in the namespaces …

I’ve a CrossbarFX deployment on Google/K8, not sure the use of multiple realms would make any difference in that context … (?)

We try to only use the bare minimum features in crossbar

It is interesting, on the one hand if the intelligence is “all” in the client, then you miss-out on some of the optimisations available by doing bits in the pipeline, and on the other hand you’re potentially far more portable and any bugs will be easier to find as they will be in local code.

Regarding gRPC, the most attractive thing to me about that is actually envoy,

Yeah, I had a look at this when you mentioned it, looks really cool … on the downside (for me) it looks like it’s http2 only and doesn’t support websockets (?)

gRPC does seem to be getting a lot of press / traction, it’ll be interesting to see how it develops …

Agreed, Im always curious how so many companies give their product away and still make money

Well I guess it’s different for every company, but I think in general, support contracts with large customers and custom coding projects probably figure. I think there

are larger companies, without core developers, or with staff retention issues, who use free software. So having someone on the end of a phone who can fix

anything for a price, is an attractive safety net for the well heeled … :slight_smile:

I’ll see if I can find out if there’s anything available yet re; pricing plans …

···

On Wednesday, 17 October 2018 01:22:14 UTC+1, Greg Keys wrote:

Ok, well firstly thanks for taking the time to detail your setup, sounds like you have something fairly substantial there. Secondly I’m going to add a caveat with regards
to the next bit; the specifics of licensing have yet to be set in stone, so what I’m about to say contains a degree of personal opinion and ‘ideas so far’.

If you look back in these groups, you’ll see I’ve been a user for many users and started asking questions about scalability maybe three years ago (!) so for me, it’s
right there at the top of my ‘must have’ list. Given my background (as an OS user) , having this as an Enterprise only feature makes no sense at all because in order

for people to see that the product ‘does’ scale, you need to be able to see it in action. To this end I’m expecting there to be a ‘free tier’ which includes HA features.

i.e. the product hasn’t suddenly gone ‘all commercial’ :slight_smile:

(which I must admit, some of the documentation might imply)

Glad to hear its hasn’t gone all commercial, that is the sense I got when reading the 11,000 per node in the documentation

Whereas the UI is designed to be an Enterprise feature, the same sort of thing applies, if nobody can see it, getting people to pay for it is going to be difficult, so again
there will need to be some sort of free tier in order to promote adoption. If you want to give FX a try with the current public instance of the UI, let me know and I’ll see
if I can sort you out with copies of stuff. (CrossbarFX is currently available as a public download)

I keep wanting to try FX but havent really had a chance, it’s hard to justify investing more time in something when

it’s future is so unclear. Any free time I have has been spent and defining our needs and looking for things to meet them.

The other thing that I had initially envisaged was “Enterprise version == prohibitively expensive” and historically I had ignored all postings relating to the Enterprise
version for just that reason. However, my understanding is that we’re talking (in terms of scale) “Google Docs pricing” rather than “Enterprise Exchange Server pricing”,

i.e. on a small scale it’s probably free, but if you’re working on a larger scale, you pay a very modest fee for a commercially supported product. So if I were to go back
to the OS community, I’m relatively convinced that I would continue to develop HA applications on the free tier, on the basis that I would probably be happy to pay for
(and be able to pay for) the enterprise stuff should I scale that far.

Just going back to your diagram, I’d not seen a use for nor used realms prior to May this year … however ( :slight_smile: ) … after discovering what they do and how easy they are
to use (specifically with the UI) I now use them (literally) for everything. In the context of your configuration for example, typically I would make 1 realm == 1 environment,

which means you get 5 routers / processes instead of 1, which immediately gives 5x the routing capacity vs using a single realm (?!) (not to mention ‘hard’ partitioning
between environments)

I tried really hard at the beginning to use realms but they only ever complicated things for us. especially now that

we are using namespaces in kubernetes clusters. We’d have to stop deploying crossbar in the namespaces and give it it’s

own namespace, this would make it a single point of failre for 5 environments versus 1 which only compounds the problem

(the Rabbit bit is interesting, I kind of came to the conclusion queues could be done in Crossbar which makes for a single messaging library, I’d be interested to hear

about the use-case if it’s something you could share?)

we are actually working on eliminating rabbitmq ourselves, it’s a huge pain, although we will still use a queue system;

we are currently looking at amazon sqs.

