by Andrey Kuznetcov
11/08/2004 | 06:53 PM
Western Digital Company appeared the first HDD manufacturer to push SATA interface for server hard disk drives into the market. Not so long ago, their Enterprise Hard Disk product line acquired a new family of solutions called RAID Edition. In fact this mere phrase implies that the HDDs from this family should be specifically adapted for their optimal performance within a RAID array.
<%BANNER[article]%>Since the name of the product family cannot be considered a sufficient reason to regard the product as a more efficient one for the given task than any other product, we are facing a very important and interesting task: to find out what are the actual differences between these new HDDs and their counterparts with similar technical specifications.
In other words, we have to check if announcement of the new hard disk drive family was just a marketing trick of the well-known HDD maker, or they have really released new products with high reliability and faster speed.
At first let’s check out what Western Digital Company says about its RAID Edition HDDs. As we got it from the info available on the company’s official web-site, their major goal was to design a storage solution boasting highest reliability, which would meet the Enterprise devices standards. In reality it means that the hard disk drives should be able to work reliably in a 24x7, continuous duty cycle environment.
The manufacturer guarantees 1 million hours MTBF (Mean Time Between Failures). RAID Edition HDDs support Time Limited Error Recovery technology (TLER), which serves to improve compatibility with RAID adapters and to prevent the HDDs from failing during error correction. Of course, these drives also support the time-tested Data Lifeguard Tools technology. This software utility works with the embedded Data Lifeguard hardware features (including shock protection, an environmental protection system, and embedded error detection and repair features) to make hard drive installation, management, and diagnostics simple and worry-free. Here I can also add that the RAID Edition HDD family uses the traditional fluid dynamic bearings that not only improve the HDD acoustic parameters (such as noise and vibration), but also improve hard drive reliability and reduce heat. The high performance of these drives is guaranteed by the 7,200rpm spindle rotation speed and 8MB buffer.
RAID Edition HDD family includes solutions with the storage capacity of 120GB, 160GB and 250GB. The HDDs are designed either for the classical EIDE interface or for the new SATA. In this review we are going to take a closer look at the top model from this family: the 250GB WD2500SD solution. Besides the already mentioned above technical specifications of this product, I would like to add that the claimed seek time is 8.9ms, the average wait time is 4.5ms, and the track-to-track seek time is 2.0ms. The internal data transfer rate is 61MB/s, while the external data transfer rate via the SATA interface can reach 150MB/s.
This solution is selling for approximately $215.
We will be comparing the new hard disk drive with the solution from Western Digital featuring similar technical specifications and claimed parameters: WD2500PD.
We will run all our tests in two work modes. In the first case we tested single hard disk drives performance for WD2500SD and WD2500PD. And in the second case we tested RAID 0 arrays of two HDDs. The WD2500PD HDDs are supplied with the default “quiet seek” set, however, we activated “fast” mode for them, because RAID Edition HDDs are shipped with fast seek as default.
RAID 0 arrays were built with the RAID controller integrated into the ICH5 South Bridge of the mainboard chipset.
The testbed was configured as follows:
We used the following benchmarking software:
With the help of the DataBase pattern we measured how fast the tested HDDs could process requests of various queue depth and write requests of various size. During our tests we used 8KB data blocks with random address.
To make the obtained results more illustrative we built three graphs for the queue depth equal to 1, 16 and 256 requests.
We will start the discussion with the linear workload. The queue depth=1 allows the older WD2500PD drive to outperform its newer rival. Its advantage is clearly seen in the standard mode as well as in case of an array. However, if in the first case the maximum advantage is achieved when the number of write requests varies between 20% and 90%, then in the second case the significant advantage is only achieved when the writes share reaches at least 70%.
On the next graph we see the HDDs performance under the workload of 16 outgoing requests.
Again the WD2500PD drive is ahead of the WD2500SD in both cases: in case of a single HDD as well as in case of a RAID array. It is evident that WD2500SD is ahead of its rival only in one mode: during RandomRead.
