New build. RAID5 on RST ok?

FAQ, getting help, user experience about PrimoCache
rajdude
Level 4
Level 4
Posts: 22
Joined: Thu Mar 07, 2019 5:24 pm

New build. RAID5 on RST ok?

Post by rajdude »

Hello everyone,
A big hello from a newbie here!

So I am about to update a 10 year old home server. The new setup will have this:
  • Supermicro server motherboard
  • Cheap Xeon CPU
  • 8 GB RAM ECC
  • 4 or 5 NAS HDDs maybe 4 TB or 6 TB each, trying to get 7200 rpm ones
  • RAID 5 created in BIOS using Intel RST
  • OS can be Windows 10 or Windows Server 2016
  • 120 GB SATA SSD for OS
  • 120 GB SATA SSD for PrimoCache
  • Usage: Lots of photos and home videos (around 70,000) Some ripped DVDs and Blu-Rays also. Currently around 5 TB used.
Questions:
  1. Will this config work well? Any teaks needed?
  2. Will I be OK with RST managing the RAID and PrimoCache accelerating only the RAID array?
(Asking because I hear PrimoCache used to have issues with RST.)

Thanks for your input!

-rajdude
User avatar
Jaga
Contributor
Contributor
Posts: 692
Joined: Sat Jan 25, 2014 1:11 am

Re: New build. RAID5 on RST ok?

Post by Jaga »

It was many years ago that RST caused problems for Primocache, but it didn't last long after I sent a report into Romex support. They squashed the bug/problem in about 36 hours. I've heard confirmation recently from them that it poses no problem at all.

What is your strategy for using Primocache on the server? With 8GB of RAM you won't have a lot for a L1 cache, but it seems like you devoted a 120GB SSD for a L2 cache (which should help quite a bit). Do you plan on caching the RAID 5 array with the SSD? It's a large enough array that you may get low hitrates if a lot of data resides on it.

One thing I'd suggest as a change to architecture - don't use RAID 5. It's outdated, and not nearly as reliable as it used to be on large drives (rebuilds stand a chance of failing, crashing the array and making it unrecoverable). Instead I'd use drive pooling software like Stablebit DrivePool. I use that on my home server (9x8TB drives and 4x4TB dual-parity drives). You can add drives on the fly to the pool, and a ton of other neat stuff.

When you couple Drivepool with on-demand parity calculation software like SnapRAID, you get the best of both worlds: re-compute parity with a simple powershell script command, recover individual files or entire drives, scrub the parity against the data to verify parity integrity, many levels of parity (up to 7 I think), etc.

You can of course still do RAID 5, but I think the ship has sailed on that architecture for the most part. I used it for many years, even on an old WD MyCloud EX4, but finally retired all use of RAID in favor of drive pooling and on-demand parity generation.

I'm a little split on using Server 2016 for a home media/file server - they removed some things from it, and you just don't gain a lot for the extra cost and administration hassle. Mine is still running Windows 7 x64 believe it or not, and doing very well. If I had to upgrade the OS, I'd probably go with W10, and simply disable all telemetry.
rajdude
Level 4
Level 4
Posts: 22
Joined: Thu Mar 07, 2019 5:24 pm

Re: New build. RAID5 on RST ok?

Post by rajdude »

Thanks for your response.

You got my proposed strategy right. I was planning to use the 2nd SSD as a cache for the array. However, I am also concerned about lot hitrates since I will have a lot of data on the array.

You see, my day job dabbles in architecture of enterprise storage arrays, used in datacenters. I work with likes of 3PAR, Compellent, NetApp, Pure storage etc. I also use them every day at work for running a large number of virtual machines. All those vendors use cache in front of their large arrays. Their cache could be SSD and/or RAM based. They all use different version of RAID, some have proprietary file systems also. They work really well...have pretty good cache hitrates, based on the performance tools they give us to use.

However, all that is way too expensive for home use. So I thought of trying out PrimoCache.
Thanks for the other recommendations, I will look into them too.
User avatar
Jaga
Contributor
Contributor
Posts: 692
Joined: Sat Jan 25, 2014 1:11 am

Re: New build. RAID5 on RST ok?

Post by Jaga »

Primocache is able to use a L2 cache volume of up to 1TB in size, so against your data store that should do *much* better for hitrates than a 250GB volume. Might be something to consider. Or even combining 2 500GB SSDs in a hardware RAID 0 configuration for the L2 volume.
neatchee
Level 5
Level 5
Posts: 49
Joined: Tue Feb 12, 2019 8:38 pm

Re: New build. RAID5 on RST ok?

