[OpenSIPS-Users] [OpenSIPS-Devel] Fwd: RTPproxy project
bogdan at opensips.org
Tue Jun 17 19:00:24 CEST 2014
Thank you for the your input.
I have also a long list of goodies to be added to a rtpproxy (fixes in
multi-stream handling, new features, etc) and this is why I started this
discussion around the RTPproxy - I wanted to understand what are the
plans for the future.
To be honest I was not aware of the latest work Maxim did on rtpproxy
(it is not so transparent via the rtpproxy.org site or mailing lists) -
but it looks interesting and before doing any steps further I will need
to look into that.
As it is not clear for me, could you detail a bit on:
"multiple opensips instances can talk to the same rtp_cluster, meaning
that you get a distributed session state map can be highly available"
What do you mean by "session state map can be highly available" ?
Also , on :
"Maybe adding rtpproxy session replication through the binary data
interface recently introduced to opensips could help with some of this"
You mean to replicate the info on sessions between multiple rtpproxy
Thanks and regards,
OpenSIPS Founder and Developer
On 14.06.2014 16:17, Bobby Smith wrote:
> For us, transcoding support, being able to fork rtcp to a reporting
> server (or even better, becoming an rtcp client), and (in a distant
> third) SRTP termination would be the big features we're looking for.
> We would love to be able to do something around the nathelper
> API/rtpproxy API to offload to a transcoder only when we need it,
> similar to OpenSIPS added sangoma support, only with rtpproxy. If the
> mechanism was added, we'd probably contribute to some codec
> translation tables (remember, g711 and ilbc are eerily similar in
> frame size). FreeSWITCH does it by converting everything down to a
> common binary format before transcode, which is how I'd envision this
> would work. Transcoding in software only is a huge value-add gain for
> us, because it allows us to continue with a software only solution and
> scale easier.
> We engineer around losing an rtpproxy on the wire (which almost never
> happens), with a few strategies around re-INVITE, silence packet
> detection, etc. It's not perfect, but it's well within the SLA we'd
> present our end users and it's probably an area of the system where we
> get the least complaints. A dropped call every few weeks is beyond
> acceptable from the generation that's used to going from 4 bars to 0
> in a subway tunnel.
> And rtp_cluster has been great for us, solely for the ability to tag
> an instance as down and bleed sessions off of it before terminating
> it. The load balancing is on parity with the features added to the
> rtpproxy module via opensips, with exception of this: multiple
> opensips instances can talk to the same rtp_cluster, meaning that you
> get a distributed session state map can be highly available, instead
> of relying upon what's in memory with opensips. That's how we achieve
> the failover features that I think the community want added. Maybe
> adding rtpproxy session replication through the binary data interface
> recently introduced to opensips could help with some of this.
> So yes, feature parity is important, but it's also important that we
> maintain reliable performance. I know Maxim has worked on some stuff
> around threading that has helped us move forward to better reliability
> with rtpp (separating command protocol from packet processing), so
> there's some progress there. The last time we did a comparison and
> made a decision between the two major entities, we just found rtpproxy
> to be much better performant at handling multiple sessions per
> instance, in the 50-60% better range. We can squeeze around 6000
> established "sessions" (if you come from an eSBC world) on an
> m3.xlarge ec2 instance and not break a sweat.
> Ultimately, I think it's good for all of the community to show that a
> project is in active development. I think it's a win for both sides,
> and discussions on where something is going are well warranted.
> And this is coming from the SP with a capital V.
> On Fri, Jun 13, 2014 at 1:55 PM, <ag at ag-projects.com
> <mailto:ag at ag-projects.com>> wrote:
> All these softwares are mature with many years in service both for
> the media relays and the SIP part. They deal find with most of the
> expected failures, which is what the customers expect. For the
> un-expected failures, well the sky if the limit for optimising
> with infinite cost/benefit ratio. I personally did not hear my
> customers asking for any more resilience or scalability for the
> media relay component, so I stopped optimising long time ago.