We try to only use the bare minimum features in crossbar, my thinking on that comes from several recommendations from micro service experts to keep your pipelines dumb and build the smarts into

your services, this is a good read https://medium.com/@nathankpeck/microservice-principles-smart-endpoints-and-dumb-pipes-5691d410700f

With regards to gRPC, I’ve been using this over the last month as an alternative interface to HTTP when talking to a service that supports both … and in terms of performance
it’s much better than http, maybe between 3 and 6 times faster in real terms. However … (again in terms of performance) it doesn’t seem to compare fare all that well against

WAMP / Crossbar … if you get round to doing any benchmarking and come up with a comparison, again I’d be interested in seeing any results?

(also, I found gRPC relatively painful to develop, maybe I’m too used to Autobahn … :wink: )

Regarding gRPC, the most attractive thing to me about that is actually envoy, if crossbar were to become more like envoy

that would probably solve all our problems. things like hot restarts without loosing connections, and drop dead simple

clustering, where we could add new instances and it would just work.

With regards to a commercial path … on the one hand there is no future in developing a product like this just for a few corporate users, but on the other hand it’s not possible
to develop the product with zero income … ideally someone would pay us lots of money to develop a completely free Enterprise version … that may still happen … :wink:

Agreed, Im always curious how so many companies give their product away and still make money. I guess its because what they give away isn’t really their primary product?

0 Likes

#10

Yes, please keep us posted once you have anything. The one company I have in mind is Elastic (did an IPO couple of weeks ago) which have their ElasticSearch database open-sourced since day one, including scaleability;

and they provide support and enterprise goodies like authn/z and SSL for a fee.

Though their model is “all or nothing” and “all” comes for rather high price starting from 5k USD per node per year - that’s why those 11k for crossbar rang a bell, at least for me.

···

On Fri, 19 Oct 2018 at 11:12, Gareth Bult garet...@gmail.com wrote:

Agreed, Im always curious how so many companies give their product away and still make money

Well I guess it’s different for every company, but I think in general, support contracts with large customers and custom coding projects probably figure. I think there

are larger companies, without core developers, or with staff retention issues, who use free software. So having someone on the end of a phone who can fix

anything for a price, is an attractive safety net for the well heeled … :slight_smile:

I’ll see if I can find out if there’s anything available yet re; pricing plans …

0 Likes

#11

Current Crossbar FX /CFC pricing can be found here; https://crossbario.com/products/edge_and_cloud/

Little bit cheaper … than enterprise support … :slight_smile:

···

On Friday, 19 October 2018 06:50:00 UTC+1, Zaar Hai wrote:

On Fri, 19 Oct 2018 at 11:12, Gareth Bult gare...@gmail.com wrote:

Agreed, Im always curious how so many companies give their product away and still make money

Well I guess it’s different for every company, but I think in general, support contracts with large customers and custom coding projects probably figure. I think there

are larger companies, without core developers, or with staff retention issues, who use free software. So having someone on the end of a phone who can fix

anything for a price, is an attractive safety net for the well heeled … :slight_smile:

I’ll see if I can find out if there’s anything available yet re; pricing plans …

Yes, please keep us posted once you have anything. The one company I have in mind is Elastic (did an IPO couple of weeks ago) which have their ElasticSearch database open-sourced since day one, including scaleability;

and they provide support and enterprise goodies like authn/z and SSL for a fee.

Though their model is “all or nothing” and “all” comes for rather high price starting from 5k USD per node per year - that’s why those 11k for crossbar rang a bell, at least for me.

0 Likes

#12

That’s really interesting.
If all my “sensors” talk to the internet, I don’t need any Edge devices, right? Same for users coming from web-browser - they just connect to Crossbar.io Fabric Cloud directly, right?

Also, for Router in Starter pack it says “1 x 4 cores”. Does it mean that crossbar router scales across multiple cores for a single realm?

···

Zaar

0 Likes

#13

Mmm, not quite, all your sensors will talk to a (CrossbarFX) edge node, but you’ll get a lot of connections per node …

CFC is a mechanism for linking / controlling all your (CrossbarFX) edge nodes, rather than for connecting sensors to.

