Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I 'm designing a distributed system and I deploy it on 1Gb cluster of Windows Server 2008 R2 But I have issue with the latency, the normal Ping takes 0.270ms and I believe this is too much, since all the machines connected locally via one 1Gbs switch.

So how I can optimize the latency in general, which will help me in .NET WCF also

share|improve this question
    
what sort of NICs and switch are you using? –  Chopper3 Jun 11 '11 at 11:33

3 Answers 3

the normal Ping takes 0.270ms and I believe this is too much

Ah, no. Sorry. It is quite ok.

need very low latency? Use technology designed for that. Infiniband.

Tere is latency in 1gb networks. 0.2ms is less than 1ms. It is ok for 99% of the applications.

So how I can optimize the latency in general, which will help me in .NET WCF also

No, it will not. At the end., a proper designed WCF based cluster system will handle decent latency without problems. Yes, indicidual packes will take longer than zero time, but on the other hand proper message protocol design and work packet granularity in a cluster means that this does not mean machines do nothing.

I do a lot of distributed systems design / operations and seriously, 0.2ms is not somthing I would orry about, unless designing a HFT system, and then using WCF in itself is a stupid move - HFT counds latency in Microseconds tehse days (note: HFT = High Frequency Trading).

share|improve this answer

There are two suspicious places to check:

  1. the host(s)
  2. the switch

It is possible that your NICs are running with a reduced rate (maybe due to some power saving state) - checking your switch's management interface will provide more information about actual link speeds. Or your host's CPU on a whole is running with reduced speed (e.g. due to SpeedStep) or falling into deep sleep and needing some time to wake up. Deactivating the power management voodoo would eliminate that as a possible cause. If your hosts are virtualized, it would be rather normal to see varying/increased latencies due to virtualization scheduling.

If your switch is either very dumb or mis-configured, it would use the store-and-forward forwarding scheme instead of cut-through. Changing that would also reduce latency somewhat.

share|improve this answer

Try connecting 2 machines with a crossover cable and check the ping results. This will verify that you have an OS issue rather than a switching issue. I suspect you have a switch issue rather than an OS issue. The OOB settings for windows server are pretty good.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.