ENTERPRISE VS. DESKTOP DRIVES – THE SELECTIVE PROCESS FOR NETWORK STORAGE
We have been asked many times to use desktop drives in server class systems as customer want to save money but are not clear why they need enterprise.
Good question, very tempting to use cost targeted desktop drives in a server environment, even in small businesses.
One of the questions we need to ask is how important is the data that is going be stored? and in often times even in the smallest of servers there is some kind of RAID array setup. When you involve multiple drives you will run into issues such as rotational vibration, which we will touch up on later.
Other issues you can run into are more basic and fundamental. There are three levels of drives most commonly available two of which are desktop and notebook drives. These are manufactured to hit a particular price point, and tend to have the shortest Mean Time Between Failures (MTBF) ratio, and consequently are aimed and produced as a low cost drive. The integration temptation in a multiple drive array from a cost ratio can become very attractive. The drives are also designed for deployment in a 8 hours a day, 5 days a week scenario in the worst case.
So you have a lower cost, less well designed, and short warranty drive really made to slot into the low cost market segmentation, the typical MTBF for desktop drives is about 700,000 hrs of usage, which may sound a lot but compare this to the Enterprise drive which with an MTBF of 1.2 Million hrs of usage measured by 24 hours and 7 days a week, these enterprise drives or nearline as they are sometimes referred to, are well designed carry a longer warranty and targeted for the server and RAID based market. The next level of drives, which one of the manufacturers (Seagate) refers to as mission critical drives (higher RPM) can yield up to 2 Million hrs of usage, these are typically spinning at 10K to 15K rpm.
When you look at the cost element, many users ask if we purchased a RAID based system we already have redundancy built in, so why worry if these are desktop drives.
The reason why it’s still an issue is when you run multiple drives in a single chassis, you have a thing called angular velocity, or revolutions per sec (RPM), rotational vibration can be caused from an adjacent drive normally caused by drives shaking not because the spindles are not balanced necessarily but its head mechanism of both drives may be at opposite ends both carrying out differing seeks, and they can interfere with each other, RV can also occur with other parts of the case like cooling fans, or just structural vibration, any and all types of vibration can effect and harm drives. When the drive head is trying to seek a track on that rotating media, it has tens of thousands of tracks per inch, and so it has to stay on a specific data track and go seek another one and it’s got to get it and if it can’t you then run into data errors. On top of that just the design nature of the desktop and notebook drives are in order of magnitude less reliable, that means a read error is going to occur earlier even if they are in the correct environment, and when you are dealing with rotational vibration issues the impact on desktop and notebook drives will be ten times more than they would with enterprise drives.
So, what does the enterprise drives have that the desktop don’t to compensate for this rotational vibration issue?
Enterprise and nearline drives have rotational vibration sensors for one thing that helps them adjust, and also they run more data integrity checks which does not make them slower.
The graph to the left depicts the effect of not having Vibration Suppression.
Hard disks installed with Vibration Suppression clearly benefit, the effect of which allows increases in both longevity and realiabity.
The other factor to consider is even though using multiple drives in a chassis with RAID may seem to provide adequate protection against data loss, this belief is false and does not necessarily provide a safe haven, when a drive drops out, be it desktop or not it will enter into a RAID re-build and depending on the RAID level used, be it RAID 5 or RAID 1 which are very common in desktop array’s, all you need is a single hiccup during that rebuilt and the whole array will be toast, and your data will have gone south!. Then if you have another back up you are thinking of getting restored, just consider the time involved and the productivity lost. Consequently all in all whether you have RAID 5 or any RAID you are not as protected as you think even less so when using desktop drives.
One other factor that desktop and notebook drives suffer from in a Multiple drive RAID array situation is as they are not designed for attachment to a RAID controller and so there is a Time Limited Error Recover (TLER) function that is built into enterprise and nearline drives to make them work better with RAID controllers. When you use desktop drives what can happen in the same situation is you can end up with false rebuilds as the drive is trying to do error correction and the controller thinks the drive has disappeared and tries to rebuild.
There are many manufacturers that are recommending using desktop drives within their chassis which are designed to limit rotational vibration and they have also build for a shorter lifecycle of around 3 years maximum.
Having considered various options, it seems clear if the data required is mission critical, and requires 24/7 access with no other backup, then stepping up to select enterprise drives seems the most prudent choice, in fact serious consideration should be given for having a secondary unit as backup to the first storage array.
CAUTION: RAID is not a substitution for not backing up, redundancy is not the same thing as back up. Even if you are running a RAID system you are still susceptible to Viruses, or accidental deletion or other human errors and disasters, so make sure you do regular backups, and have a backup for a backup.