(which comes with a UI for doing dynamic config etc)

To be honest, I don’t understand the ‘cores’ specifications so I’m going to pass on that one …

If you have a potential configuration, I’m happy to try to clarify the configuration / cost …

···

On Friday, 19 October 2018 12:24:46 UTC+1, Zaar Hai wrote:

That’s really interesting.
If all my “sensors” talk to the internet, I don’t need any Edge devices, right? Same for users coming from web-browser - they just connect to Crossbar.io Fabric Cloud directly, right?

Also, for Router in Starter pack it says “1 x 4 cores”. Does it mean that crossbar router scales across multiple cores for a single realm?

On Fri, 19 Oct 2018 at 20:37, Gareth Bult gare...@gmail.com wrote:

Current Crossbar FX /CFC pricing can be found here; https://crossbario.com/products/edge_and_cloud/

Little bit cheaper … than enterprise support … :slight_smile:

On Friday, 19 October 2018 06:50:00 UTC+1, Zaar Hai wrote:

On Fri, 19 Oct 2018 at 11:12, Gareth Bult gare...@gmail.com wrote:

Agreed, Im always curious how so many companies give their product away and still make money

Well I guess it’s different for every company, but I think in general, support contracts with large customers and custom coding projects probably figure. I think there

are larger companies, without core developers, or with staff retention issues, who use free software. So having someone on the end of a phone who can fix

anything for a price, is an attractive safety net for the well heeled … :slight_smile:

I’ll see if I can find out if there’s anything available yet re; pricing plans …

Yes, please keep us posted once you have anything. The one company I have in mind is Elastic (did an IPO couple of weeks ago) which have their ElasticSearch database open-sourced since day one, including scaleability;

and they provide support and enterprise goodies like authn/z and SSL for a fee.

Though their model is “all or nothing” and “all” comes for rather high price starting from 5k USD per node per year - that’s why those 11k for crossbar rang a bell, at least for me.

You received this message because you are subscribed to a topic in the Google Groups “Crossbar” group.

To unsubscribe from this topic, visit https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to crossbario+...@googlegroups.com.

To post to this group, send email to cros...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.


Zaar

0 Likes

#14

Te “cores” was actually the most interesting part :slight_smile: I hope the below snapshot (from the link you sent) will help to explain what I am asking about.

So, apologies if I’m taking the discussion off-topic, but I would like to understand the product.

You say CFC is for “linking / controlling edge nodes”. What do you mean by “linking”? - Will I be able to have one WAMP realm spread over multiple Edge Crossbars?

Also, the page says “Crossbar.io Fabric Edge is priced per edge device” - does it mean “per edge crossbar router”? or “per edge sensor/actuator/etc. that connects to edge crossbar router”?

image.png

···

Zaar

0 Likes

#15

Yeah sorry, I’m going to have to ask specifically how the cores / pricing relate.

The idea is to have one WAMP realm spread over multiple Edge Nodes, indeed one of the challenges I’m looking at currently is how to get the UI to cope with 1M edge nodes on 1 realm.

I know there have been lots of discussions around pricing, and looking at the table I can see how it could be interpreted in different ways … let me see if I can an authoritative comment / answer and come back to you … :slight_smile:

···

On Monday, 22 October 2018 08:53:05 UTC+1, Zaar Hai wrote:

Te “cores” was actually the most interesting part :slight_smile: I hope the below snapshot (from the link you sent) will help to explain what I am asking about.

So, apologies if I’m taking the discussion off-topic, but I would like to understand the product.

You say CFC is for “linking / controlling edge nodes”. What do you mean by “linking”? - Will I be able to have one WAMP realm spread over multiple Edge Crossbars?

Also, the page says “Crossbar.io Fabric Edge is priced per edge device” - does it mean “per edge crossbar router”? or “per edge sensor/actuator/etc. that connects to edge crossbar router”?

image.png

On Fri, 19 Oct 2018 at 22:42, Gareth Bult gare...@gmail.com wrote:

Mmm, not quite, all your sensors will talk to a (CrossbarFX) edge node, but you’ll get a lot of connections per node …

CFC is a mechanism for linking / controlling all your (CrossbarFX) edge nodes, rather than for connecting sensors to.

(which comes with a UI for doing dynamic config etc)

To be honest, I don’t understand the ‘cores’ specifications so I’m going to pass on that one …

If you have a potential configuration, I’m happy to try to clarify the configuration / cost …

On Friday, 19 October 2018 12:24:46 UTC+1, Zaar Hai wrote:

That’s really interesting.
If all my “sensors” talk to the internet, I don’t need any Edge devices, right? Same for users coming from web-browser - they just connect to Crossbar.io Fabric Cloud directly, right?

