Ping latency on all Virtual Machine

[Gamleur84]

Hi,

There is the setup :

ESX 4.0 Update 2

PowerEdge 2950

2 X CPU Quad Core E5430

8 GO

2 X Windows 2003 Virtual Machine on the physical server: File Server and Domain Controller

When I ping the domain controller and the file server, I have a reply under 1ms.

The issue is when I start a folder copy (800MB) from file server to my local computer, the reply from both server is around 20ms and the customer is complaining about slow network.

The virtual machine are on seperate network card : Domain Controller is on a vSwitch connected to a BroadCom 5708 and the File Server is on a vSwitch connected to Intel Pro 1000. VMKernel is on another Broadcom 5708.

I have a similar setup at my office with one ESX and 10 Virtual machine (SQL, Exchange...) and the same test give me a reply of 2-3ms when i'm transferring (not 20ms !)... this mean something wrong with customer server

Any idea ?

[AWo]

Welcome to the forums!

Have you installed the VMWare Tools inside the guests?

How many vCPU's are assigned to the guests?

Is the file server configured for file serving or for applications?

AWo

VCP 3 & 4

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

[Gamleur84]

Thanks for support

Domain Controller have 1vCPU

File Server have 4vCPU

File Server is only... Share Folder... No special application.

All guest have latest VMware tool from ESX 4.0 Update 2

Network Card issue with Power Edge 2950 ? Bios Option ? TOE ? TOS ?

I have the same issue with identical hardware at the same customer. I tried different Hardware Switch...

[AWo]

Reduce the vCPU count to 1 on the fileserver.

Best practices are that you start with as less vCPU's as possible. More vCPU's doen't necessarily mean more performance. The more vCPU's are assigned to a guest the longer it has to wait until it can schedule work. And if it is scheduling work all other systems have (maybe too) less ressources available.

You may need to change the guest server HAL if you use only one vCPU. Read KB 1003978 and this to check if that is needed: http://support.microsoft.com/kb/309283

AWo

VCP 3 & 4

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

Added the links, AWo

[Gamleur84]

Good point,

But In this case, I started with 1vCPU and I added 3vCPU for troubleshooting purpose...

[AWo]

However, go back to one vCPU:

Is the same virus protection software installed as in your office? Disable it for testing.

Are Jumbo Frames used or TCP Offload? If so change it , one by one.

AWo

VCP 3 & 4

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

[Gamleur84]

Symentec Endpoint Protection and i disabled it for testing with no result

How can I know if Jumbo Frame or TOE is enable or not ?

Thank you

[AWo]

Disables TCP Offload:

"Netsh int ip set chimney DISABLED"

Check: http://support.microsoft.com/?scid=kb%3Ben-us%3B912222&x=16&y=12

AWo

VCP 3 & 4

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

[Gamleur84]

Thanks,

it's a option in Windows and I already tried to disable from advanced option in Microsoft. Any command line to disable this option in ESX Hypervisor ?

[AWo]

There's no need to do that in the hypervisor.

AWo

VCP 3 & 4

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=