4

I'm using following psping command to measure the bandwith throughput for localhost on a physical server

psping -4 -b -l 8k -n 20000 localhost:1234

The throughput is around 150MB/s wich isn't all that bad but for an ipc over tcp problem we are having with an ODBMS running on the server, we decided to try the application on a standalone pc.

The performance on that standalone pc is far better than that of our server. It's throughput to localhost gets up to 800MB/s what I believe to account for most of the performance difference.

Does anyone know how to improve the localhost throughput on the server?

Server details

  • Microsoft Windows Server 2008 R2 Standard
  • Service Pack 1
  • AMD64 Familiy 15 Model 8 Stepping 0
  • 2 Six-Core AMD Opteron(tm) Processors 2427
  • 32 GB ram

Standalone PC

  • Windows 7
  • Intel(R) Xeon(R) CPU E3-1240 v3 @ 3.40GHz [Family 6 Model 60 Stepping 3]
  • 16 GB ram
  • Not so much an answer as a suggested pointer... Compare the advanced settings on the network adapters for the two machines and then compare those values to your network gear. For that kind of wild difference, I'd expect something to be just plain wrong there... – Hunter Eidson Jan 02 '15 at 02:38

1 Answers1

0

You are comparing a 6-year old server to a 3 year old PC with a much faster clocking.

If you can tune your DB so it will use all cores you will propably solve your local performance problem.

I do not think you have an IPC issue.

Nils
  • 7,657
  • 3
  • 31
  • 71
  • Looking at WPA traces, CPU performance does seem to play a role but I'm having a hard time believing it would account for the 4 to 8 time faster localhost processing. The waits on the server are a factor 8 longer than on the standalone pc. If you happen to be well versed with WPA traces *(I am not)*, I would be happy to provide you a *(private)* link to both. The frustrating part is that the answer is in those traces but I'm not knowledgeable enough to draw conclusive conclusions. – Lieven Keersmaekers Jan 04 '15 at 20:49