Hi Jacob,

Unfortunately, I don't know anything about Optane as yet. Also, although
I've come across Intel's SRT (Smart Response Technology) I hadn't heard
of its RST (Rapid Storage Technology) until now. It seems like SRT is a
subset of RST. (The similarity of the naming seems like a bad decision
to me!)

My laptop came with a small SSD (32GB) and two Intel technologies to
make use of it. The SSD had two partitions, one large and one small. SRT
was used to make the larger partition operate as a cache for the main HD.

[For those who weren't at the meeting: this caching can also be done
with the Linux bcache driver. It gives you most of the performance of an
SSD but the capacity and greater reliability of an HD, because the SSD
is used just as a cache and the HD can still be used without the SSD.]

The other technology was Rapid Start Technology (also RST!! Why are they
so stuck on those three letters?) and was used to make the small
partition into a hibernation space. This is different from regular Linux
hibernation and is actually pretty cool. It can take the processor from
sleep state to hibernation without waking up the OS. It means you can
close the laptop lid and leave the machine like that indefinitely. You
don't have to worry about it running about of power and crashing if you
forget to plug the power in at some later point. The RST firmware writes
RAM and processor state to the SSD and shuts off the power. When you
open the lid later, the firmware restores the processor and RAM from the
SSD, and wakes the OS up as if it had been asleep all along. This is
ideal for me because my laptop is a secondary machine that I don't use
very often, and I often put it to sleep and then forget about it for a week.

Unfortunately, the SRT technology that uses the SSD as a cache is
Windows-only, so I ditched that in favour of the Linux bcache driver,
which I'd already used on a different machine with excellent results.
This does the same job, but in a more Linux-friendly and transparent
way. The SRT technology isn't well known, doesn't have much
documentation, and is entirely opaque. However, the Rapid-Start
Technology worked well with Linux, so I just used it and was glad of it.

>From the quick look I had at the new RST (Rapid Storage Technology) it
claims to be Linux-compatible. However, I'd be careful. I'd want to make
sure that the driver is in the kernel mainline, otherwise you're stuck
with not being able to use a stock kernel, or at least having to use
DKMS to build the driver from scratch every time you upgrade the kernel.
I've done that in the past with proprietary graphics drivers and found
it a pain. Also, if you ever need to troubleshoot it, you may not have
access to as much info and help as you would with a Linux-native technology.

I get the impression that the main thing the new RST adds is various
kinds of RAID capability, and the claim is higher performance,
especially with SSDs. I'd evaluate this pretty carefully, since the
existing software RAID capability in Linux is mature and
well-understood. Maybe RST would boost the performance slightly, but you
have to decide if it's worth the risk and the hassle of going outside
the main Linux ecosystem. I guess RST is more like an alternative to
hardware RAID, and most server people I've talked to say they prefer
hardware RAID, so maybe it's not an apples-to-apples comparison.

My approach with new technologies, and new hardware in general, is not
to push the envelope to squeeze the absolute maximum out of the
currently-available products. If you wait six months you'll be able to
get an even faster machine, so going right to the edge of what's
currently possible buys you only a short lead. The downside is higher
cost and greater risk. There's a sweet spot which represents the best
price-performance ratio, often called the 'knee' in the
price-performance curve. That's where I prefer to stay, personally. I'll
keep some of my funds for an earlier future upgrade rather than pour
them all into a single purchase now.

The exception is when a completely new generation of technology comes
along. It make sense to get into this new era rather than being stuck
with legacy technology for the next several years. However, I prefer to
let others shake out the bugs for a few months or years before getting
into it myself, so I don't buy it if it's not tried and tested yet.
Sometimes a new technology comes along that sounds promising but never
really catches on, and then you can be left with no support in a few
years time.

Hope this helps,

clug-talk mailing list
Mailing List Guidelines (http://clug.ca/ml_guidelines.php)
**Please remove these lines when replying

Reply via email to