Build Help: ESXI Server

[Red Falcon]

I didn't see your post, john. I tend to prefer xfs. I've found that xfs seems to disable the write barrier when I am running it in a VM, claiming that the write barrier check failed (for whatever reason.) That said, the couple of times I tried playing with ext4 I gave up - the write performance was abysmal, and I wasn't doing anything customized - a straight up install.

I'll go ahead and try out XFS on my next build and see how it compares. I've heard a lot of good things about it.

I never cared for ReiserFS, and Ext2/3/4 were always very stable, but they were lacking in performance.

 

[Gregyski]

I don't know if LSI has similar settings or not.

It appears not. NCQ is only mentioned at the adapter level in the documentation. There are rather few settings for applying to the physical disks (disk write cache is one).

So I got tired of waiting for LSI to reply to the ticket I opened, so I called them. I spoke with one TSE who said I need to make sure the drive write cache is enabled, so I did that to no effect. The second TSE had absolutely no suggestions. He said I tried everything he could think of, though it was obvious he wasn't going out of his way to find a solution.

He did mention that they (LSI) only use iometer for their testing. I'm running a set of standard tests used by ESXi storage benchmarkers right now. (VMWare also only recommends iometer.) The results aren't horrible but I think it's because people post their averages of these real world scenarios but that also averages the reads and writes. It appears my reads are fine but writes are slow, so it's hard to make a good comparison with these results. I'm going to keep trying. I just wish I could find someone with a RAID5 SATA with iometer with whom I could compare some write-only results.

That said, I'm down to 10 days before anything I want to return to Newegg needs to be returned, so desperation is setting in. I need to evaluate my options. Unless these 11th hour benchmarks vindicate my setup, I don't see a happy ending in sight. But I need to determine the best course of action in this bad situation.

One thought is to keep the card, return the nearline SAS drives, and replace them with some equivalent SATA drives. The reason for this is that (a) perhaps the card doesn't work well with what it expects to be SAS drives but which, in fact, perform like SATA drives (it seems my configuration is uncommon in the wild so perhaps it's not well tested) and (b) it'll be much easier to do performance testing with SATA drives since I can use the MB for doing JBOD and software RAID comparison benchmarking. Of course, if this all turns out to be a bad card, then this is the wrong route and $90 in restocking fees down the drain plus a long wait trying to RMA the card to LSI.

Returning both the card and the drives is another option, but that'll run $165 in restocking fees plus leaves me unsure which card I should buy. There was an Areca that intrigued me, but my understanding is that ESXi's support for Areca, while official, may not be on par with LSI's. I'm also a little unsure about managing Areca cards as there's less documentation. Although the inclusion of a network management port seems excellent, if people have had good experiences working with it that way.