In Part One we have reviewed characteristics and peculiarities of 2.5" SCSI storage drives Seagate Savvio 10K.1. Having analyzed the test results of physical characteristics and firmware operating specifics of these drives for reading and writing, we can expect from Savvio very interesting and ambiguous performance results in various applications. That's what we are going to find out.
Methods of testing and the main participants are described in the previous part of this article. To make it easier to analyze the results, hard drives go on the diagrams in the same order; three operating modes of Seagate Savvio 10K.1 73 GB are placed in three bottom positions. But as we shall make sure later, it does not mean that Savvio drives are the slowest among modern 10K storage drives. :)
Performance in applications
At first, let's find out how well the hard disks are optimized for multithreaded operations. I traditionally use NBench 2.4 tests for this purpose, where 100MB files are written on the drive and read from it in multi-threaded mode.
This diagram allows us to evaluate the efficiency of multi-threaded lazy write procedures of the hard disks in real (not synthetic, as on the diagram with the average access time) conditions when the operating system works with files. Hitachi, Maxtor and WD Raptor drives rule supreme in writing several simultaneous threads, while Savvio drives are generally slower than their "full-sized" Cheetah 10K counterparts just by 10% (considering their linear transfer rates being lower by one fourth!). We can see that their manner of handling several write threads is absolutely the same as Cheetah's. The mode of fixed segmentation in Performance Mode in this case has no effect on performance, though an updated firmware version reduces Savvio's speed a little (as we already saw before, caching large blocks gets a tad worse). However, Savvio is not the slowest drive in this test — it outperforms the Fujitsu drive.
The situation with Savvio is noticeably different for multithreaded reading — storage drives from Hitachi and Fujitsu are definite leaders here, while Seagate Cheetah 10K.7 in Server Mode stuck in the third place (though the same very model demonstrates very low results in Desktop Mode in this test). The obvious outsider here is another Seagate drive — Cheetah 10K.6. All the more gratifying that Savvio with Firmware 0001 demonstrates an average transfer rate on the level with such grands as Maxtor Atlas 10K V and WD Raptor, having outperformed the old Cheetah and the new model in Performance Mode. Unlike Cheetah 10K.7, Savvio does not demonstrate such a catastrophic performance drop in this task, when the storage drive goes into Desktop Mode. There is still a 3% drop. It's also interesting that with Firmware 0002 (where, as we remember from ATTO Disk Benchmark, caching large files for reading is obviously worse, probably due to a greater number of minimum possible segments) Savvio's multithreaded reading performance gets noticeably worse.
Now let's have a look how the hard drives fair in old but still popular Disk WinMark 99 tests from the WinBench 99 package. We carry out these tests not only for the beginning but also for the middle part (in terms of capacity) of physical hard drives for two file systems. The diagrams display average results. Out of doubt, these are not "profile" tests for SCSI drives. Having published these results, we don't try to determine the fastest professional storage drive, but rather pay heed to the test itself and those readers, who are used to evaluate performance of a storage drive by WinBench99 tests. But we can note to console you that these tests will demonstrate to a certain degree the performance of these enterprise drives in typical desktop tasks.
In this case, Seagate storage drives (except for an old Cheetah 10K.6) do not shine in performance, though Desktop Mode instead of Server Mode allows to raise their performance noticeably, by about 20%. For a consolation we can note that Savvio is practically on a par with full-sized Cheetah 10K.7 and even outperforms Hitachi Ultrastar 10K300 in Business test. However, if we don't use Performance Mode in Savvio, its average performance in High-End Disk WinMark is the lowest among modern 10K models. Though it can still be reduced, if we use new Firmware 0002 — a lag from the leader may reach twofold!
Newer complex track tests for the evaluation of desktop performance from PCMakr04 and C'T H2BenchW again push Savvio into an abyss of twofold defeat to the leaders from Fujitsu, Maxtor, and WD.
