Performance testing
N5200 Pro Performance
We used IOMeter version 2006.07.27 over SMB shares and, in the case of iSCSI mode, using NTFS.
Boston kindly provided us with five Seagate 7200.10 750GB SATA disks, so big thanks are in order. The disks were jumpered-up for 3Gbps operation.
Setup
The host machine for IOMeter was as follows:
| Component | Details |
|---|---|
| CPU | AMD Opteron 146 @ 2.5GHz |
| Motherboard | ABIT AN8 Ultra (nForce 4 Ultra) |
| Memory | 2.0GB PC3200 DDR @ 209MHz |
| Disks | 4x Seagate 7200.8 250GB - RAID 10 |
| Graphics | NVIDIA GeForce 6800 GT 256MiB |
| Network | NVIDIA network controller, 1Gbps, 9000byte frames |
| OS | Microsoft Windows XP x64 |
The networking might that is the N5200 Pro's NICs can handle jumbo frames up to 16000bytes, but 8000 was the closest we could get to the host's maximum of 9000, so we set it to that.
Below is our IOMeter test regime:
| Option/test | Configuration |
|---|---|
| Outstanding I/Os | 10 |
| Individual test run time | 30 seconds |
| Read test access spec | 1MB transfers 100% sequential 100% read |
| Write test access spec | 1MB transfers 100% sequential 100% write |
| General usage access spec | 64KB transfers 50% sequential, 50% random 33% write, 67% read |
For comparison, we've included results for Thecus's four-disk rackmount 1U4500. Also in the fray is QNAP's TS-109 Pro, a single-disk solution.
The N5200 non-Pro was tested before our current NAS testing scheme was in place, so we've not included it in the graphs, simply to ensure consistency between results.
SMB performance
Windows File Sharing isn't as quick as NFS, iSCSI and a few others, but its use is widespread, so the N5200 has to deliver satisfactory throughput with it.
We don't have throughput test data for the 1U4500 below 250MiB file sizes, so please bear this in mind when considering the results.
Read performance first...
Here we can see that the 1U4500, which has the same motherboard, CPU and RAM under the hood as the N5200 Pro, edges in front of its five-disk companion.
The N5200 Pro's performance with a 1GiB test file proved significantly less than the other test subjects, which seems a little odd. We cannot put our finger on the reason but we might take a guess at the firmware having something to do with it, given the lack of architectural differences between N5200 and 1U4500.
As we've already stated, the 1U4500 is slightly quicker than the N5200 Pro. Since it has one less disk in its array, we'd wager that the reduction in RAID management overhead is what provides this edge. The single-disk TS-109 Pro does an outstanding job against the two big boys, shaming them at larger file sizes.
Also noteworthy is the fact that the N5200 Pro's performance peaks at around the 250MiB mark and starts to tail off (though not in any dramatic way) as the file size gets smaller still.
For write performance, the 1U4500 and N5200 Pro deliver similar levels of throughput, leaving the TS-109 Pro behind, although it does begin to catch up as the file size drops.
A very similar story can be told for the result of the "general" test, too. 512MiB of RAM helps deliver good throughput for larger file sizes, but get down in size, and the TS-109 Pro, with 128MiB, can put up a good fight.
The N5200 Pro's performance is in line with what we'd expect, given our experience with the N5200 non-Pro and 1U4500.
Now, let's throw multiple workers at the box and see what happens to our throughput.
Keep the N5200 Pro busy and it'll keep pumping out the data - for large requests at least. Our "general" test uses much smaller request sizes, and evidently this means that the N5200 Pro has enough requests to deal with when there's just one worker attached to it.
SMB performance is up to scratch then, for the most part. However, it's a little perturbing to see a single-disk solution outpace the RAID-bearing beasts in some situations. In others, though, the Thecus is well out in front.
iSCSI performance
We'd hope to see a performance improvement over SMB when using iSCSI, to help justify the N5200 Pro's extra cost over the N5200.
This is the first iSCSI-capable product to be processed in the HEXUS Labs, so we have no others to compare it against. We can, however, offer up our thoughts on the performance advantage over the more traditional file-transfer method of SMB.
The N5200 Pro delivered under 10MiB/s reads over SMB when dealing with a 1GiB file. Using iSCSI, that jumps to a ninja 35.64MiB/s.
As the file sizes shrink, so too does the performance difference. But whichever way you look at it, iSCSI delivers a throughput boost. Win!
RAID test
Having tested the N5200 Pro when all is well, let's see what happens when the RAID goes boom...
A dead disk doesn't damage data rates... much. There is a bit of a hit when you come to repair the array, but it's not a lot.
The real hit is the amount of time it takes to rebuild the array - in the order of several hours for our five 750GB disks. Fingers crossed that another disk doesn't die in the meantime (or you could use RAID-6 if you're paranoid).
Performance thoughts
So, we've had our first taste of iSCSI and we like it!
It wasn't as painful as the old days of SCSI, where termination and device IDs produced misery and confusion. The benefit of providing block-level access to a client is clear - things happen quicker.
SMB performance remains good, if a little flaky for larger file sizes. All aspects considered, we're fairly impressed.
