by Aleksey Meyev
01/07/2009 | 10:29 AM
In our previous article about RAID arrays we built them out of 400GB hard disk drives. Now we want to check out 500GB products as well.
The increased storage capacity (for a RAID0 it now equals the pretty-looking number of 2 terabytes) is only one aspect of the issue. More importantly, 500GB drives generally have higher recording density and, consequently, higher speeds of processing sequential data. While all the HDDs in our previous review were based on three platters, this test session includes a few HDDs with only two platters and four heads although their capacity is higher. We want to find out how they perform in RAID arrays and how much faster the new higher-density models are in comparison with the older generation. And there is also the factor of firmware that is being improved by the developers from model to model. How effectively will the RAID controller work with the new HDDs? Will the simultaneous use of the controller’s own cache and the HDDs’ cache memory be efficient? These questions do not have obvious answers and it is better just to check them out in the practical way.
500 gigabytes is far from the maximum storage capacity available today but such HDDs feature a low cost of storage. Thus, they offer an opportunity to build a large and inexpensive array of the high-performance RAID, fault-tolerant RAID5 or in-between RAID10 type.
Of course, we cannot compare all 500GB hard disk models available on the market. There are too many of them for that. The test session would take too much time and the resulting report would be too long for anyone to digest. Therefore we will only take eight models from different makers.
The first two HDDs come from Hitachi’s and Samsung’s desktop series. They are not new (newer models with two and three platters are available already) but are inexpensive, so thrifty users are going to consider them in the first place.
The next pair comes from Seagate: they are similar to the previous two models in characteristics. The difference is that the Barracuda 7200.10 is a desktop HDD whereas the Barracuda ES is meant for professional applications, i.e. for high-load 24/7 servers. You will see shortly if there is any difference in performance between these two models.
The next pair represents Seagate’s newer 7200.11 and ES.2 series. The enterprise and desktop models both feature increased recording density and each achieves the capacity of 500 gigabytes with only two platters. The cache memory of both models is increased to 32MB.
And the last pair in this review is from Western Digital’s enterprise RE2 and RE3 series. The newer RE3 disk has higher recording density and, like the abovementioned drives from Seagate, gets along with only two platters. As opposed to them, it has 16 megabytes of cache, like the older products in this review. So, we will be able to see what advantage the larger cache gives to Seagate’s products.
The following benchmarks were used:
Testbed configuration:
Like in our previous reviews, we built RAID arrays out of the HDDs using one and the same controller. Although the results are thus bound to the particular controller’s characteristics, we don’t have to guess what part of them are due to the HDDs and what, to the controller. The controller was an Areca ARC1220 installed into a PCI Express x8 slot.
The sets of HDDs were installed into the standard boxes of the SC5200 system case and fastened with four screws to the bottom panel. Each set consisted of four samples of a HDD model (all from the same batch). We built the following arrays out of them:
These array types differ greatly in terms of load the controller puts on the HDDs even when they perform the same operations. RAID0 ensures maximum performance, RAID10 adds mirroring to RAID0, and RAID5 transforms each write request to the array into four requests to the HDDs in it.
The controller was set at the Performance mode for maximum performance during the tests. This mode allows deferred writing and look-ahead reading for both the controller (in its own buffer memory) and the HDDs. Thus we will see how efficient the HDDs’ firmware algorithms are when utilized in a RAID array.
When we talk about the results of a particular HDD below, we will mean the results of the array built out of the HDDs of that model.
In this test IOMeter is sending a stream of requests to read and write 512-byte data blocks with a request queue of 1 for 10 minutes. The total number of requests processed by the disk array is over 60 thousand, so we get a sustained response time that doesn’t depend on the cache buffer size. The array cannot store all the requests in its cache and deliver a fantastically high performance.
Well, it looks like we’ve got a Samsung that works in quiet mode (with slowed-down heads), just like in the previous review. But in our review of RAID arrays based on 400GB drives we could only observe a negative effect from the quite mode under low loads in IOMeter:Database. So, it only makes the test session the more interesting. Take note that the quiet mode has no effect on the drive’s write response time – the caching of write requests neutralizes it effectively.
