HDD RAID 1 + SSD cache?

r00tb33r

Distinguished
Jan 21, 2011
143
2
18,715
Hey all,

I'm building a couple machines for family members and I'm a little behind the times on SSDs. I do have a constraint for my build, the storage must have hardware storage redundancy, RAID 1 to be specific. However, I recognize the benefits in latency/access time that SSDs bring. A 2TB SSD RAID would be cost-prohibitive but at the same time, a purely HDD storage would be somewhat of a bottleneck for a new machine.

So the question, is it possible to utilize SSD cache with a RAID 1 HDD volume?

Components I have are:
i7-7700 (OEM variety)
16GB DDR4 (2x 8GB DDR4-2400, Crucial value RAM)
ASRock Z370 Pro4 LGA1151 mobo (can return, if I must)
2x 2TB Seagate SATA III drives, 7200RPM, 64MB cache

I wanted to get 32GB Intel Optane modules and use RST Intel driver for caching, but I found this:
Intel Optane FAQ
I want to set RAID up on my system. Can I accelerate these RAID volumes with Intel Optane memory?
Intel Optane memory can't accelerate a RAID volume. An Intel Optane memory volume can reside in the same system as a RAID volume.
Note In a system with this configuration, we recommend using the Intel RST application to manage your RAID volume and Intel Optane memory volume.
So that sounded somewhat discouraging. I don't want a separate storage volume on SSD, I want the SSD (by means of a driver) to be seamlessly caching frequent storage accesses while taking advantage of the data retention safety of HDDs in RAID 1 mirror.

Can someone advise me how to materialize the setup I need?

Insight is appreciated. Thanks!

 
If I understand correctly what you want to accomplish would be similar to multiple hybrid drives that would be mirrored.

Bunch of those out there like here a tnewegg: https://www.newegg.com/Product/ProductList.aspx?Description=Hybrid%20Drive&Submit=ENE

As is kinda pointed out, when drives are in a PC (regardless of types) they are seen as physical devices. There aren't any hardware or software options available to use a seperate physical device or volume as cache for a hard drive.
 
the storage must have hardware storage redundancy, RAID 1 to be specific.

Why?
There are many other ways to lose the data, that a physical drive redundancy does not mitigate.
Drives are cheap. They can be replaced. Data cannot.

A RAID 1 does nothing for the far more common forms of data loss.

Optane on top of a HDD RAID 1?
Don't, even if it worked like that.


So, what is the actual requirement here?
What will these systems be used for?
 


So RST driver won't do it for me?

I know of ExpressCache. One of our laptops came with a 24GB SanDisk module and the ExpressCache driver. There's a tool included with it that shows data access statistics like cache hit/miss ratio. I watched it fill up and hits grow after wiping the cache, but otherwise I don't know how much of an performance improvement it offers, machine was pretty fast to begin with.
 


Because drives die. I have much experience with dying drives. I had an SSD die too (which took two senior design projects for two engineering degrees with it).

RAID 1 is by far the cheapest and easiest way to keep your machine running and data intact when a drive dies. I'm very busy these days and I don't have much time for maintenance. I've done this for years and it's kept everyone running without interruptions or too much headache for me, I just want some read access time boost now.

These systems are used for home office of family members. Important stuff is kept there, no gaming.
 
Yes, drives absolutely do die. Every day. You've seen it, I've seen it...

Data also gets corrupted, deleted, ransomed....which a RAID 1 does absolutely nothing for.
A RAID 1 presents to the user and OS as a single volume and its data.
Delete something by accident..poof, it's gone.

An actual backup can retain that data.
In the event of a physical drive fail, simply slot in a new drive and recover from the automated backup from yesterday.

I would never, ever rely on a RAID 1 as protection against data loss. Physical drive fail or otherwise.
Any company that deploys a RAID 1 also has actual backups.
 
Common configuration in 2018:
250 or 500GB SSD as the OS and application drive
2TB HDD for all the personal data
2-3-4TB external drive (or NAS box) as a target for backups, using one of the current imaging tools. Free or cheap, automated, easy, mostly foolproof.