> A better question is where would OpenSIPS project go next, beyond
> optimisations, as the outside world does not stay still and the
> perception of some of my customers is that we are being left
> behind feature-wise.
> On 13 Jun 2014, at 14:18, Bogdan-Andrei Iancu <bogdan at opensips.org
> <mailto:bogdan at opensips.org>> wrote:
>> Hi Maxim,
>> It is good to know about the rtp_cluster, but aside simplifying
>> things, it does not bring any new functionality - the LB and
>> failover between RTPproxy nodes can be done now in OpenSIPS module .
>> The most challenging thing we are looking at is the ability to
>> move calls between different instances of RTPP (for HA
>> purposes)..or some restart persistence for the sessions - without
>> something like that it's very hard to deal with SW/HW failures ;
>> there are ways to go around for scheduled stops/restarts
>> (maintenance), but non for unexpected failures.
>> Thanks and Regards,
>> Bogdan-Andrei Iancu
>> OpenSIPS Founder and Developer
>> http://www.opensips-solutions.com <http://www.opensips-solutions.com/>
>> On 13.06.2014 00:36, Maxim Sobolev wrote:
>>> Brett, on the HA/carrier-grade side there is little-advertized
>>> middle-layer component called "rtp_cluster", which in essence is
>>> load-balancing, transparent dispatcher that can be inserted in
>>> between some call-controlling component like OpenSIPS or Sippy
>>> B2BUA and bunch of RTPP instances running on the same or
>>> multiple nodes. From the point of view of that OpenSIPS it's
>>> just another RTPP instance.
>>> And it handles all logic necessary to load-balance incoming
>>> requests between online instances plus it can handle dynamic
>>> re-confiduration of the cluster and track individual nodes going
>>> up and down. The code is pretty usable, we have it deployed for
>>> several customers and it's being actively developed as well. We
>>> have it working reliably controlling up to 30-40 RTPP instances
>>> scattered over at least 5 nodes.
>>> We have at least one pretty well known service provider whose
>>> name starts with capital V using it in combination with OpenSIPS
>>> to load balance RTP traffic via bunch of Amazon EC2 instances.
>>> On Tue, May 27, 2014 at 6:52 AM, Brett Nemeroff
>>> <brett at nemeroff.com <mailto:brett at nemeroff.com>> wrote:
>>> Just wanted to add my 0.02 here..
>>> I totally agree with Bogdan. For the applications where
>>> opensips + a RTP relay make sense, HA and persistence are
>>> much more important.
>>> WebRTC and ICE are kinda applications in of themselves. And
>>> although these applications are going to grow in popularity,
>>> the "legacy" needs for an RTP relay are still massively
>>> prevalent in the space. A general push towards "Carrier
>>> Grade", resiliency and redundancy I think is much better for
>>> the project as a whole.
>>> Not only that, consider that applications requiring ICE or
>>> WebRTC will greatly benefit from HA / persistence, but not
>>> so much the other way around :)
>>> On Sun, May 25, 2014 at 6:30 AM, Bogdan-Andrei Iancu
>>> <bogdan at opensips.org <mailto:bogdan at opensips.org>> wrote:
>>> As always, the truth is in the middle.
>>> I agree RTPP is behind on certain things (and this is
>>> why we want to do them), but on the other hand it is a
>>> good platform with other good features (missing on the
>>> other relays). RTPP has better ability in individually
>>> controlling the stream (audio /video), ability to set
>>> timeouts and onhold with no conflicts, ability to
>>> generates events on timeout, more flexibility in
>>> handling symmetric / asymmetric NATs, ability to do
>>> media injection (playback), ability to do call recording
>>> What neither mediaproxy, nor rtpengine have is a
>>> mechanism for implementing RTP failover (for ongoing
>>> calls) or restart persistence . This is something we
>>> want to look into. I would love to have ICE and WebRTC
>>> on my media relay, for the HA and persistence are more
>>> important I would say.
