<div dir="ltr">Bogdan, with regards to the media relays clustering what's the advantage of sharing that load info between each signalling node versus having each node tracking this independently? In my view the latter could be more reliable and much less complicated construct. The only disadvantage is that you'd get more command load on relays, however at least with the RTPproxy pulling load stats is very lightweight operation so even with tens of signalling nodes pulling single media relay every 1-10 seconds it won't cause any noticeable performance degradation on the relay. On the flip side, each signalling node would get accurate view from its vantage PoV, so in the case of geographically distributed system when signalling node can only see subset of all media nodes it would still be able to make proper decisions. This is the approach we use in the rtp_cluster and it works pretty well with cluster size of up to 5 signalling and 10 RTP handling nodes, 40-50K media sessions in total. It can also give you accurate RTT information, so your signalling node can not only factor in the load but also proximity or each and every media relay.<div><br></div><div>As far as the load tracking is concerned, I think the approach to implement "b2b-driven routing" using API that is specific to each particular b2b is somewhat wasteful and is not very future-proof. What we would like to see instead, is for opensips to publish some kind of API (preferably SIP-based, using OPTIONS or SUBSCRIBE/NOTIFY mechanism) to pull this information out and let each b2b vendor to implement proper hooks. Then it can go as far as making this info some king of RFC.</div><div><br></div><div>Anyhow, just my $0.02c. Not volunteering to do opensips side (ENOTIME), but if opensips project comes up with the reasonable b2bua-agnostic load query API to use we might look at implementing it in the sippy [py/go]-B2BUAs.</div><div><br></div><div>-Max</div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Nov 8, 2017 at 9:31 AM, Bogdan-Andrei Iancu <span dir="ltr"><<a href="mailto:bogdan@opensips.org" target="_blank">bogdan@opensips.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000">
<tt>Hi Nuno,<br>
<br>
On the Asterisk part, the plan is to do exactly what we already
have for FreeSWITCH (see
<a class="m_3371668786977425150moz-txt-link-freetext" href="https://blog.opensips.org/2017/03/01/freeswitch-driven-routing-in-opensips-2-3/" target="_blank">https://blog.opensips.org/<wbr>2017/03/01/freeswitch-driven-<wbr>routing-in-opensips-2-3/</a>)<br>
<br>
In terms of clustering media relays, is about the ability to share
the state (enabled/disabled) between all the cluster nodes using
the media relays. Optionally, we are looking in adding the ability
to balance the traffic between the relays, in a cluster-level
aware (all the nodes in the cluster will share the information on
the load of the media relays )<br>
<br>
Regards,<br>
</tt><span class="">
<pre class="m_3371668786977425150moz-signature" cols="72">Bogdan-Andrei Iancu
OpenSIPS Founder and Developer
<a class="m_3371668786977425150moz-txt-link-freetext" href="http://www.opensips-solutions.com" target="_blank">http://www.opensips-solutions.<wbr>com</a>
</pre>
</span><div><div class="h5"><div class="m_3371668786977425150moz-cite-prefix"><br></div></div></div></div></blockquote></div>
</div></div>