4

I am a (Java) developer currently working in Investment Banking and I have always wondered why it is so difficult to increase the capacity of mapped drives on the *nix servers we run our applications on.

I realise there may be cost restraints, but this does not explain the gulf between server hard drive capacity and personal hard drive capacity. For example, I have more capacity on the machine I work on than the production server used for our applications. I also find it baffling that when new capacity is requested (in the region of 10s of GBs), so much business justification is required (I understand the justification requirement if TBs is required).

Can someone explain the difference between the disk in my PC and the disk in a server; are they really the same or is the gulf in capacity due to different hardware requirements. Conversely, are the physical disks the same but the infrastructure requirements much greater when used in a server (i.e. RAID, backup etc).

Richard
  • 143

2 Answers2

4

Enterprise-grade hard drives are built with a different philosophy to consumer-grade drives - that is, speed is not king so much in the field of server storage devices; reliability is.

For a server, what you're looking for is a high MTBF value (Mean Time Between Failures) which gives a rough idea of how long the drive will operate for before probability suggests it will start experiencing old-age related issues. For the drives in my workstation, that value isn't specified, because they're bog-standard Seagate consumer drives. For the drives in my servers, they're around 1.4 million hour MTBF rated.

The harder you push the envelope, the lower your stability will become; overclocking is a prime example of this. When creating hardware, the more you force the same technology to do, and the closer to the limit you bend it, the less confident you can be that it will survive in a datacentre environment, churning away day after day for multiple years between swapouts - because that's the point of a server's hard drive. Desktop PCs are used very differently, get a lot less work (in varying setups) in higher bursts, and are built according to what sells - capacity and speed.

Xyon
  • 1,593
  • 9
  • 15
1

Hard disks are cheap; network storage devices aren't (if you want something that can happily be used by a lot of people, you pretty much have to buy a NetApp at $20k). Tape backup isn't cheap either and gets less convenient as you need more tapes every night.

pjc50
  • 6,186