Also, for Router in Starter pack it says “1 x 4 cores”. Does it mean that crossbar router scales across multiple cores for a single realm?

On Fri, 19 Oct 2018 at 20:37, Gareth Bult gare...@gmail.com wrote:

Current Crossbar FX /CFC pricing can be found here; https://crossbario.com/products/edge_and_cloud/

Little bit cheaper … than enterprise support … :slight_smile:

On Friday, 19 October 2018 06:50:00 UTC+1, Zaar Hai wrote:

On Fri, 19 Oct 2018 at 11:12, Gareth Bult gare...@gmail.com wrote:

Agreed, Im always curious how so many companies give their product away and still make money

Well I guess it’s different for every company, but I think in general, support contracts with large customers and custom coding projects probably figure. I think there

are larger companies, without core developers, or with staff retention issues, who use free software. So having someone on the end of a phone who can fix

anything for a price, is an attractive safety net for the well heeled … :slight_smile:

I’ll see if I can find out if there’s anything available yet re; pricing plans …

Yes, please keep us posted once you have anything. The one company I have in mind is Elastic (did an IPO couple of weeks ago) which have their ElasticSearch database open-sourced since day one, including scaleability;

and they provide support and enterprise goodies like authn/z and SSL for a fee.

Though their model is “all or nothing” and “all” comes for rather high price starting from 5k USD per node per year - that’s why those 11k for crossbar rang a bell, at least for me.

You received this message because you are subscribed to a topic in the Google Groups “Crossbar” group.

To unsubscribe from this topic, visit https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to crossbario+...@googlegroups.com.

To post to this group, send email to cros...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.


Zaar

You received this message because you are subscribed to a topic in the Google Groups “Crossbar” group.

To unsubscribe from this topic, visit https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to crossbario+...@googlegroups.com.

To post to this group, send email to cros...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/crossbario/80de67e9-e8d2-4ab6-9220-e379acd3b02e%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.


Zaar

0 Likes

#16

Hi Zaar,

let me chime in;) I'll add some infos ..

Te "cores" was actually the most interesting part :slight_smile: I hope the below snapshot (from the link you sent) will help to explain what I am asking about.

So, apologies if I'm taking the discussion off-topic, but I would like to understand the product.
You say CFC is for "linking / controlling edge nodes". What do you mean by "linking"? - Will I be able to have one WAMP realm spread over multiple Edge Crossbars?

"linking"

this means an edge node maintains a management connection (WAMP) to CFC so that it can be remotely managed and monitored. this connection does not carry any app traffic - only control rgd crossbar itself.

"one realm spread over multiple edge nodes?"

you can define one app realm, and have all settings and configuration automatically _replicated_ to multiple edge nodes

ultimately, you will also be able to have one app realm also _span_ multiple edge nodes, which will require at least one cloud anchored node too (with public routable IP)

in such a topology, an edge node routes events/calls destined for the other edge node over the cloud node (because the assumption is that the edge nodes are not reachable from public internet)

Also, the page says "Crossbar.io Fabric Edge is priced per edge device" - does it mean "per edge crossbar router"? or "per edge sensor/actuator/etc. that connects to edge crossbar router"?

the former.

an edge device (hardware, like an industrial PC, or a Pi) running a Crossbar.io Fabric Edge node counts as one.

an Autobahn (or general WAMP) client does _not_ count (is free)

Cheers,
/Tobias

···

Am 22.10.2018 um 09:52 schrieb Zaar Hai:

image.png

On Fri, 19 Oct 2018 at 22:42, Gareth Bult <garet...@gmail.com > <mailto:garet...@gmail.com>> wrote:

    Mmm, not quite, all your sensors will talk to a (CrossbarFX) edge
    node, but you'll get a lot of connections per node ...

    CFC is a mechanism for linking / controlling all your (CrossbarFX)
    edge nodes, rather than for connecting sensors to.
    (which comes with a UI for doing dynamic config etc)

    To be honest, I don't understand the 'cores' specifications so I'm
    going to pass on that one ..

    If you have a potential configuration, I'm happy to try to clarify
    the configuration / cost ..

