I've done my own research in the UccAPI logs as well as opening a ticket with Office 365 support but the results are still muddled and confusing. I have yet to do full on packet captures to really nail it down...simply for lack of time.
My question is: does the Lync client really talk straight peer-to-peer for its IM and video call communications when not in a "Lync Meeting"? (Note: for me it's Lync Online/Office 365, but it should function very similar to an on-premise deployment)
Massimo's question here: How can two Lync client talk when there is no route between them? states:
Now, as everyone knows, Lync clients are only managed by the Lync servers, but talk directly between themselves for actual communications; unless of course a conference is involved, in which case they connect all to the front-end servers which acts as a MCU.
However, after working with MS support on a desktop sharing issue within Lync, the tech stated that the Lync server is always the MCU, even during basic IM messages. He said that even though 2 clients may be on the same LAN they will still relay their traffic through the Lync Online server, in essence acting as the proxy for the call/IM.
Looking at the UCCAPI.log on my client I see a lot of activity back and forth to the Lync server itself, but never really direct peer-to-peer traffic:
From: "Bob Smith"<sip:bob@foo.com>;tag=e7833a30;epid=1275117fecd
To: <sip:mary@foo.com>;epid=1d2a1221b4;tag=173235356c1
Call-ID: 07ec7f7a031111111111121100
CSeq: 1 INVITE
Record-Route: <sip:sippoolbl20a04.infra.lync.com:443;transport=tls;ms-fe=BLES03.infra.lync.com;opaque=state:Ci.Rfea2c00:Ti.W85ERLQ6ZEmmt2GgmW1EdwAA;lr;ms-route-sig=cf2qT3Y_wo-oN53Ui0Fro_lRokEb9YLApu_iBPn0fgV_0m71111AAA>;ms-rrsig=cfZAEgWED1u9111111RclA3N488uz7fXX7CSIQATO_8m7DA1rdtBTAAA;tag=F4A2EFB01111111111114FB97F116
Record-Route: <sip:sippoolbl20a04.infra.lync.com:5061;transport=tls;ms-fe=BL20AS04.infra.lync.com;opaque=state:F:Ti.W85ERLQ6ZEmmt2Ggm1111wAA;lr>;tag=4A2AA729141D0111177C6EA4C8AC6E7
Record-Route: <sip:sippoolbl20a04.infra.lync.com:5061;transport=tls;ms-fe=BL20AS02.infra.lync.com;opaque=state:T:Ti.W85ERLQ6ZEmmt2Ggm1111wAA;lr>;tag=0C9E9993ECA011111111893C255B4855
User-Agent: UCCAPI/15.0.4517.1004 OC/15.0.4517.1004 (Microsoft Lync)
Proxy-Authorization: TLS-DSK qop="auth", realm="SIP Communications Service", opaque="C09B2801", targetname="BL2003.infra.lync.com", crand="f2258f89", cnum="684", response="9b515b67187e3c1111ca712d797771e449df"
Content-Length: 0
So, am I interpreting the logs and what the MS tech said incorrectly or not?
The reason I ask the question in bold is because we are trying to decipher how much Lync client traffic is "lan client to lan client" vs. "lan->wan->O365 Lync server->wan->lan". I want to determine if the connections are routing over the fastest links available to them for Lync client communications.