[Webinar] Streamline your web hosting managementRegister Today

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 1560
  • Last Modified:

EqualLogic iSCSI network not optimized

We have an EqualLogic 6000XV running in our production network. Two of the HBA/network ports are currently plugged in and active. While running some crude speed tests on the machines currently talking to the array, we see that traffic is doing some weird things. While copying large files from a Windows volume on the EqualLogic across the local network to another machine's internal drive, we see the two (MPIO) network adapters come up to about half speed, stay for a second, drop to almost nothing, then repeat. At absolute peak, speed seems to max out around 500Mbps. Everything is on 1Gb links.

So the iSCSI links appear to be doing these strange bursts, and then only to about half the max link speed. The iSCSI links go through a Dell PowerConnect 6220 with MTU turned up to 9216 and flow control on. The local network is running fine - when doing copies from internal disks to internal disks we see the LAN links max out all the way through.

I've checked out stats in the EqualLogic's Group Manager and there don't appear to be any errors on the network. The two unused links on the active controller are disabled.
0
j_h_2
Asked:
j_h_2
  • 7
  • 3
  • 3
2 Solutions
 
gheistCommented:
might be flow control
basically if buffer of one port is full it slows all sources.
i dont know much about qos you have in your switch. if it resembles traffic shaping like HFSC then do it.
0
 
j_h_2Author Commented:
Do you mean a lack of flow control? Flow control is on for every port.
0
 
gheistCommented:
presence of flow control is a problem.
you need to adjust shaping to make sure flow control never kicks in
0
Get expert help—faster!

Need expert help—fast? Use the Help Bell for personalized assistance getting answers to your important questions.

 
j_h_2Author Commented:
Hmm I actually just checked everything in the config again and though the switches are set to use flow control, the NICs in the attached server(s) are not.
0
 
gheistCommented:
no problem with iSCSI... fix your IP network....
0
 
Michael WorshamInfrastructure / Solutions ArchitectCommented:
Even though you stated the EqualLogic's Group Manager does not show any errors, but does the Dell PowerConnect 6220 show any CRC errors?

As for the hosts doing the data retrieving/storing, are these Linux or Windows servers? If the latter, which version -- 2003, 2008?
0
 
j_h_2Author Commented:
No errors on any ports on the 6220. We actually have two 6220's running and no errors on either.

It's Windows Server 2008 x64. SQL Server 2008 on the machine.
0
 
Michael WorshamInfrastructure / Solutions ArchitectCommented:
On the Windows Server 2008, do you have TOE enabled? The reason I ask is that under Windows 2003 and also seen under 2008, the TCP Offload Engine (TOE) part of the Scalable Networking Pack can at times cause network performance issues.

Post the output of this command: "netsh interface tcp show global"

0
 
j_h_2Author Commented:
Output:

C:\Windows\system32>netsh interface tcp show globalQuerying active state...TCP Global Parameters----------------------------------------------Receive-Side Scaling State          : enabledChimney Offload State               : automaticNetDMA State                        : enabledDirect Cache Acess (DCA)            : disabledReceive Window Auto-Tuning Level    : normalAdd-On Congestion Control Provider  : ctcpECN Capability                      : disabledRFC 1323 Timestamps                 : disabled
0
 
Michael WorshamInfrastructure / Solutions ArchitectCommented:
Try the following configuration and see if this helps any:

netsh interface tcp set global chimney=disabled
netsh interface tcp set global rss=disabled

Information about the TCP Chimney Offload, Receive Side Scaling, and Network Direct Memory Access features in Windows Server 2008
http://support.microsoft.com/kb/951037
0
 
j_h_2Author Commented:
Okay, this is actually a failover cluster we're on, so I'll make changes on the inactive box and fail it over tonight to see what we get. Thanks for the input and I'll report back in the morning.
0
 
j_h_2Author Commented:
We're still having an issue, but it looks like a different issue now. Closing this one.
0
 
j_h_2Author Commented:
Good, clear answers, but turns out that we needed to perform update drivers, update switch configs, do a few other things. Thanks for the responses!
0

Featured Post

The 14th Annual Expert Award Winners

The results are in! Meet the top members of our 2017 Expert Awards. Congratulations to all who qualified!

  • 7
  • 3
  • 3
Tackle projects and never again get stuck behind a technical roadblock.
Join Now