Server OS drive, RAID1 or single SSD?

theinv

Limp Gawd
Joined
Apr 17, 2008
Messages
216
I am building a web server with dual xeon CPUs. The server will be running 24x7 and may have very high traffic.

The server will be running windows server 2008 R2 with hyperV. The web server will be running on virtual machines in the data drive (RAID1 or RAID5)

Currently, I have a 256GB SSD drive. Initially, I was thinking of using a single SSD for the server OS drive, but now I think it is too risky. From what I have read, SSD drives have a limit on the number of write operations, who knows when it will reach its end of life suddenly. If the SSD drive goes bad, I need to reinstall the main OS.

I want to do RAID1 on the OS, but I don't want to spend another $500 to get another SSD.

So here is my question, is two 7200RPM drives in RAID1 better and more reliable than a single SSD for OS drive? I will use the enterprise class drive for the RAID1.
 
Last edited:
RAID 1 OS Drive.

High traffic usually means downtime costs money. RAID 1 is simply more fault tolerant.

If you need the SSD access times, then SSD RAID 1. If that is too expensive, you probably want to look at 2.5" 15K RPM SAS.
 
Definitely RAID1, it has increased read speeds which really helps an OS drive, and the redundancy is good.
 
Based on your description

1. The server OS only serves as Hyper-V Host. For uptime protection, yes you can use RAID-1 7200RPM SATA drives.

2. It is not clear from description how many disk slots you have on the server.
3. You mention web server will be running as VM on data drive (RAID-1 or RAID-5)
3a. this is the part you can actually put SSD into good use. Web server usually has a lot of Read IOPS, a good use case for SSD, which you can optimize with slight investment.

Cheers
 
More than one SSD would be needed though. Having only one SSD is not secure enough on a hardware-level for a server. If two were in raid 1, then he'd be good, but having only one is not a good choice imo, considering how often they tend to die compared to HDDs.
 
Based on your description

1. The server OS only serves as Hyper-V Host. For uptime protection, yes you can use RAID-1 7200RPM SATA drives.

2. It is not clear from description how many disk slots you have on the server.
3. You mention web server will be running as VM on data drive (RAID-1 or RAID-5)
3a. this is the part you can actually put SSD into good use. Web server usually has a lot of Read IOPS, a good use case for SSD, which you can optimize with slight investment.

Cheers

Thanks.

The server has 6 SATA slots.

Regarding #3, There are a lot of reads for web server, but there are also a lot of writes too due to logging activities. I am not sure whether the number of writes will shorten the life of SSD.

Anyone know how SSD works when it reaches the end of its life? Will it die suddenly or there are signs alert you it is about to die?
 
More than one SSD would be needed though. Having only one SSD is not secure enough on a hardware-level for a server. If two were in raid 1, then he'd be good, but having only one is not a good choice imo, considering how often they tend to die compared to HDDs.

Yep, you are absolutely correct!

Thus we need more info how theinv is going to layout storage for his data drive where his VM is sitting, and how many free disk slots on the server, and web server VM OS, plus probably other things if it is expecting serious 24x7 usage.

Cheers
 
Thanks.

The server has 6 SATA slots.

Regarding #3, There are a lot of reads for web server, but there are also a lot of writes too due to logging activities. I am not sure whether the number of writes will shorten the life of SSD.

Anyone know how SSD works when it reaches the end of its life? Will it die suddenly or there are signs alert you it is about to die?

Thanks.

The server has 6 SATA slots.

Regarding #3, There are a lot of reads for web server, but there are also a lot of writes too due to logging activities. I am not sure whether the number of writes will shorten the life of SSD.

Anyone know how SSD works when it reaches the end of its life? Will it die suddenly or there are signs alert you it is about to die?

following is just idea, not final. Please evaluate seriously with professional, my plan is not tested. You should have better options from other pro.

Port-0 OS RAID-1 7200RPM SATA/SAS-nearline(if this is SAS controller with SATA devices)
Port-1 OS RAID-1 7200RPM SATA/SAS-nearline(if this is SAS controller with SATA devices)
Port-2 DATA RAID 15K or 10K-RPM SAS (if this is SAS controller with SATA devices, else 10KRPM-SATA)
Port-3 DATA RAID Intel X25E-32GB
Port-4 DATA RAID Intel X25E-64GB
Port-5 DATA RAID Intel X25M G2 120GB

Port (2/3/4/5) combined to form RAID group. You have many options when using FOSS software RAID to skew the IOPS runtime operation. Since you are set on Windows, I do not have enough idea on that.

You can mix and match. Make sure SSD size is different. I suggest X25E for safety. If you are confident you can put multiple G2. I recommend career protection above minor cost saving since your project sounds like major one.

Again, if your job depends on it, please consult with real professional.

Cheers.

Edit : To save cost, the DATA RAID portion you can alternate with SATA/SSD-X25M-G2 combination. Need to be very careful with software behaviour to optimize runtime. Some people believe by the time you are done with analysis, you might as well go with SLC X25-E
 
Last edited:
One thing not considered above (I think anyway) is a hot spare capability. If you have sufficient monitoring and 24/7 support then maybe you can get away without a hot spare capability for each disk pool I present but imho for anything that has an SLA or price attached I wouldn't personally configure a system without one.

Not sure why the assessment is on IOPS here - I wouldn't think a web service would typically be bound by disk performance - more likely to be network bandwidth as the constraint, but maybe I am wrong :)

On the disk front first question is whether the RAID config supports hot spares or hot global spares.

Assuming it does and taking the leap of faith that from the description above it doesn't sound like you need too much space, so I would at least consider buying 6 drives of the same type, RAID 1 for OS, RAID 5 using 3 disks for the VM storage and set 1 disk as a global spare to leap in if one of the others fail.
 
If you have THAT much logging going on, and aren't concerned about losing them, why not have logs write to SSD?

Keep the OS on the disks like you mentioned, and keep your hosting data on an array as well.

So
Port 1 - Array 1 (Raid1) - OS drive 7200rpm sata
Port 2 - Array 1 (Raid1) - OS drive 7200rpm sata
Port 3 - Array 2 (Raid1) - Webserver Data 10k velociraptor or 7200rpm sata
Port 4 - Array 2 (Raid1) - Webserver Data 10k velociraptor or 7200rpm sata
Port 5 - SSD For logs

or run Raid 10 for OS & Web data w/ ssd for logging. with the last port holding a hot spare.
 
Back
Top