The third case we are going to consider corresponds to 256 requests queue depth. Just like in the previous case WD2500PD appears faster. WD2500SD HDD outperforms its counterpart only in case of 0 writes.
Well, it is evident that WD2500SD hard disk drive cannot boast as aggressive lazy writing, as its “desktop” fellow. Also I dare suppose that WD2500SD features less aggressive read ahead, which allows it to prove faster during random reading.
In order to better study all the nuances about the hard disk drives performance, we tested the sequential read and write speeds, which showed how quickly the testing participants can process ordered requests with linearly growing address. The obtained results are given in the table and on the diagrams below.
Let’s start our discussion with the sequential reading. The diagram above shows that WD2500SD is the winner. It outperforms its opponent as a single drive as well as in the array. It manages to achieve its maximum advantage when working with 4Kb and 8KB data blocks. The overall advantage in sequential read speed is seen when working with the data blocks from 2KB to 1024KB.
The next diagram reflects the sequential write speed of the testing participants. Here the laurels are again won by WD2500SD HDD. This solution is more efficient in both cases than its rival, especially when we have an array. It yields a little bit only in the regular work mode when processing 1KB and 2KB data blocks.
With the help of the WorkStation pattern we imitated the work of the hard disk drive as if it were working in a typical workstation, when the workload is limited by the maximum request queue depth of 32 requests.
In the first diagram you see the situation when the entire storage capacity of the hard disk drives was tested. WD2500PD is generally faster here. The new WD2500SD HDD appears defeated in both cases.
Since the advantage of WD2500PD solution in terms of lazy writing has already revealed itself in DataBase pattern, the results obtained for WorkStation pattern were very easy to predict. Large share of write requests, which is pretty typical of this pattern, doesn’t leave a single chance for WD2500SD: neither in the RAID array, nor in the single-drive mode.
Let’s see how the disks ratings in this pattern compare. To calculate the ratings we apply the following formula, which you should already be familiar with:
Performance = Total I/O (queue=1)/1 + Total I/O (queue=2)/2 + Total I/O (queue=4)/4 + Total I/O (queue=8)/8 + Total I/O (queue=16)/16 + Total I/O (queue=32)/32
The diagram shows that both integral parameters of the overall product efficiency are higher by WD2500PD.
One more situation we decided to check out when running the benchmarks for the WorkStation pattern deals with involving only the first 32GB of storage space of each tested hard disk drive. Again, WD2500PD looks more convincing here, as it outperforms WD2500SD in the regular mode. However, when we have an array involved, the latter manages to beat WD2500SD only in case of high requests queue depth.
The diagram with overall performance indexes demonstrates pretty unexpected results, as it might seem at first glance. WD2500PD hard disk drive wins only in three cases out of four. It yields to WD2500SD when working in an array, due to higher importance of the results obtained for low queue depths.
FileServer and WebServer patterns were used to test hard disk drives in the environment typical of storage subsystems in file- and webservers.
Let’s see what results the hard disk drives show in the FileServer pattern. WD2500PD appears to be somewhat faster. Its advantage over the rival grows even bigger as the number of requests in the queue increases.
The diagram shows integral performance ratings of the testing participants, equal to average requests processing speed under all sorts of workload. WD2500PD HDD is not dramatically faster than the opponent but its advantage is still evident in both cases.
The results of our tests in the webserver pattern suggest that WD2500SD should be regarded as the winner. The diagram also shows that this hard disk drive appears faster than WD2500PD regardless of the requests queue depth.
As you remember, WD2500SD was faster than the opponent in DataBase pattern when we had random reading requests. Therefore, it is not at all surprising that in WebServer pattern these drives feel perfectly at home.
On the diagram with overall performance indexes, which we calculated the same way as in the previous case, we see an evident victory of WD2500SD HDD.