Nevertheless, the situation is not as bad as it seems, as Savvio is always faster here than the full-sized Cheetah 10K.7! It's great. Savvio's (as well as Cheetah's) performance grows instantly (by one third!) in Performance Mode, so that they already outperform the drive from Hitachi and start to catch up with the leaders.
Still greater gain from Performance Mode is demonstrated in H2BenchW test. Performance of both Seagate drives (by the way, Savvio is not worse than Cheetah 10K.7 here as well) grows more than by 60% (!), though it is still lower than in the old Cheetah 10K.6. However, in Server Mode this respected test shows that the modern storage drives from Seagate are nearly three times as slow as the leaders. These results cannot be explained by slower linear transfer rates, especially as Savvio is one of the best in terms of random access time. It means that the blame for these ungainly results lies solely with the firmware of storage drives as far as caching policies for reading and writing are concerned.
What concerns Adobe Photoshop swapping performance, Seagate Savvio is not very fast — it's evidently slower than all the other 10K models, including Cheetah 10K.7. Even Performance Mode can't help it catch up with the competition.
Intel Iometer tests
Enough of wearing our storage drives out with the tasks, which are not quite typical of their profile. Let's proceed to tasks, to which these storage drives are actually intended. We use special patterns in Intel IOmeter to imitate hard disks operating in various enterprise-class applications. At first — traditional popular patterns, offered by Intel and Storagereview.com — imitation of various servers (DataBase, File Server, Web Server) and a Workstation.
We shall publish only Seagate results as well as the results of Maxtor Atlas 10K V (as the fastest server 10K model), in order not to overcrowd detailed graphs. The diagram below will display the average results of all the other storage drives.
And here we are finally witnessing the first fruit of the successful design of Seagate Savvio — while Seagate drives don't shine in small queues (more typical of desktop applications), with the queue depth of 16, Savvio catches up with the current leader from Maxtor. In case QD=64, it obviously outperforms it! Although Maxtor's average random access time is even a tad better than in Savvio! That is its firmware obviously possesses some advantages in this respect — just remember Savvio's best random write result in H2BenchW (see Part One of the review). Both Savvio's firmware versions demonstrate practically the same results, but in Performance Mode the little storage drive suffers a severe performance drop at 8-64 queues. Moreover, Savvio is obviously faster than full-sized Cheetahs 10K — especially at a large queue depth.
Imitation of file-server operations shows a similar picture — Savvio manages to slightly outperform Maxtor drive at 64 queue (though Maxtor's advantage in linear transfer and seek rates is evident) and always keeps an advantage over Cheetah 10K. Performance Mode is contra-indicated here.
Our model under review fails to outperform the leader from Maxtor in web server pattern with absolutely no write commands (Savvio's advantage in write access time over the other storage drives does not affect the result). It's another proof that our conclusions are correct. But the situation is not as bad as it seems — Savvio's results in this test were not very good with Firmware 0001, it was outperformed even by all Cheetah 10K drives. But when we updated it to Version 0002, the situation in the web server pattern got cardinally better, Savvio not only managed to outperform all Cheetahs, but it also caught up with Atlas 10K V. Thus, we have finally found out the reason why Savvio programmers changed the policy of adaptive caching files for reading in the second firmware version — having degraded caching of large data blocks for reading, they evidently improved the read rate for small blocks (we have already noted that in the results of ATTO Disk Benchmark). That's especially noticeable at large queue depth for reading, where the gain from firmware upgrade may reach 35%!
According to the average server load, Savvio 10K.1 is currently one of the fastest storage drives for such tasks, having confidently outperformed the majority of 3.5-inch 10K SCSI drives. You are recommended to use Firmware 0002 and leave Performance Mode alone. :)
The situation in the workstation pattern is less peachy for Savvio — even though it gets a tad faster at large queue depth (thanks to Firmware 0002) and it works still faster in Performance Mode (unlike server patterns), it obviously lags behind the leaders from Maxtor and Fujitsu and even from a former hero of Seagate Cheetah 10K.6.
