Articles: Networking
 

Bookmark and Share

(3) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]

Performance

We could not get together four identical HDDs for this test. Instead, we collected four 500GB models from different generations: Fujitsu MJA2500BH, Hitachi HTS545050B9A300, Toshiba MK5055GSX and Samsung HM500JI. The other test conditions are the same as in our previous NAS tests: a mainstream PC with a dual-core processor and a Gigabit Ethernet controller works as a client. It runs 32-bit Windows Vista and is connected to the NAS via Gigabit Ethernet switch. Jumbo Frames are turned on. Intel NASPT 1.7.0 is used as the benchmark.

As we’ve got four different HDDs today, let’s first see if the HDD model affects the performance of the NAS. We used single-disk volumes in this test.

The results suggest that the HDDs differ considerably with streaming data patterns. There is no overall winner, though. The Fujitsu is good at reading whereas the Samsung is the best at writing. On non-streaming operations the difference is negligible. So, you can choose HDDs for your NAS without looking for the highest-performance model. Instead, you should consider such factors as availability, price, capacity and personal preferences concerning particular brands.

Next let’s check out the NAS’s performance in different RAID modes. We had no problems using different disks in RAID arrays, by the way. There was no incompatibility or anything although the results perhaps would have been somewhat higher with identical HDDs.

The second graph shows the speed of RAID0 with different number of disks. Here and in the third diagram the average speed of our four models is shown as the speed of a single disk.

RAID0 helps improve the performance with streaming operations. The speed grows up to 37MBps at reading and 55MBps at writing. This is an impressive result for such a compact device and is largely due to the 1.2GHz processor. Comparing these numbers to the performance of other models based on the same processor, the DS409slim is much faster.

The third diagram shows the performance of the fault-tolerant RAID5 and RAID6 arrays. We guess RAID1 is not a good idea for this device in terms of overall cost. RAID6 isn’t quite appealing due to the same reason.

Calculating checksums is not an easy task for the ARM core. Fortunately, this is only necessary for writing data. The write speed of the RAID6 is only half that of the single disk. The 4-disk RAID5 is good, delivering a speed of 30MBps. Reading is much better. The RAID5 accelerates to 33MBps while the RAID6 is comparable to a single disk. Note also that the differences are less conspicuous with patterns consisting of a large number of small files.

Judging by the results, the 4-disk RAID5 is the best option if you need fault tolerance. This array can survive the loss of one disk and delivers good enough performance.

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]

Discussion

Comments currently: 3
Discussion started: 11/10/09 07:53:17 AM
Latest comment: 11/23/09 08:58:37 AM

View comments

Add your Comment