Let’s get back to the results, though. First of all, we can note the obvious thing once again: the array type does not influence the array’s read response time. This parameter is only determined by the characteristics of the HDDs the array is built out of.
Writing is a different story as it is affected by the caching algorithms. The situation is even more complicated since we’ve got two caches: in the controller and in the HDDs. The Western Digital RE3 is the best in this test. Its write response time is excellent irrespective of the array type.
The Seagate drives produce curious results here. The two new models with large cache are worse than the older models in RAID0. So, better specifications do not always mean better performance. The old Seagate ES is considerably superior to its desktop counterpart. The new series feels much better than the older one in RAID10, and the response time of the desktop and enterprise models is almost the same in each pair. In RAID5 the new series is better than the older one but the new enterprise model is inferior to the new desktop model whereas within the older series it is just the opposite.
The Hitachi and the Samsung are very poor at writing in RAID5. The Western Digital RE2 produces mediocre results all the time.
In the Database pattern the disk array is processing a stream of requests to read and write 8KB random-address data blocks. The ratio of read to write requests is changing from 0% to 100% (stepping 10%) throughout the test while the request queue size varies from 1 to 256.
You can view tables with the results by the following links:
We’ll discuss the results for queue depths of 1, 16 and 256.
At high percentages of reads the slowed-down Samsung cannot compete with the others under low loads in RAID0. The Western Digital RE3 is good again, boasting very effective deferred writing. The Seagate Barracuda 7200.11 (it is quite faster than its enterprise counterpart), Hitachi and Western Digital RE2 stand out among the others.
As soon as there is a queue of requests, the Samsung joins the others and delivers modest performance. The Western Digital RE3 is splendid still, although the Seagate 7200.11 is competitive to it at high percentages of reads. The Seagate ES.2 behaves somewhat oddly: it is as fast as the leaders at pure random reading but slower than them at other loads. Its behavior is different from that of its desktop counterpart and it has a sudden performance hit at pure random writing. There must be some flaw in its firmware algorithms.
The old Seagate 7200.10 is a loser in this test. It is slower than every other drive including its enterprise counterpart.
When the load is very high, Western Digital claims victory in RAID0. The Seagate 7200.11 is confidently second while its enterprise counterpart’s performance fluctuates wildly depending on the ratio of reads to writes, which is not normal.
The Seagate Barracuda 7200.10, the Samsung or the Hitachi sink to last place alternately depending on the ratio of reads and writes, yet the old desktop model from Seagate seems to be the overall loser.
The Seagate doesn’t feel good at reading in RAID10, too. It accelerates fast when the ratio of writes grows up but can’t reach the top position due to a slowdown at a very high percentage of writes.
The Western Digital RE3 is in the lead again. The other HDDs are similar to each other in this test, except that the old drives from Seagate are somewhat worse.
When the load grows higher, the picture resembles what we have seen with the RAID0 arrays: the Western Digital RE3 is in the lead, followed by the new desktop Seagate which is characteristically faster than its enterprise counterpart. The Seagate 7200.10 is the loser again.
We don’t see anything new at the maximum load. We’ve got the same leaders and the Seagate ES.2 still has a fluctuating performance except that it is now all right at high percentages of writes. The Seagate Barracuda 7200.10 is on the losing side, again. Interestingly, as opposed to Seagate’s new series, in the old series the enterprise model is faster than the desktop one, especially at high percentages of writes.
RAID5 is the most complex type of an array because each write request to the array is translated into four requests to the HDDs. Of course, the Samsung with its slowed-down heads is again much slower than the others at the queue depth of 1. We have the same leaders: Western Digital RE3 and Seagate 7200.11. Third place is contested by the Western Digital RE2 and Seagate ES.2. This is an odd contest as the Seagate ES.2 has such advantages as higher recording density and a larger cache, but Western Digital’s enterprise product seems to be very efficient at utilizing what resources it has.