    On Friday, 19 October 2018 12:24:46 UTC+1, Zaar Hai wrote:

        That's really interesting.
        If all my "sensors" talk to the internet, I don't need any Edge
        devices, right? Same for users coming from web-browser - they
        just connect to Crossbar.io Fabric Cloud directly, right?

        Also, for Router in Starter pack it says "1 x 4 cores". Does it
        mean that crossbar router scales across multiple cores for a
        single realm?

        On Fri, 19 Oct 2018 at 20:37, Gareth Bult <gare...@gmail.com> > wrote:

            Current Crossbar FX /CFC pricing can be found here;
            https://crossbario.com/products/edge_and_cloud/ ...

            Little bit cheaper .. than enterprise support .. :slight_smile:

            On Friday, 19 October 2018 06:50:00 UTC+1, Zaar Hai wrote:

                On Fri, 19 Oct 2018 at 11:12, Gareth Bult > <gare...@gmail.com> wrote:

                     > Agreed, Im always curious how so many companies
                    give their product away and still make money

                    Well I guess it's different for every company, but I
                    think in general, support contracts with large
                    customers and custom coding projects probably
                    figure. I think there
                    are larger companies, without core developers, or
                    with staff retention issues, who use free software.
                    So having someone on the end of a phone who can fix
                    anything for a price, is an attractive safety net
                    for the well heeled .. :slight_smile:

                    I'll see if I can find out if there's anything
                    available yet re; pricing plans ...

                Yes, please keep us posted once you have anything. The
                one company I have in mind is Elastic (did an IPO couple
                of weeks ago) which have their ElasticSearch database
                open-sourced since day one, including scaleability;
                and they provide support and enterprise goodies like
                authn/z and SSL for a fee.
                Though their model is "all or nothing" and "all" comes
                for rather high price starting from 5k USD per node per
                year - that's why those 11k for crossbar rang a bell, at
                least for me.

            -- You received this message because you are subscribed to a
            topic in the Google Groups "Crossbar" group.
            To unsubscribe from this topic, visit
            https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe.
            To unsubscribe from this group and all its topics, send an
            email to crossba...@googlegroups.com.
            To post to this group, send email to cros...@googlegroups.com.
            To view this discussion on the web visit
            https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com
            <https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com?utm_medium=email&utm_source=footer>.
            For more options, visit https://groups.google.com/d/optout.

        -- Zaar

    -- You received this message because you are subscribed to a topic in
    the Google Groups "Crossbar" group.
    To unsubscribe from this topic, visit
    https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe.
    To unsubscribe from this group and all its topics, send an email to
    crossbario+...@googlegroups.com
    <mailto:crossbario+...@googlegroups.com>.
    To post to this group, send email to cross...@googlegroups.com
    <mailto:cross...@googlegroups.com>.
    To view this discussion on the web visit
    https://groups.google.com/d/msgid/crossbario/80de67e9-e8d2-4ab6-9220-e379acd3b02e%40googlegroups.com
    <https://groups.google.com/d/msgid/crossbario/80de67e9-e8d2-4ab6-9220-e379acd3b02e%40googlegroups.com?utm_medium=email&utm_source=footer>.
    For more options, visit https://groups.google.com/d/optout.

--
Zaar

--
You received this message because you are subscribed to the Google Groups "Crossbar" group.
To unsubscribe from this group and stop receiving emails from it, send an email to crossbario+...@googlegroups.com <mailto:crossbario+...@googlegroups.com>.
To post to this group, send email to cross...@googlegroups.com <mailto:cross...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/crossbario/CAGQ_5kVdqFP2B0qJ3cr0C3LNwRv_atF3%3Di4sKNMfsuyGrOiJAA%40mail.gmail.com <https://groups.google.com/d/msgid/crossbario/CAGQ_5kVdqFP2B0qJ3cr0C3LNwRv_atF3%3Di4sKNMfsuyGrOiJAA%40mail.gmail.com?utm_medium=email&utm_source=footer>.
For more options, visit https://groups.google.com/d/optout.

0 Likes

#17

That sounds quite complex.
I have to be honest I’m a bit lost on the crossbarFX, none of it looks like it fits our needs.

I’m getting the sense that crossbarFX is moving more toward being a router in the cloud that iot can connect to, which

I can see being a great choice for iot users.

