We’ve long heard stories about how 2.5 inch drives would save the environment by providing “green” storage; but with ever-increasing demands for performance power usage might just have been put on the back burner. So I figured a comparison of 2.5 and 3.5 inch drive’s power usage might be worthwile.
For smaller storage needs; the difference is obvious. If all you’re looking for are one or two 146 or 300 GB drives, there’s no question about it: 2.5 inch is the way to go. When you’re looking for the highest storage densities things become very different; in the 7200rpm “we don’t care about performance, give us massive storage” segment (also known as “second tier storage”, “near-line storage” etc), 3.5 inch SATA is the way to go. 2.5 inch drives currently max out at 500 GB, which just isn’t enough.
Which just leaves the high performance storage; 10.000 and 15.000 rpm drives for more demanding applications. For most server vendors, you can get space for two 2.5 inch drives for every 3.5 inch; I selected two comparable drives, the Seagate Cheetah NS.2 and Savvio 10K.3. Both are 10.000 rpm drives; here are there specs according to Seagate’s data sheets:
| Model | Savvio 10K.3 | Cheetah NS.2 |
|---|---|---|
| Form factor | 2.5 inch | 3.5 inch |
| Capacity | 300 GB | 600 GB |
| MTBF | 1.600.000 hours | 1.600.000 hours |
| BER | 1*10^16 | 1*10^16 |
| Latency | 3.0 ms | 3.0 ms |
| Random seek | 3.6 ms | 3.8 ms |
| Random write | 4.2 ms | 4.4 ms |
| Idle power | 3.5W | 6.22W |
| Operating power | 6.15W | 9.75W |
| More information | data sheet | data sheet |
As you can see, using a 3.5 inch drive instead of two 2.5 inch drives saves you both power and money, while leading only to a small loss in performance. In addition, you halve the risk of a drive failure; both drives have exactly the same MTBF according to Seagate’s specs, which means that when using two 2.5 inch drives the chances of a drive failing are much higher.
This doesn’t mean 2.5 inch drives are completely useless; but buying them just for their green image is not always the smartest thing to do.
Related posts:
Your RSS reader
Daily e-mail
Twitter
Whoa! “In addition, you halve the risk of a drive failure; both drives have exactly the same MTBF according to Seagate’s specs, which means that when using two 2.5 inch drives the chances of a drive failing are much higher.”
These are interesting figures, but I would check with a statistician before arriving at the above conclusion. Bear in mind that the probabilities of failure for the two drives are Independent of each other. I don’t know how you halve the risk of failure. IANAStatistician, so please ask one. On the face of it, the fact that you have two independent drives suggests to me that the two-drive solution is far more reliable than the one-drive solution.
I’m not a statistician either, but if you’d want to build a RAID-array of, say, 3 TB, you’d need 6 600 GB drives for a RAID-5 array, or 7 for a RAID-6 array (capable of withstanding either a single or two drive failures).
If you’d want to reach the same capacity with 2.5 inch drives, you’re stuck with 300 GB drives and will need a lot more (16 and 17 respectively). (Numbers are from this RAID calculator)
If you have 16 2.5 inch disks running, the risk that two of them fail at approximately the same time, destroying all your data even though running RAID-5, is much larger than two drives our of 6 failing. In the 2.5 inch case, that would mean 12.5% of the drives failing; for the 3.5 inch case 33% of the drives need to fail to get data loss!
Regarding "As you can see, using a 3.5 inch drive instead of two 2.5 inch drives saves you both power and money, while leading only to a small loss in performance."
You are way off. Two 10K 2.5" drives give twice as much performance (IOPS@response_time) as one 10K 3.5" drive. Why do you think that TPC-C benchmark systems always use the largest number disks, and use the smallest capacity disks available?
A better comparison would be to use two 7,200RPM 2.5" laptop-class disks in place of a single 10K RPM 3.5" disk. Because power consumption increases as the cube of RPM, the two 7,200RPM disks use only 1.2 watts each and together they give 20% more IOPS for 1/3rd the power.
You can read more here, this is the seminal research paper on the topic:
ftp://ftp.cs.rochester.edu/pub/papers/systems/04.tr837.Power-effic_server-class_perf_from_arrays_of_laptop_disks.pdf