Professional / Workstation disk setups, what are people running

[JasonM]

Just putting this out to the forum to get feedback. It's a question to anyone who uses their computer for professional work and aiming for maximum up-time. Basically anyone who depends on their computer for an actual living.

Currently i'm running Samsung PRO SSD's but with Intel Raid Recovery (off-line Raid 1) on either Western Digital Black or RE drives.

Last week however my Samsung 840 PRO failed, Raid Recovery allowed the machine to boot from HDD. I had a Crucial MX 100 spare and have rebuilt to this, however it's a temporary situation as looking at reviews I don't trust that drive. To be hones I've had a lot of SSD's fail over the last few years, and after the 840 Pro failing, I'm thinking of relying more on HDD's.

One option is Intel RAID recovery using to Western Digital RE4 drive, and then adding Intel Rapid Response technology to a RAID 1 volume. This way if the SSD fails it's only the cache volume, and the RAID 1 volume is still intact.

So just opening it up to the forum, I know there's professional workstation setup's out there, what disks / SSD's and how are they configured?

 

[Quartz]

Do you want your workstation to survive a drive failure without hiccup and no loss of work? You want it to report that the drive has failed but you just want to replace the drive and keep working during the rebuild and keep working until you can obtain a replacement drive? For this I suggest you look at a proper hardware RAID controller. Not a FakeRaid controller.

 

[JasonM]

Hi Quartz,

I was looking for something that if the OS drive fails, then the computer can keep running, at least after a reboot and removing the failed drive.

Like SSD's, however this is the 5th thats either lost data or failed totally, had best luck with Samsung 830's however.

Curious what disk set-ups people run.

 

[Quartz]

So something like one of these Adaptec RAID controllers, then, running RAID 1 sets. They're not cheap, but then you need to consider the value of your time.

 

[JasonM]

There is no issue with the Intel RAID, it always works well for me. The Issue I have is SSD's failing.

I was just trying to create some discussion about the SSD's / HDD's / SSD caching on workstation type setups. Drives that have high amount of writes, or computers that are on 18 hours a day. I'm thinking about this stuff and wondered what other people do.

 

[Deleted member 138126]

I suspect there is something very badly wrong if you've lost 5 SSDs. Maybe you need a power conditioner or a UPS that doesn't just pass the power through (i.e. one that uses the mains to feed the battery, but runs the load off the battery so you always get pure power).

For that matter, how do you know it's not the Intel RAID that's causing you all these problems? Have all 5 SSDs died in the same workstation? I also don't understand how RAIDing an SSD with a hard drive can work -- they are literally 100-1000x apart in terms of latency. Presumably you are just getting the performance of the hard drive, not the SSD?

So many things wrong...

 

[Deleted member 138126]

There is no issue with the Intel RAID, it always works well for me. The Issue I have is SSD's failing.

Better that you get to the bottom of why you have so many SSDs failing -- that ain't normal. Running a computer 18 hours a day is nothing special, a lot of people on this forum probably hammer their computers way harder than any professional workstation user would in terms of heat, dust, number of devices inside and out, multiple power-hungry graphics cards, and so on.

 

[JasonM]

Some of the SSD's are on different machines. I have 4 Computers + 1 Laptop in my house.

These are whats failed.

Kingston 96GB, died needed replacing. Separate Computer.
Kingston 120GB, kept running but it's performance dropped off, stuttering mouse / hanging. Separate Computer
Samsung 830, lost all data but once formatted worked ok again. Seperate Computer, working ok in this one.
Crucial M4 64 GB, died. - This computer
Samsung 840 PRO died, only drive that died in RAID. - This computer.

I do have another 4 SSD's in use that have been issue free.

The RAID i'm running is RAID recovery, it's an off-line RAID 1 where you update data to second disk manually. If your primary drive fails, then the second drive takes over but it's only as recent as when it was last updated.

You raise a good point about RAID possibly damaging the 840 PRO, however this setup had been running since late 2012/2013 without any issue, then one morning the drive reported an error in BIOS.

 

[JasonM]

What I think killed the Crucial M4 was it was used as a dedicated page file.

I'm a software developer, and I had some Neural Networks that were generating crazy data-sets, despite having 32GB in this machine I had over 100GB page files in use (this was some .Net data sets), I was spanning page over over various SSD's and I think this killed the M4. However the Samsung 840 PRO just died one morning for no obvious reason.

 

[Quartz]

There is no issue with the Intel RAID, it always works well for me. The Issue I have is SSD's failing.

The SSDs keeping failing is actually of secondary - well, less immediate - importance. You want to keep working despite any failure. Your time costs money, after all.

 

[Quartz]

What I think killed the Crucial M4 was it was used as a dedicated page file.

I very much doubt that. Check out the SSD Endurance Experiment. The Samsung 840 Pro managed over 2PB of writes.

I'm a software developer, and I had some Neural Networks that were generating crazy data-sets, despite having 32GB in this machine I had over 100GB page files in use (this was some .Net data sets), I was spanning page over over various SSD's and I think this killed the M4. However the Samsung 840 PRO just died one morning for no obvious reason.

It would still make sense to have a box with 128 or 256 GB RAM in that case.

 

[rickh]

To answer your question, I am also a software developer working on data analysis problems, and have been using consumer samsung SSDs in my PCs since the first generation without failure. My current Samsung 850 EVO has been flawless. I take a disk image every week, and all my files are on a different SVN server.

I personally have little faith in 'fake' raid controllers. The only time I've ever lost HD data was when I used Intel/Windows RAID. If I wanted RAID, I'd stick to 'real' RAID cards with battery-backup and caching memory from the likes of LSI, Areca, etc.

 

[david25]

Consider using DC/enterprise SSDs

Intel S3500 / S3510
Intel S3610
Intel S3700 / S3710


Intel S3500 is very good, see

http://www.extremetech.com/computin...might-be-the-only-reliable-drive-manufacturer

 

[JonJ678]

I keep all the data on an obsessively backed up home server running the ZFS variant of raid 1, on the grounds that workstations die occasionally. At some point I'll probably switch to redundant servers.

If the workstation dies, there's no significant data loss, so I can boot the "spare" and carry on. The spare is actually the previous workstation which is probably still working.

This isn't optimal. In particular, if the home server dies I'll have a tedious backup restore to wait for, and if the workstation dies the predecessor is probably running an obsolete operating system.

Aiming at a single, ultra reliable workstation strikes me as doomed to failure. Computers die too often in too many ways.

edit: I think the SSD is a single crucial M4 which I expect to fail at some point