The Samsung joins the others at the higher load, and all the three-platter drives, excepting the Western Digital RE2, go close to each other. The performance of the Hitachi and Seagate ES.2 grows up suddenly at pure reading. Note also that the Seagate ES.2 behaves more predictably in RAID5.
There are no changes in the standings at the maximum load.
Summing this test up, we must note the excellent results of the Western Digital RE3. If it is also that good with sequential operations, it is going to be the leader of this review despite the last-generation Seagate models having more cache memory and the same recording density.
The Western Digital RE2 must be noted, too. It is obviously the best among the three-platter products.
The Seagate 7200.10 seems to be the worst HDD in this test session so far. Curiously, its enterprise counterpart performs much better.
The disk arrays are tested under loads typical of servers and workstations. The names of the patterns are self-explanatory. The Workstation pattern is used with the full capacity of the array as well as with a 32GB partition. The queue depth is limited to 32 requests in the Workstation pattern.
We will be discussing diagrams but you can click the following links for the detailed results and performance ratings:
There is no point in discussing the Web-Server diagrams individually because the numbers change but the standings do not. So, our comments will refer to the behavior of every type of the array under this reads-only load.
The Western Digital RE3 is obviously the best at long queue depths. The two new drives from Seagate share second place and challenge the leader at short queue depths. The Western Digital RE2 stands out among the lower-density disks, and the Hitachi is good, too. The Seagate 7200.10 is the worst drive here (its enterprise counterpart is much better), but the Samsung is poor as well. We should keep it in mind, however, that the Samsung has the same specs but slowed-down heads.
There are differences in the File-Server pattern due to the presence of write requests. Therefore we will discuss the diagrams separately.
Quite expectedly, the Western Digital RE3 and the Seagate Barracuda 7200.11 take first and second places when working in RAID0 (and, running a little ahead, in the other types of the array). It is not so clear with third place. Due to the performance fluctuations we have seen in the Database pattern the Seagate ES.2 is not always third but loses occasionally to the lower-density and older Western Digital RE2. This serves to demonstrate that firmware algorithms are very important. The next contesting pair is the Hitachi and the Seagate ES. The former is better at low loads and the latter, at high loads. Last place goes to the Seagate 7200.10 although the Samsung is almost as slow as it at long queue depths.
The overall picture with the RAID10 arrays is roughly the same as with the RAID0 arrays but the Seagate ES.2 is now confidently third.
When it comes to RAID5, there is some fighting again with the same participants. Like in our previous reviews, the Hitachi expresses its dislike of this type of the array and joins the group of losers. The Samsung struggles with RAID5, too. It is the overall worst drive in this subtest.
The Seagate 7200.11 catches up with the Western Digital RE3 in RAID0 under Workstation load. Thus, they share top place here. This combination of load and array type seems to be difficult for Western Digital’s products: the RE2 is rather slow at medium queue depths, so the Seagate ES.2 easily makes to the top three.
The Samsung is poor at short queue depths whereas the Seagate 7200.10 is last at long queue depths.
When we switch to RAID10, the RE3 is in the lead again while the RE2 beats the ES.2 at short queue depths. There are no changes among the losers: the Samsung is poor at short queue depths and the Barracuda 7200.10 is in last place at long queue depths.
The leaders do not change with RAID5. The Samsung and the Hitachi show their dislike of this array type at short and long queue depths, respectively.
When the test zone is limited to a 32-gigabyte partition, the standings change considerably. For example, we’ve got four drives fighting for top place in RAID0: the Western Digital RE3 and the Seagate 7200.11 are now challenged by the Seagate ES.2 and the Hitachi. The latter has found a suitable subtest at last! The Samsung now competes with the Western Digital RE2 and Seagate ES at short and long queue depths, respectively. The Seagate 7200.10 is left alone in its miserable last position.
Take note of the behavior of Western Digital’s products. They feel much better than their opponents at very long queue depths, delivering a nice performance boost then.
