P410i Performance Change

[Steve Sparlin]

What I'm trying to figure out is why the most recent few ML350G6 servers we've built don't perform the same as ones we did a few months ago. Read on for a few details...

At my company when we sell a ML350G6, our standard config is the P410i array controller with a 256 MB cache module and RAID-5. Usually we fill them with SATA drives. We don't install a battery for the array cache, nor do we enable the write caching. Obviously, this is not an ideal setup for high performance but for our small office clients they fit the bill at a decent price.

The past 2 or 3 of these servers are MUCH slower with disk access than those we built a few months ago with the exact same hardware and configuration. Using the same firmware and drivers as the older servers has made no difference, so it doesn't seem to be an issue with firmwares or drivers.

If we enable write caching on these new ML350's they perform like the older ones but we aren't using write cache with the older ones. What has made the difference? I'd be interested to see if anyone has noticed the same problem. We've decided to just buy the batteries and enable write caching to solve this for now but I'm really curious as to why the same hardware performs so differently.

Some of these servers have Windows on them and some have Novell Netware. The most recent one has Netware on it. Without the write cache enabled, I can't copy 2 GB of data to it before the server and workstation both become unresponsive and/or I start to get file copy errors. In console monitor I see disk requests jump up to 1,000 almost instantly when I start to copy data. Within a few minutes my volume is no longer accessible at the workstation (it isn't dismounting) and the server is unresponsive. All that can be done is to power the server off.

Has anyone else noticed this same issue recently? Can someone point me to an advisory from HP I've missed? Calling HP support hasn't produced any solution - there doesn't seem to be any bad hardware in use here. The hardware just isn't performing right. We've elected to purchase the battery for the write cache and install those to solve the problem but I still want to know why...

[wobbe]

Strange story.

Quote: I can't copy 2 GB of data to it before the server and workstation both become unresponsive

This sounds like network issues to me.

Also afer you setup a raid five it can take a long time before all parity data has been calculated.

[Steve Sparlin]

I've verified good connectivity on the network while this happens. At first, I was convinced it had to be a network issue too. I lose access to data on the server, I don't lose connectivity to the server. I can still login, use remote console, etc. - all I seem to lose is access to the server's volumes.

I have also made sure to let the parity operations fully complete before starting the data copy to ensure this isn't causing problems too. I'm at a loss to explain this one so far.

Thank you for your input.

[Bleeder]

Steve, did you ever find a solution to this?

[Steve Sparlin]

Bleeder: No, I never figured out why the performance from the controller was different. We just started including a battery backed write cache module in all of our servers and enabled write cache on the controller.

That was all I could figure out to "solve" the issue.