When it comes to network readiness for Lync, media bandwidth isn’t everything network latency is just as important and often forgotten when it come to voice.
With Lync deployments it is being coming more and more common that they will employ SIP trunk whether centrally, regionally or in a distributed manor so latency related more importantly to distance between two locations is a massive factor when Lync is deployed in a regional topology with SIP trunks.
Network Latency equates to 8.2 microseconds per mile, or 0.82 milliseconds per 100 miles. therefore for 200 miles it would be 1.64ms; 2000 miles it is 16.4ms or 4000 miles 32.8m, but that it only one direction and not roundtrip. 4000 miles round trip latency is 65.6ms. But that doesn’t take into account relaying (on each and every router/switch) on a hardware switch will be in the 4-to-20 microsecond range but in practice it should be 25 microseconds per hop. And then serialization latency for framing (won’t bore you here with that) and queuing.
Anywhere from 5 to 20 hops can be a reasonable assumption and you can easily assumed 1ms per hop after all the factor are included. Plus remember to case this twice for roundtrip again.
The Lync Documentation states for network requirements
“Provision your network to ensure a maximum end-to-end delay (latency) of 150 milliseconds (ms) under peak load. Latency is the one network impairment that Lync Server media components cannot reduce, and it is important to find and eliminate the weak points.”
Although this is one direction, it is under peak load, generally speaking less than 200ms round trip time (RTT) latency is good, greater than 200ms is poor and over 500ms is bad. But for planning I was told that ≤ 60ms one-way latency was optimal and ≤ 100ms is acceptable. So plan for 120-150ms roundtrip. (I think that is right, I’m sure someone will correct me)
Here are some example distances for you (and checking to see if people are reading my blog):
- Singapore → Sydney = 3,918 miles
- New York → Rio De Janeiro = 4,820 miles
- Calgary → Buenos Aires = 6,809 miles
- Miami → San Jose = 3,069 miles
- London → Johannesburg = 5,634 miles
- London → Dubai = 3,400 miles
- Dubai → Singapore = 3,626 miles
- Perth → Auckland = 3,322 miles
- Anchorage → Houston = 3,271 miles
- Vancouver → Montreal = 2,287 miles
- Reykjavik → Redmond = 3,605 miles
- The Hague (Netherlands) → Las Vegas = 5,337 miles
- Kuusankoski (Finland) → London 1,203 miles
So if we take our the magic 4000 miles with 10 router hops (20 roundtrip) this is already up near 100ms and less not forgot beyond the on-premise SBC and the ITSP’s SIP cloud and their exit on-to the PSTN.
These are all very important as there are a few things we can’t change one of them being the laws of physics and the fact that Light travels is generally fixed when travelling through either a optical or cooper cable as well as where country are in the world.
Design Best Practices for Latency Optimization
Network Infrastructure Requirements