0

I'm in a confusing situation right now: I have 2 MS Server 2008, hosting a WCF-Webservice which is basically reading data from an SQL-Server and forwarding it to the client.

We use NLB to loadbalance those two applicationservers.

During Loadtests there is now a strange reading in the performance-logs: the CPU and MEM Load is quite even throughout those server, but only one server is building up write-queues on the HD, the other isn't.

Is there any relation between those readings and the NLB?

(The servers are only hosting this application: nothing else and are configured the same way)

MADMap
  • 133
  • 5
  • Same servers? Same discs? Same configuration? Can it be one disc subsystem is faster, thus not queueing up, while the other is overloaded? ;) – TomTom Aug 29 '11 at 14:41
  • This was an idea: but the servers were newly bought to do only this job: and they are identical in any way. – MADMap Aug 29 '11 at 14:51

1 Answers1

0

It really depends on what the WCF application is doing to know if those counters will matter. You need to work with your dev team to see if it is a problem.

mrdenny
  • 27,174
  • 4
  • 41
  • 69
  • I'm also a dev and have insights in the service: there is no writing to the disk from the service itself (except logging which was deactivated for the loadtest). So I can exclude that. But basically my question was wheter NLB has to write something on the HD to perform? – MADMap Aug 29 '11 at 14:45
  • It has not. NLB does no writing. – TomTom Aug 30 '11 at 03:44
  • NLB doesn't touch the disk. – mrdenny Aug 31 '11 at 00:25