Disk Defragmenter - How to make it work.

[Bambro]

@dzero

Aren’t there some people who are of the opinion that putting SSD’s in RAID doesn’t result in a performance boost. I never followed the arguments closely, but it seems to me that data striping would have to be beneficial regardless of the media.

Would you be able to say if the increase is mostly linear - as in 2 drives = twice as fast, 3 ==> 3x, etc.? I would imagine that at some point you would have to hit a wall, but IDK. Thanks.

I believe it depends on your controller and the I/Ops it can handle. I have had as many as 4 SSDs in the array and it was considerably faster. I also had them running in a 1+0 at one point for testing so I could sustain multiple failures. I removed them and sold the 2 other SSDs as it was just overkill at that point. But at some point you will reach a bottleneck due to the reaching the maximum I/Ops that your controller can handle.

I think it was determined while specing out my SAN at work that the theoretical limit for a spindle was around 150 I/Ops. In order to get the high I/O we required for the databases we spread them across 14 spindles which also left a lot of room for downward growth. Now an SSD is going to have much higher I/Ops than a spindle, but you should still see improvement when adding more drives to the array…and if you have the ability to span controllers then I would expect even higher benchmarks.

[voltaire]

Besides access speed, another reason to defrag is that when your HD starts to fail and a certain sector dies, that bad sector will tend to affect less things all at once. It’s better to lose one big chunk of data from a limited amount of sources than it is to lose a whole bunch of tiny pieces of data from a whole lot of different sources.

But hopefully you have everything backed up, anyway.

[Mgalindo13]

@voltaire

Besides access speed, another reason to defrag is that when your HD starts to fail and a certain sector dies, that bad sector will tend to affect less things all at once. It’s better to lose one big chunk of data from a limited amount of sources than it is to lose a whole bunch of tiny pieces of data from a whole lot of different sources.

But hopefully you have everything backed up, anyway.

Excellent point!