7200/7400 Performance

[Patrick]

We've been looking at the Storserv 7200 & 7400 for a couple of weeks now but when checking online and through the local reps but I've been unable to see anyone deploy these units in production yet. Does anyone have an performance numbers on these units? Configurations would be useful, also if you have any comparisons from your now legacy arrays would be helpful.

[wkdpanda]

Deployed one of these for a customer a couple of weeks ago. Interesting that the "Service Processor" can now be a VM instead of physical box.

Overall, they are moving from a T-400 to a two node 7400. Performance has been good, easily exceeding the older T-400.

-Andrew

[NielsChr]

We have recently taken out new 7400 dual controler in use
Performance will ofcouse depend on the number and types of disks used.
we have 96x900Gb FC + 16x200Gb SSD

We come from a Netapp Fas2040 dual controller with aprox same number of disks (no SSD), and performance have gained significantly - about a factor 4-5 since our Netapp was not runing well.

here are a simple performance mesurement from sisa sandra, where I compare a revodrive (witch is considered as one of the fastes disk system)

[Patrick]

We have recently taken out new 7400 dual controler in use
Performance will ofcouse depend on the number and types of disks used.
we have 96x900Gb FC + 16x200Gb SSD

We come from a Netapp Fas2040 dual controller with aprox same number of disks (no SSD), and performance have gained significantly - about a factor 4-5 since our Netapp was not runing well.

here are a simple performance mesurement from sisa sandra, where I compare a revodrive (witch is considered as one of the fastes disk system)

What kind of workload are you throwing at it? FC or iSCSI?

[Arkturas]

It also depends on the number & type of disks (SSD/FC - fast class, NL SAS) + Raid type.
One thing I wasn't aware of is that you cant maximize node cache usage unless you fully populate all the cages. ie cache is allocated on a per disk basis.

[NielsChr]

Sorry for the late reply (I wasn’t watching this topic - I do now)

We did switch from Netapp (NFS) to 3PAR (FC) and I can still see the huge difference in performance and scalability. Netapp+NFS sucks while 3PAR+FC rocks.

With Netapp there was always issue about CPU usage on the controllers (usually going between 50-80% CPU Util), issues about bandwidth on the Ethernet etc., 3PAR have a very low CPU util on the controllers (usually in my case less than 5% util) and there is always bandwidth on the FC and low latency. (As a side note: we do have NFS separated VLAN, dedicated NIC's and Qos when using NFS - so theoretically is should have been alright, but not to my experience.)

My workload is a mix of VMWARE and HYPERV - primary VMware - I have a mix of approx. 120 MS Servers running on 10 physical hosts, webfrontend, applications, SQL Exchange etc. - All hosts have now redundant 8 GB FC to the 3PAR. seems to me like a good solution and I have newer regretted going away from Ethernet for SAN usage.

[Richard Siemers]

We did switch from Netapp (NFS) to 3PAR (FC) and I can still see the huge difference in performance and scalability. Netapp+NFS sucks while 3PAR+FC rocks.

Its good to hear you say that... I've always supported FC deployments and never seen a good reason to convert any of them to iSCSI, or as you mentioned to NFS with regards to ESX. There are several storage vendors out there that completely lack FC connectivity options and will spin stories about how ethernet is better than FC... its gets pretty evangelical sometimes.

[nsnidanko]

Please see attached file. This reports shows what you get from 108 10K SAS disks on a busy 7400 (2 controllers). It demonstrates load on physical disks after cache, etc. We are happy with performance.