PERC H310 - LSI 9211-8i - $50
- Thread starter Lance Joseph
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
question for you guys who know better, is the H310 flashed to IT mode work just as well as an LSI 9211-8i? I read on some FreeNAS forum that the H310 was sort of crippled in that its command queue depth is very shallow versus other HBAs. I was thinking of using a few H310 in IT mode for a ZFS storage server, but wondering if that's not the best choice here relative to getting another LSI 9211-8i based card?
I'm sure I read somewhere that the H310 natively was set to 25 with Dell firmware, but was capable of a queue depth of something like 600 when running the LSI firmware? Maybe I'm wrong...
In ESXi I think you can ssh in and run esxtop, d then f to bring up QSTATS and look down the AQLEN column for your adapter
I flashed 3 H310 controllers for use with ESXi and vSAN. I used this guide to get them on LSI firmware with the correct queue depth.
How-to Flash Dell Perc H310 with IT Firmware To Change Queue Depth from 25 to 600 - ESX Virtualization
How-to Flash Dell Perc H310 with IT Firmware To Change Queue Depth from 25 to 600 - ESX Virtualization
Hmm...I know that I posted this earlier but it's disappeared, or it's ended up in another thread...
Anyway, if you still want one, these are not bad for the money: Refurbished Dell HV52W PERC H310 8-Port SAS+SATA RAID Controller w/ (1) Cable | eBay
Anyway, if you still want one, these are not bad for the money: Refurbished Dell HV52W PERC H310 8-Port SAS+SATA RAID Controller w/ (1) Cable | eBay
I grabbed one from this seller for $37 shipped
Dell 3P0R3 PERC H310 6gbps PCI-E raid controller | eBay
Should be here in the next few days
Dell 3P0R3 PERC H310 6gbps PCI-E raid controller | eBay
Should be here in the next few days
so, once these are flashed to IT mode, and the queue depth problem is gone, am I going to be able to get full bandwidth from 8x SATA 6Gbps drives? for example, if I strip them in ZFS (no parity) or mdadm raid-0, am I going to see >800MBytes/sec sequential reads?
I see no reason why not, assuming that the rest of your hardware is capable of dealing with the data
is that based on experience or a guess?I see no reason why not, assuming that the rest of your hardware is capable of dealing with the data
Experience, but there are loads of things that come into play here, if you want to actually make use of that performance, not least of which might be number of disks in the array, disk access times, latency, read cache, stripe size, workload and etc. For performance, you need low latency and really fast access times. For that you'll be wanting good SSD's rather than magnetic media.
That being said, the rest of the hardware and software, also has to be able to keep up, otherwise you'll be playing the "push the bottleneck about" game
That being said, the rest of the hardware and software, also has to be able to keep up, otherwise you'll be playing the "push the bottleneck about" game
Experience with a slower configuration (RAIDZ1 of 5 * 8TB) on a 9201-16i. Those systems replicated over the LAN and got about 700MB/sec. Given that the graph shows a transfer of some 16TB, this is purely disk-limited as there's no way to cache all those writes on a SSD - eventually things will "bog down" as the SSD fills and is flushed to spinning rust. The downward trend is due to slower data transfers as you move toward the middle of the drives from the outside.
