Linux 3.15 Can Resume From Suspend 7~12x Faster

[tmpdir]

Sata is missed from the benefit?

No its not. Its even mentioned in the test case.

[droste]

The original Intel blog post had both ATA and SCSI patches, but the patch that was merged into 3.15 was only the SCSI one.
What happened to the ATA patch? And who has SCSI devices in their laptops anyway?

IIRC almost all SATA, PATA, USB and Firewire(?) devices "talk" SCSI in the linux kernel.

[BSDude]

IIRC almost all SATA, PATA, USB and Firewire(?) devices "talk" SCSI in the linux kernel.

I actually wanted to ask/say the same thing. I know this is how it works in FreeBSD. Not entirely sure about Linux.

[liam]

It's suspicious that all of the benchmarks are on systems with HDDs. I suspect that there is far less benefit for machines with SSDs.

I'm not sure it's suspicious so much as being succinct. I suspect, though don't know, that ssd machines resume much faster as their firmware is probably designed to suspend/resume the drive as fast as possible. Obviously they are also faster when it comes to reads/writes but given that these should be sequential the difference should only be, roughly, 3-8 times. Regardless, he should've at least addressed this point, I agree.
Something I wish at least some OEMS would start including is a dedicated ssd attached to the pcie bus and sized slightly larger than ram to be used for hibernation. Ideally it would become platform independent by being addressed by acpi(or is uefi now more appropriate?).

[Luke_Wolf]

I'm not sure it's suspicious so much as being succinct. I suspect, though don't know, that ssd machines resume much faster as their firmware is probably designed to suspend/resume the drive as fast as possible. Obviously they are also faster when it comes to reads/writes but given that these should be sequential the difference should only be, roughly, 3-8 times. Regardless, he should've at least addressed this point, I agree.
Something I wish at least some OEMS would start including is a dedicated ssd attached to the pcie bus and sized slightly larger than ram to be used for hibernation. Ideally it would become platform independent by being addressed by acpi(or is uefi now more appropriate?).

If you're going to do that it should be the size of the upper limit of RAM that the system can handle that way you don't break hibernation when the user decides to upgrade his RAM, which will also help with the longevity of the device by giving it more room to do wear leveling.

[liam]

If you're going to do that it should be the size of the upper limit of RAM that the system can handle that way you don't break hibernation when the user decides to upgrade his RAM, which will also help with the longevity of the device by giving it more room to do wear leveling.

I did say "sized slightly larger than ram"😀
The devices i'm thinking where this makes the most sense are ultrabooks where RAM is, generally, not upgradable and battery life is so very important (along with general speediness).
For more general purpose machines what you say makes sense but would substantially add to the cost since 32GB is becoming the norm for new laptops. That's a pretty decent sized sad, and would be wasted if the user only had, say, 8GB installed. Add for wear leveling i'm not sure that's an issue given even 2bit/cell nand ~10000 cycles. So hibernating 3 times per day gives you ~10 years of life (you'd still need over provisioning, of course).
Having read the post it does seem to be the case that ssd firmware is far quicker than hdd. On the order of > 10X, our roughly, .5s. A special purpose built ssd that only needed what's provided by uefi, and has no need of an advanced controller since it still only ever perform sequential operations should be both cheaper (eventually), use less power (if only slightly), and be faster than a general purpose ssd.

[mendieta]

I wonder if chromebooks will finally be upgraded to a newer kernel. My son's Acer c710 is still running 3.4, which forces you to use crouton with an older ubuntu image, sigh

[chinoto]

This may explain why when I resume my computer, it is unresponsive (just a second) until my harddrive spins up. I guess it'll be kinda neat for it to be immediately responsive.