วันเสาร์ที่ 7 สิงหาคม พ.ศ. 2553

3.5'' Vs. 2.5'' SAS HDDs: In Storage, Size Matters : Conclusion

Clearly, even the latest 2.5” Ultrastar C15K147 doesn’t stand a chance against the 3.5” Ultrastar 15K600 if you look at performance. This is because of the 3.5" model's higher data density and more aggressive performance. However, the 2.5” drive shines when it comes to power consumption, requiring half or even less power than the 3.5” drives, while performance differences aren’t even remotely close to the 50% or higher difference in power consumption.


In general, 2.5,” 15,000 RPM SAS drives will deliver 25% less performance than 3.5” drives, but efficiency increases at least twofold. Our results here are specific to Hitachi's drives, but we would expect similar results with Fujitsu or Seagate drives, although the performance results would have been different.

The performance "issue" with 2.5" drives can be managed through adjusting the drive count. Two 2.5” SAS drives in RAID will outperform one fast 3.5” SAS drive in every workload. At the same time, the power required won’t be higher and capacities may be similar.

If you take our results to the next level, you can easily imagine that a simple RAID with four 3.5” drives could be replaced by six to ten 2.5” drives. These would not consume more power than the 3.5” models, but resulting performance would be 50% to 2.5x higher, depending on the specific configuration. Meanwhile, power consumption will remain below the level of the 3.5” RAID.

All that’s left to consider is cost, and this has to be contemplated carefully. Performance storage may offer the best cost effectiveness when using a small number of SSDs, but as soon as you need capacity, flash memory might quickly become too expensive. That's when you’ll turn to arrays with 15K drives. For best price/capacity, 3.5” drives at 7,200 RPM are probably the most sensible choice unless you also need a certain performance level. In that case, you'll find best compromises with 2.5” hard drives.

3.5'' Vs. 2.5'' SAS HDDs: In Storage, Size Matters : Benchmark Results: Temperature, Power, And Efficiency


Surface temperatures are lowest on the new 600GB 3.5” drive, probably because it's capable of dissipating heat most efficiently over its entire surface. The predecessor was the hottest product in this analysis, and the 2.5” drive rests in the middle. We assumed that the 2.5” drive would win the temperature test, although ultimately the differences don’t matter much. All of these drives require active cooling.


The 2.5” drive only requires one-third of the power consumed by the 3.5” drives. This is extremely important in server environments, where hundreds of drives can be required for top performance, capacity, and redundancy.


Shifting to a sustained HD video stream doesn't change 3.5” drives' power consumption much. They're already at a pretty high level. The 2.5” drive still requires less than half the power for this task.


Power efficiency is best on the 2.5” drive because it blends speed (while not being the fastest) with only half the power of the 3.5” drives.


Lastly, there's workstation I/O performance per watt. The 2.5” drive is even better in this test since the power required to deliver peak I/O performance is much lower than on the 3.5” drives.

3.5'' Vs. 2.5'' SAS HDDs: In Storage, Size Matters : Benchmark Results: PCMark Application Performance

As always, I'd like to point out that PCMark Vantage isn’t a server or workstation benchmark, as it doesn’t reflect typical application scenarios for these hard drive products. However, it's still a decent tool for exploring raw performance.

The new Ultrastar 15K600 wins in all subtests. The older Ultrastar 15K450 and the 2.5” C15K147 run head to head, but there are differences between one benchmark section and the next.

3.5'' Vs. 2.5'' SAS HDDs: In Storage, Size Matters : Benchmark Results: I/O Performance And Access Time


The differences in database I/O performance aren’t huge, but they are noticable. The 2.5” drive does well here.


File server performance isn’t the Ultrastar C15K147's strong suit. The two 3.5” drives deliver better performance in this test.


We found similar results for the Web server test.


Lastly, the workstation test pattern confirms that the 3.5” drives are superior in I/O performance when relying on an individual drive. However, space and power consumption savings enable administrators to operate twice as many 2.5” drives as 3.5” options. In such configurations, the 2.5” array will always outperform 3.5” setups on performance, capacity, and efficiency. Cost might be an issue, but you can always look for the best price/capacity/performance sweet spot.

Access Time


Access times are slightly longer on the 2.5” drive.

3.5'' Vs. 2.5'' SAS HDDs: In Storage, Size Matters : Benchmark Results: Throughput And Interface


Seeing 115 to 196 MB/s sequential read throughput is pretty impressive. The 2.5” Ultrastar C15K147 is actually faster than the 450GB 3.5” Ultrastar generation. Although SSDs provide higher peak and average transfer rates, hard drives are still more reliable when it comes to delivering consistent performance at certain levels.


Our Iometer streaming read test confirms that the latest 15,000 RPM 2.5” drive is as fast as the last generation in 3.5” model.


Interface performance reflects the peak bandwidth when reading from or writing into the drive’s cache memory.

3.5'' Vs. 2.5'' SAS HDDs: In Storage, Size Matters : Test Setup And Transfer Diagrams

System Hardware
HardwareDetails
CPUIntel Core i7-920 (45 nm, 2.66 GHz, 8MB L3 Cache)
Motherboard
(Socket 1366)
Supermicro X8SAX; Revision: 1.1; Chipset: Intel X58 + ICH10R; BIOS: 1.0B
RAM3GB DDR3-1333 Corsair CM3X1024-1333C9DHX
System HDDSeagate NL35 400GB
ST3400832NS; 7,200 RPM, SATA 1.5Gb/s, 8MB Cache
Test HDDs3.5” 15,000 RPM (Fujitsu MBA3147RC)
2.5” 10,000 RPM (Toshiba MBF2600RC)
Power SupplyOCZ EliteXstream 800W
OCZ800EXS-EU
Benchmarks
Performance Measurementsh2benchw 3.12
PCMark Vantage 1.0
I/O PerformanceIometer 2006.07.27
Fileserver-Benchmark
Webserver-Benchmark
Database-Benchmark
Workstation-Benchmark
Streaming Reads
Streaming Writes
System Software & Drivers
DriverDetails
Operating SystemWindows Vista Ultimate SP1
Intel ChipsetChipset Installation Utility 9.1.0.1007
AMD GraphicsRadeon 8.12
Intel Matrix Storage8.7.0.1007


3.5'' Vs. 2.5'' SAS HDDs: In Storage, Size Matters : 2.5” Test HDDs: Hitachi Ultrastar C15K147


Finally, Hitachi's Ultrastar C15K147 is representative of 2.5” high performance enterprise hard drives, offering up to 147GB capacity. This isn’t much, but it's still more than most SSDs. In addition, Hitachi's next generation, with up to 300GB capacity, is only a few months away.

This drive offers the same characteristics as the 3.5” drive: 64MB cache and a SAS 6Gb/s interface. Hitachi uses two platters for this product, and we found that power consumption is amazingly low for a 15,000 RPM drive. A 4.7W idle and and 7.3W peak load is roughly half the power consumption of most 3.5” drives. The C15K147 doesn't run much cooler, though, still registering 59°C versus 61°C and 64°C for the 3.5” drives.

Performance-wise, the C15K147 can't reach the Ultrastar 15K600's 195 MB/s throughput, but it reads and writes data almost exactly at the same speed as the Ultrastar 15K450. Application performance outstrips the 15K450 a bit, while I/O performance is slightly better on the 3.5” drives. However, 2.5” drives have significant advantages over 3.5” models when it comes to storage density in rackmount environments. The space and power envelope required to run four 3.5” drives will allow up to ten 2.5” drives in the same 1U.