Having run a few synthetic IOMeter benchmarks let’s turn to the results of the popular WinBench99 test set. Disk WinMark results in the tables below are highlighted with red if they are the lowest and with blue if they are the highest. Take a look:
WD2500PD
WD2500SD
WD2500PD RAID 0
WD2500SD RAID 0
At first let’s check out the performance of our HDDs when they have been formatted for FAT32 file system:
This time we decided to discuss the results of Business Disk WinMark and High-End Disk WinMark separately on purpose. Let’s start with the first one.
In both cases we considered here WD2500SD is faster than WD2500PD. We could point out however that its performance is faster in case of a single drive compared with the results obtained for the array.
The next diagram refers to High-End Disk WinMark. Finally we can see WD2500SD win a brilliant victory: it appears much faster than the rival when we test single hard disk drives, as well as when we test RAID arrays.
Now let’s see what we get when the drives are formatted for NTFS file system. On the first diagram we have the results for Business Disk WinMark. WD2500SD is considerably faster here although in the array it slowed down a lot.
The efficiency of our hard disk drives in High-End Disk WinMark test gives us another reason to state that WD engineers’ efforts were not wasted: WD2500SD hard drive is faster than WD2500PD. You can get an idea about this advantage when you look at the speed the single WD2500SD demonstrated compared to WD2500PD working in an array.
The last benchmark we are going to use in this test session is X-bit’s own FC-Test. We tested our hard drives in a usual way with the help of five patterns differing from one another by the number of files involved and their size. Overall we performed four types of operations. First we created files in the tested array on the first 32GB section (Create). Then we read the entire set of files from the drive (Read). Then we copied this set of files within the same section (Copy Near). And in conclusion we copied the whole set of files from the first 32GB section into the second section of the same size (Copy Far).
Let’s start with FAT32 file system.
The first diagram shows how quickly the files were created when the HDDs were formatted for FAT32 file system. We can see that WD2500SD is far ahead of the rival here in all five patterns and in both modes tested. The performance difference is especially noticeable when we compare the arrays’ performance.
File reading results into the same flawless victory of the WD2500SD over WD2500PD. Just as in the previous case the gap between the two testing participants grows bigger when we have arrays working.
Copying files within the same section makes WD2500SD look even better. It appears almost twice as fast as the opponent in most patterns. This is actually true for both testing modes.
The last diagram shows how quickly the drives allow copying the file sets from one section to another. Again WD2500SD is ahead in all patterns.
Now let’s check the situation with NTFS file system.
WD2500SD again manages to create files faster than WD2500PD. Although here the performance advantage of our leader is no longer as great as in case of FAT32 file system. But again the maximum performance advantage can be achieved when we have an array of HDDs.
The next diagram shows the read speed of the testing participants. WD2500SD is an indisputable leader here in all patterns. It appears the most efficient when working in an array again.
File copying within the same section shows that WD2500SD is considerably faster than WD2500PD, especially in patterns with larger files.
The last diagram confirms once again how fast the new WD2500SD actually is. When we copy files from one section to another, our hero again wins the laurels being far ahead the WD2500PD. As usual, it advantage is most evident in patterns with larger files.
As we have already mentioned in the beginning of this review, Western Digital engineers primarily aimed at increasing the reliable functioning time of their RAID Edition drives under heavy and intensive workload. Unfortunately, we cannot yet state if they managed to accomplish this task 100%, because we couldn’t carry out such a long-term testing session this time, so that we could make any serious conclusions about their reliability.
The only thing we can state with all certainty is that during this one-week test session none of the WD2500SD drives failed. And so did none of the WD2500PD drives, too :)
However, we managed to study carefully the features of the new WD2500SD responsible for its high efficiency and fast performance in single-drive mode as well as in a RAID 0 array. So, we have every right to say that WD did a great job in creating a successful product with principally new product features.
Despite the fact that WD2500SD lost in a few synthetic IOMeter tests, it managed to show its best in less synthetic programs such as WinBench99 and FC-Test. These results seem very optimistic and allow us to say that potential buyers of this product will get what they are looking for. All this together with higher claimed reliability makes this drive a really worthy choice.