We don’t use crossbar for iot.

we use it more like nginx where our client connects to the router in order to access api endpoints instead of restful end points.

It’s really import that we run everything on our hardware in our data centers (AWS), I can’t imagine we would ever have all of our services connecting to

a router service hosted elsewhere. We just need the router to scale so that its no longer a single point of failure.

I’m getting the sense that we are probably not the target market for crossbar, am I wrong about that?

···

On Monday, October 22, 2018 at 2:36:15 AM UTC-7, Tobias Oberstein wrote:

Hi Zaar,

let me chime in;) I’ll add some infos …

Am 22.10.2018 um 09:52 schrieb Zaar Hai:

Te “cores” was actually the most interesting part :slight_smile: I hope the below
snapshot (from the link you sent) will help to explain what I am asking
about.

So, apologies if I’m taking the discussion off-topic, but I would like
to understand the product.

You say CFC is for “linking / controlling edge nodes”. What do you mean
by “linking”? - Will I be able to have one WAMP realm spread over
multiple Edge Crossbars?

“linking”

this means an edge node maintains a management connection (WAMP) to CFC
so that it can be remotely managed and monitored. this connection does
not carry any app traffic - only control rgd crossbar itself.

“one realm spread over multiple edge nodes?”

you can define one app realm, and have all settings and configuration
automatically replicated to multiple edge nodes

ultimately, you will also be able to have one app realm also span
multiple edge nodes, which will require at least one cloud anchored node
too (with public routable IP)

in such a topology, an edge node routes events/calls destined for the
other edge node over the cloud node (because the assumption is that the
edge nodes are not reachable from public internet)

Also, the page says “Crossbar.io Fabric Edge is priced per edge device”

  • does it mean “per edge crossbar router”? or “per edge
    sensor/actuator/etc. that connects to edge crossbar router”?

the former.

an edge device (hardware, like an industrial PC, or a Pi) running a
Crossbar.io Fabric Edge node counts as one.

an Autobahn (or general WAMP) client does not count (is free)

Cheers,

/Tobias

image.png

On Fri, 19 Oct 2018 at 22:42, Gareth Bult <gare...@gmail.com > > > mailto:gare...@gmail.com> wrote:

