X-bit labs

500GB HDD Shootout: Seagate Barracuda 7200.10 and Others! (page 11)

Category: Storage

by Alexey Volkov , Nikita Nikolaichev

[ 10/31/2006 | 02:57 PM ]

Performance in Intel IOMeter

Average Read/Write Response Time

So, the first test we’ll compare the hard drives in is about measuring the average response time of the drive when requested to read or write a sector. The goal of our measurements is to find the read access time and see how aggressive the deferred write algorithms are (roughly speaking, to estimate the number of cache segments allotted to store write requests).

In order to do this, we use IOMeter to bombard the drive for 10 minutes with a stream of requests to read and write 512-byte data blocks with a request queue depth of 1. The drive has to process over 60 thousand requests, so we get a sustained disk response time that doesn’t depend on the amount of cache memory. The results are sorted by the read response time:

The numbers are quite interesting. Hitachi’s drives are in the lead when it comes to reading, and Seagate’s ATA-interfaced drive from the Barracuda 7200.9 series is the last one here. Either to achieve quiet operation of the drive or to separate sharply “fast” SATA drives from “slow” ATA ones, Seagate slows down its ATA models since the Barracuda 7200.7 series.

As for the write response time, Maxtor’s four drives are in the lead and are closely followed by the Hitachi with a 16MB buffer. Seagate’s drives have the worst results, and two of them have downright poor performance with deferred writes. We can explain this for the NL35 drive by Seagate’s desire to increase data storage reliability on server-oriented drives, but this explanation doesn’t work for the ST3500641AS. It must have suffered for the other model (we mean they tested the firmware on it).