>>> Bogdan-Andrei Iancu
>>> OpenSIPS Founder and Developer
>>> http://www.opensips-solutions.com <http://www.opensips-solutions.com/>
>>> On 24.05.2014 01 <tel:24.05.2014%2001>:59, Muhammad
>>> Shahzad Shafi wrote:
>>>> To be honest, i have stopped using rtpproxy for over 2
>>>> years now. It is not evolving as fast as it should be,
>>>> specially in the context of ICE and WebRTC technologies.
>>>> I would like to suggest that opensips team should
>>>> consider adding support for rtpengine from SIPWise,
>>>> For now mediaproxy from AG Projects is the only good
>>>> choice for handling media in opensips with ICE support
>>>> (though it still lacks WebRTC features).
>>>> Thank you.
>>>> On 2014-05-23 14:55, Bogdan-Andrei Iancu wrote:
>>>>> Going for a public exposure on this question to Maxim,
>>>>> maybe we will get an answer here.
>>>>> -------- Original Message --------
>>>>> Subject: RTPproxy project
>>>>> Date: Mon, 14 Apr 2014 15:03:31 +0300
>>>>> From: Bogdan-Andrei Iancu
>>>>> To: Maxim Sobolev
>>>>> CC: Razvan Crainea
>>>>> Hello Maxim,
>>>>> Long time, no talks, but I hope everything is fine on your side.
>>>>> I'm reaching you in order to ask about your future plans in regards to
>>>>> the rtpproxy project? We see no much activity around it and other media
>>>>> relays are popping around.
>>>>> RTPP is an essential component for us, we invested a lot of work, we
>>>>> have many patches (extensions) for it (which we want to push to the
>>>>> public tree, but there is no answer on this) and we are also looking for
>>>>> investing a lot into big future plans (as adding more functionalities).
>>>>> Now, my question is - what is your commitment and disponibility for the
>>>>> RTPP project ? depending on that we what to re-position ourselves, as we
>>>>> do not want to waste time and work on things which are out of control.
>>>>> Best regards,
>>>>> Bogdan-Andrei Iancu
>>>>> OpenSIPS Founder and Developer
>>>>> http://www.opensips-solutions.com <http://www.opensips-solutions.com/>
>>>> Mit freundlichen Grüßen
>>>> Muhammad Shahzad
>>>> CISCO Rich Media Communication Specialist (CRMCS)
>>>> CISCO Certified Network Associate (CCNA)
>>>> Cell:+49 176 99 83 10 85 <tel:%2B49%20176%2099%2083%2010%2085>
>>>> MSN:shari_786pk at hotmail.com <mailto:shari_786pk at hotmail.com>
>>>> Email:shaheryarkh at googlemail.com <mailto:shaheryarkh at googlemail.com>
>>>> Users mailing list
>>>> Users at lists.opensips.org <mailto:Users at lists.opensips.org>
>>> Users mailing list
>>> Users at lists.opensips.org <mailto:Users at lists.opensips.org>
>>> Devel mailing list
>>> Devel at lists.opensips.org <mailto:Devel at lists.opensips.org>
>>> Maksym Sobolyev
>>> Sippy Software, Inc.
>>> Internet Telephony (VoIP) Experts
>>> Tel (Canada): +1-778-783-0474 <tel:%2B1-778-783-0474>
>>> Tel (Toll-Free): +1-855-747-7779 <tel:%2B1-855-747-7779>
>>> Fax: +1-866-857-6942 <tel:%2B1-866-857-6942>
>>> Web: http://www.sippysoft.com <http://www.sippysoft.com/>
>>> MSN: sales at sippysoft.com <mailto:sales at sippysoft.com>
>>> Skype: SippySoft
>> Devel mailing list
>> Devel at lists.opensips.org <mailto:Devel at lists.opensips.org>
> Users mailing list
> Users at lists.opensips.org <mailto:Users at lists.opensips.org>
> Users mailing list
> Users at lists.opensips.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Users