Mmm, not quite, all your sensors will talk to a (CrossbarFX) edge
node, but you'll get a lot of connections per node ...
CFC is a mechanism for linking / controlling all your (CrossbarFX)
edge nodes, rather than for connecting sensors to.
(which comes with a UI for doing dynamic config etc)
To be honest, I don't understand the 'cores' specifications so I'm
going to pass on that one ..
If you have a potential configuration, I'm happy to try to clarify
the configuration / cost ..
On Friday, 19 October 2018 12:24:46 UTC+1, Zaar Hai wrote:
    That's really interesting.
    If all my "sensors" talk to the internet, I don't need any Edge
    devices, right? Same for users coming from web-browser - they
    just connect to Crossbar.io Fabric Cloud directly, right?
    Also, for Router in Starter pack it says "1 x 4 cores". Does it
    mean that crossbar router scales across multiple cores for a
    single realm?
    On Fri, 19 Oct 2018 at 20:37, Gareth Bult <gare...@gmail.com> >  > >         wrote:
        Current Crossbar FX /CFC pricing can be found here;
        [https://crossbario.com/products/edge_and_cloud/](https://crossbario.com/products/edge_and_cloud/) ...
        Little bit cheaper .. than enterprise support .. :)
        On Friday, 19 October 2018 06:50:00 UTC+1, Zaar Hai wrote:
            On Fri, 19 Oct 2018 at 11:12, Gareth Bult >  > >                 <gare...@gmail.com> wrote:
                 > Agreed, Im always curious how so many companies
                give their product away and still make money
                Well I guess it's different for every company, but I
                think in general, support contracts with large
                customers and custom coding projects probably
                figure. I think there
                are larger companies, without core developers, or
                with staff retention issues, who use free software.
                So having someone on the end of a phone who can fix
                anything for a price, is an attractive safety net
                for the well heeled .. :)
                I'll see if I can find out if there's anything
                available yet re; pricing plans ...
            Yes, please keep us posted once you have anything. The
            one company I have in mind is Elastic (did an IPO couple
            of weeks ago) which have their ElasticSearch database
            open-sourced since day one, including scaleability;
            and they provide support and enterprise goodies like
            authn/z and SSL for a fee.
            Though their model is "all or nothing" and "all" comes
            for rather high price starting from 5k USD per node per
            year - that's why those 11k for crossbar rang a bell, at
            least for me.
        --
        You received this message because you are subscribed to a
        topic in the Google Groups "Crossbar" group.
        To unsubscribe from this topic, visit
        [https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe](https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe).
        To unsubscribe from this group and all its topics, send an
        email to crossbario+...@googlegroups.com.
        To post to this group, send email to cros...@googlegroups.com.
        To view this discussion on the web visit
        [https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com](https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com)
        <[https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com?utm_medium=email&utm_source=footer](https://groups.google.com/d/msgid/crossbario/0e4d482a-4e65-4bbd-8a9d-a1d11ddcab83%40googlegroups.com?utm_medium=email&utm_source=footer)>.
        For more options, visit [https://groups.google.com/d/optout](https://groups.google.com/d/optout).
    --
    Zaar
--
You received this message because you are subscribed to a topic in
the Google Groups "Crossbar" group.
To unsubscribe from this topic, visit
[https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe](https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe).
To unsubscribe from this group and all its topics, send an email to
crossbario+...@googlegroups.com
<mailto:crossbario+unsub...@googlegroups.com>.
To post to this group, send email to cros...@googlegroups.com
<mailto:cross...@googlegroups.com>.
To view this discussion on the web visit
[https://groups.google.com/d/msgid/crossbario/80de67e9-e8d2-4ab6-9220-e379acd3b02e%40googlegroups.com](https://groups.google.com/d/msgid/crossbario/80de67e9-e8d2-4ab6-9220-e379acd3b02e%40googlegroups.com)
<[https://groups.google.com/d/msgid/crossbario/80de67e9-e8d2-4ab6-9220-e379acd3b02e%40googlegroups.com?utm_medium=email&utm_source=footer](https://groups.google.com/d/msgid/crossbario/80de67e9-e8d2-4ab6-9220-e379acd3b02e%40googlegroups.com?utm_medium=email&utm_source=footer)>.
For more options, visit [https://groups.google.com/d/optout](https://groups.google.com/d/optout).


Zaar


You received this message because you are subscribed to the Google
Groups “Crossbar” group.

To unsubscribe from this group and stop receiving emails from it, send
an email to crossbario+...@googlegroups.com

mailto:crossbario+unsub...@googlegroups.com.

To post to this group, send email to cros...@googlegroups.com

mailto:cross...@googlegroups.com.

To view this discussion on the web visit
https://groups.google.com/d/msgid/crossbario/CAGQ_5kVdqFP2B0qJ3cr0C3LNwRv_atF3%3Di4sKNMfsuyGrOiJAA%40mail.gmail.com

<https://groups.google.com/d/msgid/crossbario/CAGQ_5kVdqFP2B0qJ3cr0C3LNwRv_atF3%3Di4sKNMfsuyGrOiJAA%40mail.gmail.com?utm_medium=email&utm_source=footer>.

For more options, visit https://groups.google.com/d/optout.

0 Likes

#18

I’m getting the sense that we are probably not the target market for crossbar, am I wrong about that?

Target market is an interesting question. From a personal perspective I started using Crossbar (quite some time ago now) for an IoT project … well … sort of …

custom door entry system … so lots of Raspberry Pi’s with Camera’s maglocks and RFID readers all talking to a central Crossbar server over WAMP. As a part

of the project I had to produce a UI showing open doors, who was in each room etc … so I hung that off Crossbar too … at which point it dawned on me that IoT

aside, Crossbar / WAMP was a great solution for real-time reactive websites and online applications, and that’s the use-case that’s sort of held my interest.

So the pundits reckon 20 billion IoT devices out there by 2021, and there are already a billion websites … so in theory both markets have lots of potential.

… from a product capability perspective, I would say “yes” … from a marketing perspective, maybe not at the moment (?) , but that’s not so say

the use-case isn’t being considered … if I can expand a little on the basis of my understanding … (someone will hopefully correct me if I get a bit wrong :slight_smile: )

CrossbarFX comes as a binary which will run in either master or edge mode.

Master mode is for linking edge node’s together (and managing edge nodes)

Edge mode provides something for your WAMP clients (and/or edge devices) to connect to.