The RE3 regains its leadership in RAID10. Then, the Hitachi and the RE2 go neck and neck with Seagate’s two new models at short queue depths. Their higher recording density doesn’t help at all here. It is only at a much longer queue depth that the Hitachi and ES.2 (yes, the ES.2 rather than the lower-density RE2) fall behind the two others, the ES.2 being the worst drive among the four.
Western Digital’s products feel at ease in RAID5 when the test zone is limited: the RE3 is unrivalled. The RE2 competes with the new drives from Seagate and beats them at medium queue depths. Interestingly, the Hitachi is not so bad with this array type now. It is the Samsung and the 7200.10 that are the slowest drives here.
IOMeter is sending a stream of read and write requests with a request queue depth of 4. The size of the requested data block is changed each minute, so that we could see the dependence of the array’s sequential read/write speed on the size of the data block. This test is indicative of the highest speed the array can achieve.
You can click the following links for tabled results:
When reading from the RAID0 arrays, the three-platter models are all on the losing side. The Samsung is the best among them, the Hitachi being the worst one. The new two-platter drives are over 30% (and the leading Western Digital RE3 about 50%) faster than their three-platter opponents.
It is all clear with the losers which are all of the three-platter models. The results of the RE3 indicate that data is being read from both disks in a mirror, which leads to such an impressively high performance. Take note that this speed is achieved with 64KB data chunks, which is a very frequently used size.
It is all nice and tidy with RAID5: the three-platter drives are slower and go close to each other. The new HDDs enjoy a nice lead over them, the RE3 still being the fastest among them. Besides excellent firmware algorithms and fast read/write heads, the RE3 seems to have platters with somewhat higher recording density than the other HDDs have. We will check this out in WinBench’s data-transfer graphs.
It is the controller and its memory that affect the speed of writing in RAID0. The higher-density drives have but a small advantage. The Hitachi is the only obvious loser. The controller doesn’t work efficiently with it, and its top speed is much lower than that of any other HDD.
The HDDs split up into two groups when working in RAID10 arrays. Take note that the higher-density HDDs take the lead as soon as 4KB data chunks. It means that their advantage is going to be conspicuous in typical disk usage scenarios.
Writing is a traditionally difficult trial for RAID5 arrays. The combination of the controller’s and HDDs’ algorithms must be effective for the array to show good performance. The RE3 and the Barracuda 7200.11 are still in the lead, though. The Samsung and the ES.2 are good, too. The Hitachi and the Barracuda 7200.10 are slow.
The multithreaded tests simulate a situation when there are one to four clients accessing the virtual disk at the same time – the clients’ address zones do not overlap. We will discuss diagrams for a request queue of 1 as the most illustrative ones – the speed doesn’t depend much on the number of applications at a request queue of 2 and longer.
You can click the following links for full results:
It is yet another test won by the Western Digital RE3. It is the best at multithreaded reading in RAID0. Its performance hit with multiple data threads is small.
Seagate’s new series drives behave in an odd way. They did not like to switch from one to two threads but improved their performance when there were even more threads to process.
The Western Digital RE2 and the Hitachi are the best of the older drives here.
The RE3 is just excellent when reading in a RAID10 array. It not only reads from two disks at once but also improves its performance when processing an even number of threads. The controller ideally distributes the read threads among the different drives in the mirrors.
The other HDDs behave like in the RAID0 arrays: the new models from Seagate again improve their speed when switching from two to three threads and take second and third places. The RE2 and the Hitachi are still the best among the three-platter products.
There are no changes when we switch to RAID5.
The large total amount of cache memory allows to perform multithreaded reading in RAID0 with a minimum loss irrespective of the number of threads. There are a couple of interesting things, though.
First, the Hitachi and the Seagate 7200.11 are the only drives to switch to multiple threads without any loss of speed. Interestingly, the enterprise counterpart of the latter model is among those two HDD that suffer the heaviest performance hit (the Western Digital RE3 is the other one; it has weak spots after all).