However, our little one goes on a par with Cheetah 10K.7 (having outperformed it only in PM mode). Here is another proof that caching policies in PM are slightly different in Savvio and Cheetah 10K.7.
Now let's proceed to simpler loads, such as random reading and writing of large and small files, which is also typical of these storage drives' profile.
The best drives at reading large (about 1 MB) files at random addresses across the entire disk are Fujitsu (not shown here), Savvio 10K.1 in PM mode, and old Cheetah 10K.6. The last two storage drives are almost twice as fast as Savvio in Server Mode as well as Cheetah 10K.7. It's only natural (see above) that Savvio (with Firmware 0002) obviously draws back in this test (the blue curve for this case is right under the green one for Cheetah 10K.7). So, despite the random access, this load is more suitable to Desktop-optimized storage drives.
It's also hard to compete with Fujitsu and WD storage drives in the Large File Write test, but Savvio looks quite good here, outperforming Maxtor Atlas 10K V, Seagate Cheetah 10K.7, and being slightly slower only than the old Cheetah 10K.6. Firmware updated, Savvio climbs down a little here, it works in PM exactly as in Server Mode (graphs are practically matching!). Let's see how it works with small files, where Savvio has much more trump cards.
It's again very difficult to compete with Fujitsu in Small File Read test, though Savvio is quite capable of reaching the top three places, if it uses Performance Mode. It works in a default mode (that is in Server Mode), its performance is low here, on the level of Seagate Cheetah 10K.7.
Unfortunately, the situation with writing small files isn't much better for Savvio — all the three modes demonstrate absolutely the same results, which are obviously better than in Cheetah 10K.7, but generally a tad worse than in Cheetah 10K.6 and especially than in such leaders as Fujitsu and Maxtor storage drives.
In the test, where large files are copied at random addresses across the entire disk (randomness is more typical of server than desktop loads as well) Maxtor storage drive is again beyond competition, having outperformed its competitors manifold. Savvio drive demonstrates moderate results here, being a tad faster than Cheetah 10K.7 (especially in Performance Mode).
A gap between the competitors in the Small File Copy test gets noticeably shorter, though the leaders remain the same. Savvio works faster here than Cheetah 10K.7, it has a chance to compete with its competitors, if it uses Performance Mode.
According to the geometric mean of the six previous graphs, modern SCSI 10K drives from Fujitsu and Maxtor take up the lead in these operations (random reading, writing, and copying of large and small files). They are followed by the old WD Raptor and Seagate Cheetah 10K.6. The fifth place is suddenly taken by Savvio 10K.1 in Performance Mode! It outperforms Cheetah 10K.7 (also in PM), Hitachi Ultrastar 10K300, and is nearly 1.5 times as fast as itself in Server Mode. In this case, new Firmware 0002 works a tad slower than old Version 0001. On the whole, it's a very good result for the storage drive, which is much slower than the others in linear transfer rates.
The other patterns complete the picture of Savvio 10K.1 performance.
Our model under review is a tad faster than Cheetah 10K.7 in the defragmentation imitation test, being slower than all the other drives (PM is beneficial here - it provides 23% performance gain).
What concerns streaming read/write in large and small blocks, our storage drive acts exactly as Cheetah 10K.7 — quite unpretentious, just another proof of its purely server mission to the prejudice of consumer applications. However, just a fact that it's actually no worse than Cheetah 10K.7 here, which has a much higher transfer rate, is praiseworthy.
