Does anyone know how to adjust the staggered spin-up (SSU) delay on an IRST RAID?
I have 5 HDDs in a RAID 5 configuration and the default delay between drives seems to be about 3 seconds, which means total spin-up time is over 15 seconds. Most of my networked devices will time-out waiting for that. (I'm on v.184.108.40.2067 BIOS firmware running IRST v.220.127.116.112, since v.18.104.22.1682 drivers don't seem to support idle power management.)
As I noted, reverting to IRST v.22.214.171.1242 did work around the problem with the current drivers not supporting Link Power Management at all. So, the drives will spin down when idle.
However, the staggered spin-up takes so long that many networked devices will time out waiting for it. My last RAID system used a staggered delay of only 1 second, which served to reduced the combined power draw and yet was fast enough to avoid timeouts. Unfortunately, I can find nowhere to set this delay in the BIOS or IRST user interface, and I'm hoping that there may be a registry setting that I can just modify.
Are you saying that it's a bug? I didn't mean to suggest that.
I was just under the impression that someone decided to play it safe and made the default staggered spin-up delay much longer than it really needs to be. Usually, however, these settings aren't hard coded and we should be able to override with a registry setting. Alternatively, maybe we can just disable the staggered spin-up feature altogether. While a future IRST release may expose these settings in the UI, why can't we access it via the registry now?
You may need to wait for another community member that can help with registry values, that would be considered out of our scope of support.
So, there's a thread that's been active over a year on the bug with RST drivers after v14 not allowing HDDs to spin down in RAID mode. Frankly, I think it's pretty sad that Intel has taken so long to address that problem, but I think it indicates that no one there is taking these bug reports all that seriously.
It cannot be by design that it should take over a minute to spin up a 5-drive RAID. That's just unreasonable, and no other RAID hardware I've used has anywhere near that kind of a delay.
Another thing that I've noticed is that the RAID device (DEV_2822) doesn't seem to handle the spin-up in any coordinated way that brings the entire RAID online at once. So, as I go to access directories on the RAID volume, it will spin up just the drives needed to service the block data request and no others. This may seem like a reasonable design decision, but it means that users will soon be waiting for the next drive to spin up from scratch...and then the next...and then the next...and then the next...and then the next. This not only takes forever, but it means that IO is blocked five times instead of just once. Other RAID hardware doesn't do that.
Link Power Management should power all RAID drives down during inactivity together, and it should restore them together on the first IO request. Drives should be staggered to reduce startup power draw, but only by a second or two. Ideally, this would be configurable via registry setting, but I think the default should be only one second.