Overall Performance on P4300 SAS Starter

[JasonE]

Same problem here. Two P4300 G1's with SATA drives, two Procurve 2900 switches, no errors, just poor performance. This SAN should be able to saturate a 1Gb link easily. Using ALB and jumbo frames. Don't need HDTUNE or IOMETER to tell me what is obvious, that a single sequential file copy is topping out at 40-45MB/s...which is unacceptable. Moved a file and printer server to it, and originally planned to move an Exchange box and an SQL box over to it (Hyper-V), but I'm worried about the lack of bandwidth, even with Exchange 2010, which requires less. Guess I'll contact support (again), and try to get someone that knows something, and just doesn't try to pass the buck. If anyone knows of a P4300 installation that can saturate a 1Gb link properly, please reply. (just so we know it's possible!)

[Patrick Terlisten]

Hello,

when you're getting 965 IOs and every IO is 8 KB, then you will get round about 8 MB/s.

MB/s = IOPS x IO Size

You can fill a 1 GbE Link in two ways: Big IOs and less IOPS, or small IOs and many IOPS. IOPS is limited by the cache and number of disks.

Regards,
Patrick

[JacobS_4]

I understand the math behind IOPS x IO Size = Throughput. I'm not sure how that is 100% relevant to the slow throughput many of us are seeing though. I have 4 SAS nodes in a single cluster. That's 48, 15K RPM disks. Even at a conservative average of 150 IOPS per disk I should be able to get 7200 IOPS. I see an average IO size of 22K in my environment. That meens I should be able to push well over 1 Gbps, or 125 MBps. Obviously there is TCP/UP overhead, but I definitely should be able to get more the 40 MBps. My network and switch ports aren't saturated, I'm not seeing any errors across my network, and my host CPU levels are low. I definitely feel like this is a Lefthand bottleneck and it's very frustrating.

[Amana Living]

Sorry to revive a dead thread, but we are seeing the same value in performance. 30 - 45 MB/s. I have been tracking down a backup issue. Our backup server is only utilising 25% of 1Gbps.

I've gone through our vsphere config, and our blade Flex10 config and tested network throughputs with iperf and can max out all data links.

I then setup a basic FTP server and copied a 5 gig file between two VM's and can only get 30-45 MB/s transfer rate.

Since I know the data LAN's can support 2.5Gbps so that's not the bottle neck. The storage network is on a 4.8 Gbps NIC's and is part of the same Flex10 domain.

So far it is looking like the Lefthand cluster, I do not want to configure Multipathing at the moment as we will in the future be going to multi site SAN.

Any help is appreciated.

[NXS]

Anyone ideas on how to boost performance? with 6 P4500G2 sas (600 GB 15K) nodes with ESX4.1 ??

K

[Thomas Halwax]

I already postet this in an other thread but I was not able to set a link, sorry. So again:

We have a two node P4300 cluster (SAN iQ 9) with ALB on redundant HP 2910al switches. Clients are two DL380 G7 with ESXi 4.1.

Attached is a screenshot of a robocopy job and of the HP SAN Performance Monitor. As you can see we are able to reach 122 MByte/s (max. for a 1 Gbit/s link is 125 MByte/s).

Source of the robocopy job is a Win2003 server using the MS iSCSI initiator, target is Win2003 server on VMFS.

All volumes are Network RAID 10 (volumes mirrored).

Flow control, jumbo frames and Rapid Spanning Tree are enabled.

Of course this is no IO test but it shows that a P4300 cluster can operate at the max. throughput limit of a 1 Gbit/s link.

Thomas