Drive burst speed vs sustained transfer rate?

[Roll]

Just been running various benchmarks on my drives/arrays just out of general interest really.

Of the two benchmarks, burst speed and sustained transfer rate, I was wondering which gives the best indication of how well drives perform against each other in real life situations (eg. starting Windows, loading an application or game)?

It was interesting that the array on my first system reported the better burst speed whilst the array on my other system had a higher sustained transfer rate.

Thanks for replies

 

[rpstewart]

Real world performance will be determined by a mix of access time and sustained transfer rate.

The burst rate is really just a measure of how fast the drive can send the contents of its cache to the PC, which is great if what you want is in the cache but there isn't a lot of intelligence in the cache population algorithm. The drive doesn't know what the meaning of the data being requested is or where to find the next logical bit of it, that's all done a the OS level so the cache is likely to just be the next few sectors following on from the last request. Again great for contiguous files BUT the cache can only be replenished at the sustained transfer rate to it's all a bit pointless really.

The drive or array's access time will determine how quickly, on average, a particular sector of the disk can be located. This will determine the delay in finding non contiguous file fragments or the next file the OS requires. The sustained transfer rate will then determine how fast that data can be read off the disk and pumped down to the PC.

 

[Roll]

Thanks rpstewart

So I suppose that really the most meaningful benchmark is the sustained transfer rate, and that this is the best way to measure how drives/arrays perform against each other in real world situations.