Acoustic noise and power consumption
Double-platter Seagate Savvio 10K.1 is subjectively much quieter than all the other modern SCSI storage drives. It concerns active seek modes as well as idle noise. Though the rotation noise from platters is not that low as in the majority of modern desktop 3.5-inch storage drives with the spindle rotational speed of 7200 rpm, you will hardly leave it running at night in your bedroom. :)
Savvio's power consumption is significantly lower than in 3.5-inch 10K models — both due to a more compact spindle motor (lighter platters of fewer numbers) and a shorter bracket with heads, which requires less efforts to move. Savvio consumes very little at startup (see the details in our previous article), its typical power consumption in Idle mode is 4.5 W (according to specifications, it's 5.1 W), which is much lower than in any 3.5-inch ATA drive! In active seek mode, Savvio's power consumption grows to 9.1 W (though it must be 8.1 W according to the specifications), which is still much lower than in other 10K models (the closest counterpart among SCSI models is single-platter 74 GB Cheetah 10K.7 with its 12.5 W in seek mode; by the way, WD Raptor WD740GD demonstrates 11.3 W here). In case of typical intensive system operations with a storage drive (access time takes up 100%, but 20% of this time fall on the seek time) Savvio consumes about 7 W, while 74 GB Cheetah 10K.7 and WD740GD storage drives consume about 9 W and 10 W correspondingly.
So, Savvio evidently demonstrates a power saving advantage (with the same capacity and similar performance), which certainly favors their usage in 1U servers and blade servers. But if we estimate power consumption of a typical cage consisting of 8 Savvio storage drives, installed in two 5-inch bays of the server case (see the photo in the beginning of Part One of this article), we'll find out that even in case of unhurried operations (nearly Idle), the cage (total capacity of RAID 50 is about 440 GB) will require 8x4.7=37.6 W, while active operations will raise its power consumption to about 60 W. You should be ready for that! For example, a standard cage (similar in its capacity and performance) made up of four 147 GB SCSI storage drives (RAID 5, 440 GB) occupying three 5" bays is even less power consuming — 30-50 W. To say nothing of the price — only the drives in the first cage will cost $2400, while the second option will be within $1700-2000.
So, out of doubt, the new product Seagate Savvio 10K.1 is very interesting and possesses a number of irrefutable advantages, such as small dimensions and weight, high capacitivity (however, it's still higher in 300 GB 3.5-inch SCSI models), reduced power consumption and heat dissipation, quieter operation, higher efficiency in terms of the room it occupies, and many others. Savvio allows to cardinally review the approach to building compact servers (1U, blade) and data storage devices by offering the new quality of these systems, unreal before this storage drive appeared. Savvio can also noticeably improve scalability of more traditional medium-sized data storage systems. But the main Savvio's application so far is compact systems. As in case of larger data storage systems, modern 3.5" storage drives offer better prices (all other things being equal). Their capacity reached 300 GB already. Besides, even larger Savvio arrays would have problems competing with 15K drives in the high-performance segment.
Good news - according to our test results, Seagate Savvio 10K.1 drives generally offer higher efficiency than their full-sized Seagate Cheetah 10K.7 brothers. But Savvio drives still cannot compete with the best representatives of modern enterprise 10K models in performance. Savvio's firmware is quite queer. It provides excellent performance results for typical server loads, where Savvii are currently one of the fastest 10K models.
Thanks to the total sum of these features, Seagate Savvio 10K.1 definitely merits the Original Design award.
The long and successful life of these products will be also ensured by the fact that they are manufactured not only for the traditional parallel SCSI, but also for the deeply professional Fibre Channel as well as for the fashionable and promising Serial Attached SCSI.
At the same time, Savvio storage drives are not perfect. Among their noticeable disadvantages we can note their low capacity (for these days), quite low linear transfer rates, not very fast (relative to the potentially possible) seek mode, and some strange things about the firmware, which inherited the drawbacks of unhurried operation in many applications from Cheetah 10K.7. Dubious advantages include the current retail price — from $200 for the junior Savvio model (37 GB) and $300 and higher for the senior 74 GB model (for example, a single-platter Cheetah 10K.7 74 GB costs about $230; for $300 and something you can find a 147 GB SCSI model). However, the pricing policy for exclusive enterprise products is another story. We can only be happy that Seagate did not raise the price for Savvio still higher.
But we don't put a period to our review of Savvio 10K.1 yet. We hope to test SAS models in due course. We shall soon publish an article with Seagate Savvio RAID tests, as drive arrays are their main application.
Alex Karabuto (firstname.lastname@example.org)
November 15, 2005.
Write a comment below. No registration needed!