It is different with the RAID10 arrays: the RE3 regains its leadership whereas the two new products from Seagate are no good at multithreading. Both are more or less good with two threads but slow down with more threads.
There are no obvious losers with RAID5, but we have two leaders. The RE3 is a traditional leader already while the Samsung is a new face at the top.
For this test two 32GB partitions are created on the RAID array and formatted in NTFS and then in FAT32. After that a file-set is created of the hard disk. It is then read from the disk, copied within the same partition and then copied into another partition. The time taken to perform these operations is measured and the speed of the array is calculated. The Windows and Programs file-sets consist of a large number of small files whereas the other three patterns (ISO, MP3, and Install) include a few large files each.
This test produces too much data, so we will only discuss the results of the Install, ISO and Programs patterns in NTFS which illustrate the most characteristic use of the arrays. You can click the following links for the full results:
Let’s discuss the NTFS results first.
Western Digital’s drives are both ahead of the others when creating files on a RAID0 array. They are especially good with small files. The Samsung and Hitachi are also good but are too sensitive to the average file size. The two new products from Seagate are a disappointment. Their higher recording density does not show up, so their firmware obviously needs improvement.
Writing in RAID10 does not differ from writing in RAID0: the standings are the same except that the RE3 has an even bigger advantage over the others.
There are some changes in RAID5. The Hitachi falls out of the leading group and the Samsung is only good with large files. The Seagate 7200.11 performs much better here (but its enterprise counterpart is slower).
It is only with very large files that the HDDs differ much in RAID0 arrays. The three HDDs with higher recording density are all in the lead, the Barracuda 7200.11 being ahead of the RE3.
The Samsung and the Hitachi are alternately in the lead with small and medium-sized files but the higher-density products are ahead with large files, the RE3 obviously reading from both disks in a mirror again.
There is nothing interesting when we switch to RAID5: the HDDs with high-density platters only go ahead with large files.
Copying within the same partition on a RAID0 is like reading: the trick is in having high-density platters and reading large files.
The picture does not change much in RAID10: the same three drives are ahead with large files. The Hitachi and the Samsung show their speed with small and medium files, though.
The Seagate ES.2 falls behind the leaders in RAID5 but the Samsung is surprisingly good with large files. The Samsung behaves oddly in RAID5: it is good at processing small files but does not like to work in RAID5 under server-like loads.
The standings are overall the same in the copy far test.
Here are the arrays’ data-transfer graphs recorded in WinBench 99:
The following diagram compares the arrays by the read speed at the beginning and end of the partitions created on them.
Everything is quite logical here except that the results of the Seagate ES.2 in RAID5 are lower than those of its desktop counterpart. As for the Western Digital RE3, we can say that its recording density is indeed higher than that of its opponents. Anyway, recording density along would not have helped it win most of our tests (you can remember that the three-platter Western Digital RE2 was often as fast as the two-platter Seagate ES.2, for example).
First of all, we would like to note the excellent performance of the Western Digital RE3. Thanks to high recording density, fast mechanics and effective firmware algorithms, this HDD delivered the highest performance in most of our tests. This product is a good development of Western Digital’s server product series but its predecessor Western Digital RE2 is quite good, too. The RE2 was the best among the three-platter models in our server tests and occasionally challenged its two-platter opponents with larger cache.
Seagate’s products leave an ambiguous impression. On one hand, the new generation is more effective than the older one and this can hardly be explained by the increased recording density alone. On the other hand, it is unclear why the enterprise model is slower than its desktop counterpart in RAID arrays and why it has firmware flaws that showed up in IOMeter: Database. So, Seagate still has a lot of work to do yet.
Samsung and Hitachi have got some work to do on their firmware, too, if they care about this market sector because RAID arrays are not used only for processing small files. Their HDDs have strong points, but they are not so easy to see.
Generally speaking, it is good that hard disk drives develop in other ways than just towards higher recording density. The performance of the Western Digital products in this test session proves once again that good firmware contributes a lot to the resulting performance of an HDD.