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 have W2008 R2 server on colocation. I am getting weird cyclical lags - several times a minute latency jumps and then gradually drops back.

I can tell becuase the service I consume sends me local timestamps. The remote server synced with Stratum 1 with atomic clock and very precise.

I synced with local Stratum 2 and normally bias between my local time and the remote server is +-2ms. And it was for 2 months.

Now I am getting those cyclical latency jumps.

I told that to the colo support and they said they see nothing. They told me to use WinMTR and send them stats.

I pinged the IP address which is immediate hop to my box and it shows jumps in ping times. I.e. up to 450 ms.

The colo support said something that ICMP traffic is deprioritised on their routers and that is expected...

I cant believe this.

Surely many of you were in the situation like this. I just have no experience and I am not system admin anyway.

What would you guys do to make the colo admins to do something about it? What should I ask or what can I do on my side to prove the point?

P.S. the remote server is firewalled and I cant ping it to provide MTR to the end-point.

share|improve this question
    
WinMTR will actually tracert from you to the end and display where the bottleneck is hanging u –  Prix Aug 22 '10 at 19:03
    
In my case it doesnt work that way. I see "No response from host" for all IPs between me and the last address. However I see them in usual Windows tracert and they are pingable. –  Boppity Bop Aug 22 '10 at 19:23

2 Answers 2

up vote 0 down vote accepted

"The colo support said something that ICMP traffic is deprioritised on their routers and that is expected...

I cant believe this. "

I used to tell people posting here that pinging across the internet was not reliable because ICMP was low priority. After getting down-voted a couple of times I stopped. It is commonplace for ICMP traffic to have low priority.

It is not at all clear to me what the configuration is that you are having difficulty with.

My guess is that it starts off something like this:

W2008_R2_server ------ Switch ------ Router ------ ??? ------ end_user

share|improve this answer
    
?? - two more routers... however I have 500ms jumps on the "Router" already. –  Boppity Bop Aug 23 '10 at 6:35
    
In terms of the server / end user the problem is that the service is slow, or not working, or fails sometimes, or other? –  dbasnett Aug 23 '10 at 11:54
    
server - user: is that data coming from end-user (which is server rather) is lagging. Apparently that is faulty NNTP server in the colo environment which just went amok after 2 months of stable +/- 2ms error. Now it is 250ms.... Network is fine.. –  Boppity Bop Aug 23 '10 at 19:58

You might consider setting up response measurements from the co-located server to "near" (within the co-lo) and "far" (outside/beyond the co-lo) samplings too. That will provide more information as to what may actually occurring within the co-lo and/or its connectivity.

share|improve this answer
    
what are "response measurements"? are you talking WinMTR stats? –  Boppity Bop Aug 22 '10 at 19:39
    
Yes or any other tool that you choose to record the network latency performance over time. –  user48838 Aug 22 '10 at 22:32

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.