In the event of a dead drive, slot in a new one, and recover from last nights backup.
In the even of a major virus or ransomware...again, recover from last nights backup.
 


True.

Users in question are mature enough to not delete files by accident.

Data backups are made, though not frequently enough. I will soon correct this as I have deployed a FreeNAS 8-drive ZFS2 file server on the home network to which I can automate backups.

Even though I can have my backups all figured out, I'd hate to have the scenario where the sole OS drive is dead, and the machine is unusable until I find time to perform repairs.
 


Everyone has deleted a critical file by accident. Everyone.
If you haven't, you will.

Or an OS update does.
Or the cat walks across the keyboard
Or the laptop battery dies during an update.
or, or, or, or....anyone of a zillion other ways to 'lose data'.

A RAID 1 can be OK, if and only if there is an actual, automated, tested backup routine to go with it.
Don't rely on it to be the One True data protection method.
 


So once we determine that RAID 1 is fine, what do we do to speed up access time (back to the subject of the thread)?
 


A real SSD as the OS drive.
To ward off the rare physical drive fail, you could probably do 2x SSD (250GB or 500GB each) RAID 1 for the OS, and 2x HDD RAID 1 as the data location.

I wouldn't, but you could.
 


Have to agree with this. If extra snappiness is the goal there's no comparison to properly running from an SSD, and it's not hard to point the various user folders to a D drive to automatically store everything on the HDD (or HDD array if you choose). Caching is a much smaller benefit and only a benefit at all to a subset of the information whose effectiveness depends on how well some software algorithm guesses what you'll need in the cache.

 


SSD RAID is still expensive. Is there no common good SSD caching solution? Of course, SSD RAID would have been the ultimate solution for everyone if money is of no object, but there has to be a cost-effective alternative.

Also, what's the way to tell Windows at installation time to put OS and applications on one volume, but user profile and user data on another volume? I'm sure there's a way to do it post-installation, but I always wondered if there is a way to do it during installation.
 


I'm not aware of a way to do it during installation. I prefer to keep data drives disconnected at install just to make sure the system reserved (and other, for UEFI) partitions are placed on the same physical drive as C since for some reason Windows doesn't automatically do this.
 


2x 250GB SSD = $140
Yes it is more than HDD, but that's what it costs.

For the user profile etc on the second drive?
You can't and don't do that during the install. At all.

You redirect the Libraries to other drives after the OS install is done.
Win 7 & 8: http://www.tomshardware.com/faq/id-1834397/ssd-redirecting-static-files.html
Win 8.1 & 10: http://www.tomshardware.com/faq/id-2024314/windows-redirecting-folders-drives.html
 


An SSD cache is only fast across whatever data is retained in that small cache space.
And any write applications only happen at 5400 or 7200RPM HDD speed.
SSHD = 8GB SSD space
24 or 32GB SSD as cache is also small.

A 250 or 500GB SSD is fast across the whole drive.
 
Well, what do you know. Another day, another dead drive. Drive in my work PC died. I don't get to pick what my workplace buys... If it had RAID 1 I would be working in full capacity today. Now I'm reduced to catching up on secondary tasks until I get my machine fixed.

So... SSD RAID. I looked up some M.2 PCIe 3.x x4 NVMe drives, looks like no go for SSD RAID, and I don't want SATA SSDs.

Looks like I will have to experiment with one of the above 3rd party SSD caching solutions to get them working with a larger and faster SSD drive. Might even go with ExpressCache again, since I'm familiar with it. Will be pretty ideal if I get it working.

PS Looks like Z370 chipset will not work with Kaby Lake CPU through artificial BIOS lockout, which is a surprise as it was supported during development. Not cool. That's a second boo at Intel in one day (the other being no Optane for RAID). Anyway, that's off-topic.
 
I got ExpressCache working with an old 128GB Samsung SATA SSD I had laying around, as a proof of concept. Seems to work. Machine boots noticeably quicker too. Suggest a read benchmark?

I ordered a couple of those M.2 PCIe 3.0 x4 modules in 120GB capacity.