So, to run a realm across multiple nodes you would need;

a. crossbarfx running in master mode on an ip that all desired edge nodes can connect to

b. two crossbarfxs running in edge mode such that they can connect to the master

c. the as-yet unreleased bit of software that provides same-realm routing between edge nodes via the master node

(so the realm would run on the edge nodes and the master node would provide the interconnect)

My understanding is that there will be a free-tier version of this such that you could set up two edge nodes for basic scaling, but once you start to expand, then

you expand into a paid-for per-node tier.

So, whereas FX will do a lot of things you don’t need, there are some important features going in there that you might find useful. In addition to the scaling and the

delivery format (deploying Crossbar as a pre-compiled [python] binary is very convenient) FX has the ability to access and control “docker” on the machine

on which it’s deployed … so if you build your client applications / microservices as docker images, it can pull, deploy, start, stop, update and monitor your client

applications from within the UI, and changing the crossbar configuration no longer means having to edit config.json … :slight_smile:

At first glance Docker might not appear of any use of you’re not a docker user, but it’s effectively providing package management and partitioning for your microservices with very little fuss.

(I wasn’t really a Docker fan, however in this context it’s sort of growing on me …)

Anyway, hope this helps a little …

···
0 Likes

#19

So, whereas FX will do a lot of things you don’t need, there are some important features going in there that you might find useful. In addition to the scaling and the

delivery format (deploying Crossbar as a pre-compiled [python] binary is very convenient) FX has the ability to access and control “docker” on the machine

on which it’s deployed … so if you build your client applications / microservices as docker images, it can pull, deploy, start, stop, update and monitor your client

applications from within the UI, and changing the crossbar configuration no longer means having to edit config.json … :slight_smile:

May be that makes sense in IoT, but living with Docker & Kubernetes for the last 3 years, the above statements may sound like you’ll be competing with the latter :slight_smile:

I already have K8s/Helm for deployment/scaling/etc., Prometheus for monitoring, Graylog for log collection and so on. For me, WAMP is just another transport bus inside my

application cluster, along with HTTP/REST, persistent pub/sub, etc. With all do respect, I find it hard to imagine considering Crossbar as my deployment manager

(if this is what you mean).

···

Zaar

0 Likes

#20

Agreed, it does appear crossbarFX is going down the path of orchestration, scheduling and management.

We too are using kubernetes for that. We recently switch from using Rancher cattle, Im not looking to change to yet another

orchestrator, kubernetes does everything we ever wanted and more.

···

On Monday, October 22, 2018 at 7:44:18 PM UTC-7, Zaar Hai wrote:

On Tue, 23 Oct 2018 at 13:19, Gareth Bult gare...@gmail.com wrote:

So, whereas FX will do a lot of things you don’t need, there are some important features going in there that you might find useful. In addition to the scaling and the

delivery format (deploying Crossbar as a pre-compiled [python] binary is very convenient) FX has the ability to access and control “docker” on the machine

on which it’s deployed … so if you build your client applications / microservices as docker images, it can pull, deploy, start, stop, update and monitor your client

applications from within the UI, and changing the crossbar configuration no longer means having to edit config.json … :slight_smile:

May be that makes sense in IoT, but living with Docker & Kubernetes for the last 3 years, the above statements may sound like you’ll be competing with the latter :slight_smile:

I already have K8s/Helm for deployment/scaling/etc., Prometheus for monitoring, Graylog for log collection and so on. For me, WAMP is just another transport bus inside my

application cluster, along with HTTP/REST, persistent pub/sub, etc. With all do respect, I find it hard to imagine considering Crossbar as my deployment manager

(if this is what you mean).

At first glance Docker might not appear of any use of you’re not a docker user, but it’s effectively providing package management and partitioning for your microservices with very little fuss.

(I wasn’t really a Docker fan, however in this context it’s sort of growing on me …)

Anyway, hope this helps a little …

You received this message because you are subscribed to a topic in the Google Groups “Crossbar” group.

To unsubscribe from this topic, visit https://groups.google.com/d/topic/crossbario/m0yE-wNAbYU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to crossbario+...@googlegroups.com.

To post to this group, send email to cros...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/crossbario/a819a677-3995-425d-bd95-b0f94468847e%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.


Zaar

0 Likes