Post by neatchee »

I think it's important to be clear that the hitrate on a 120gb L2CACHE in front of 5+TB of data entirely depends on your usage pattern.
  • If you're usually watching the same few TV shows/movies/photos/etc, and only dipping into the majority of your data for the occasional trip down memory lane, you'll have a decent hitrate
  • If you're constantly cycling through your entire library and cover the entire 5+TB before repeating anything, your hitrate will be abysmal
Usage pattern is far more important here than the contents of the drive, per se. As you obviously know, the cache is going to be populated with the most frequently used data. So it's really up to your behavior to determine whether 120gb of cache is enough :)

I agree with Jaga that RAID5 is basically obsolete at this point for any serious implementation. But if you're not really looking to bring your work home with you, and just want something that works and is easy to set up and forget about, then it should be alright. Just be aware of the risks as Jaga described and consider periodic offsite backups.
cichy45
Level 4
Level 4
Posts: 38
Joined: Sun Oct 14, 2018 3:34 pm

Re: New build. RAID5 on RST ok?

Post by cichy45 »

I support what Jaga said. RAID5 is obsolete, RAID6 has disadvantage with 3 disks failing you lose all data. My setup is similar (although much smaller with 5x1TB data HDDs + 2x1TB SnapRAID Parity HDDs). As for caching strategy I went with 2x 128GB SSD in RAID0 (for good simultaneous read/write performance) as write-cache only.

Setup Idle-Flush with INFINITE timing set. This produces behavior that never flush data to disk when it is active and allows to flush cache at the same time that you are writing to cache (lets say Pool with disk A B C cached, while files are being written (cached) to B and at the same time cache is being emptied from files that go on disk A (currently idle)), so this gives effective cache capacity >256GB. I do not cache reads from my pool as it is storage only. I mostly write to it locally (and this is quick) and rarely retrieve data from it over LAN (it is still quick, as single HDD can go up to 200MB/s so 2x faster than typical 1Gbps LAN.
User avatar
Jaga
Contributor
Contributor
Posts: 692
Joined: Sat Jan 25, 2014 1:11 am

Re: New build. RAID5 on RST ok?

Post by Jaga »

neatchee wrote: Sat Mar 09, 2019 6:46 am I think it's important to be clear that the hitrate on a 120gb L2CACHE in front of 5+TB of data entirely depends on your usage pattern.
Very true. I made a blanket statement that assumes people use different data regularly. If you do the same things with the same files every day, a small cache will serve perfectly well.

Good to hear the other confirmations on architecture and which way to go.
rajdude
Level 4
Level 4
Posts: 22
Joined: Thu Mar 07, 2019 5:24 pm

Re: New build. RAID5 on RST ok?

Post by rajdude »

Ok, so I am very curious........if RAID is obsolete......what do people actually use now? Maybe a JBOD with backup, and pray that a disk does not fail? ;)

Now I will admit, I did not know what snapraid is (just googled it)...reading up on that now....but what if those 2 parity RAID disks die first? Do you lose all the data? Probably not.

Most important question for me is : how are the read write speeds? Looks like it will be the same as a single drive.
With RAID 5 (or 6) the more disks in an array, the more IOPS or MB/s you get. Thats why lots of slow disks = a fast disk array.

As for backup, yes of course I backup. I use a two tier backup strategy:
Backup to an external disk + cloud backup (real backup meaning proper versioning, not just sync)
I use CrashPlan for business, for now.

Not saying my way is the best.... :D it could be obsolete by now. Hence I am searching for better options and am open to suggestions like yours. Thanks!
neatchee
Level 5
Level 5
Posts: 49
Joined: Tue Feb 12, 2019 8:38 pm

Re: New build. RAID5 on RST ok?

Post by neatchee »

RAID 6 for capacity or RAID 10 for performance.

RAID 6 will eventually become "obsolete" too, but that's a bigger concern for enterprise right now than it is for home use.
User avatar
Jaga
Contributor
Contributor
Posts: 692
Joined: Sat Jan 25, 2014 1:11 am

Re: New build. RAID5 on RST ok?

Post by Jaga »

rajdude wrote: Tue Mar 12, 2019 12:03 am Ok, so I am very curious........if RAID is obsolete......what do people actually use now? Maybe a JBOD with backup, and pray that a disk does not fail? ;)
The software I posted above - Stablebit Drivepool. It makes a group of drives look like one large drive without the drawbacks of hardware RAID. Pairing the SnapRAID software with it is a great solution to completely replace RAID 5/6/etc.
rajdude wrote: Tue Mar 12, 2019 12:03 amNow I will admit, I did not know what snapraid is (just googled it)...reading up on that now....but what if those 2 parity RAID disks die first? Do you lose all the data? Probably not.
If however many drives you have holding SnapRAID's parity somehow all die at once, you just replace the drives and rebuild the parity. Nothing happens to the data. The beautiful thing about SnapRAID is it can have up to 7 levels of redundant parity. And you can split a parity set across multiple volumes (like I do). I have 4 4TB drives doing the work of 2 8TB parity volumes (4TB+4TB x2).

Drive Pooling really is a better overall solution, if you automate the parity calculation for protection (like with a powershell command and the Windows scheduler). With dual parity on 4 drives I don't even feel the need for backup anymore. The chances of losing 3 or more drives at once is almost zero for my environment. And replacements are super easy for me to source on the same day.
Post Reply