February 5, 201016 yr Promo code: EMCYNYS53 http://www.newegg.com/Product/Product.aspx?Item=N82E16822152202&nm_mc=EMC-IGNEFL020510&cm_mmc=EMC-IGNEFL020510-_-HardDrives-_-L0F-_-22152202 This seems to be the most reliable 2TB if the reviews on Newegg are to be trusted. This drive went on sale for $149.99 at Newegg for about a day a couple of months ago. http://lime-technology.com/forum/index.php?topic=4944.0
February 9, 201016 yr Can anyone provide an estimate on write speeds with these drives used both for data and parity drives? Is 20MB/sec write performance likely with the drives connected to the SATA ports on the MB?
February 9, 201016 yr If your talking about writing over the network that will be highly dependent on your network. The drive should be more then capable of doing 20MB/s on its own or when doing internal writes from one drive to the other, but may not reach that if you are writing over the network.
February 9, 201016 yr If your talking about writing over the network that will be highly dependent on your network. The drive should be more then capable of doing 20MB/s on its own or when doing internal writes from one drive to the other, but may not reach that if you are writing over the network. Sorry, I guess I didn't explain myself very well. I'm new to unRAID, but its my understanding that write speeds on parity-protected arrays are one of the significant downsides. I get why they would be slow, based on the fact that it has to do 4 disk operations to do one write. I have some notion from the FAQ on what factors into write speeds, but there aren't a lot of data points available with concrete examples of write performance. The user benchmarks page doesn't even discuss write speeds. So really, what I mean is I don't know how much slower a system using 5400/5900RPM drives would be than 7200RPM, all other things being essentially equal. So, really I'm talking about network writes to a parity-protected array. I wouldn't think that would be terribly network-dependent. It sounds like write performance on unRAID systems is slow enough that even relatively cheap gigabit network devices wouldn't end up being the bottleneck. Certainly that could be the case for read speeds, but I wouldn't think so for writes.
February 9, 201016 yr If you are running unRAID stock (meaning no add-ons), then the rotational speed of the parity drive only matters if your data drives are faster as well. So having a 7200 rpm parity drive and 5400 rpm data drives will 'feel' the same as having all 5400 rpm parity and data drives. If you are concerned about write speed, adding a cache drive of any speed to your array will significantly improve your 'perceived' write speeds. I generally see between 60 and 70 mb/s write to my cache drive (over a gigabit network). However, if you are running add-ons, such as torrents, then a faster parity drive can make a big difference. I run unRAID stock, so I can't give you any concrete numbers to back that statement up.
February 9, 201016 yr I understand that it doesn't make sense to run a parity drive faster than all your data drives. However, I'm still unclear how much of a write performance hit there is from going from a 7200RPM to a 5400RPM drive. I would think it would be terribly significant, but I could easily be wrong. I'm not interested in running any addons. I don't think I need terribly fast write speeds, which is why a cache drive doesn't interest me initially, but I would like to try to make sure I can get something around 20MB/sec for network writes. Does that mean I would need to go with a cache drive, or 7200RPM data/parity drives, or is that doable with 5400RPM drives?
February 9, 201016 yr I understand that it doesn't make sense to run a parity drive faster than all your data drives. However, I'm still unclear how much of a write performance hit there is from going from a 7200RPM to a 5400RPM drive. I would think it would be terribly significant, but I could easily be wrong. I'm not interested in running any addons. I don't think I need terribly fast write speeds, which is why a cache drive doesn't interest me initially, but I would like to try to make sure I can get something around 20MB/sec for network writes. Does that mean I would need to go with a cache drive, or 7200RPM data/parity drives, or is that doable with 5400RPM drives? A recent post by a user indicated they were getting 30MB/s over the lan. This post by BubbaQ calculated the MAX you'll ever be able to obtain with most drives. If you are writing 1GB to unRAID, you have to read 1GB and write 1GB. There has been a lot of discussion of the optimization from reading a bunch at once, then queuing the writes and writing them back in a batch. Consider a drive (such as mine) that reads at 100MB/sec and writes at 75MB/sec. So theoretically it takes 10 seconds of reading and 13.3 seconds of writing for a total of 23.3 seconds to write 1GB to unRAID. That means 43MB/sec of maximum theoretical throughput. Using this formula, and I/O specs for your drives, you can easily calculate the maximum theoretical throughput for your system. This does not take into consideration network overhead, nor other overhead switching from read to write, and rotational latencies on specific disks. A recent post from one user indicated they are getting 30+MB/s on their new unRAID server writing to 7200 RPM drives. http://lime-technology.com/forum/index.php?topic=5286.msg49130#msg49130 Joe L.
February 9, 201016 yr I should have data on 5400 RPM drives within the week. I have been testing "stock" unraid with various hardware configurations. I have a WD EARS drive on the way (64MB Cache, 5400 RPM) to throw it in the mix. Interestingly, I have measured over the wire performance differences of about 7MB/s on large (4GB) file transfers with identical hard drives and only changing the CPU and motherboard (processor and I/O controller combinations). The best performance I have measured so far is 36.9 MB/s average read performance and 26.8MB/s average write performance using ICH10 onboard I/O controllers on 10,000 RPM drives for LARGE file transfers (4GB files). This is performance over the gigabit network. When I get the EARS drive in, I plan to test it on the same configuration to measure the impact of the slower RPM with the ICH10 controller. I also have some 7200 RPM drives to test. However, there are significant performance differences even between drives with identical RPM and cache due to different disk geometry, controllers, etc. You can spend some time comparing here (I'm not a huge fan of Tom's Hardware Guide, but this seems to be the most current and complete HD comparison tool I have found. My old favorite, storagereview.com seems to have taken a vacation...): http://www.tomshardware.com/charts/2009-3.5-desktop-hard-drive-charts/h2benchw-3.12-Avg-Write-Throughput,1013.html There are some good 7200 RPM drives that approach this same level of performance (the WD Black drives, for example), so I know you can get where you want to be at 7200 RPM. Using BubbaQ's theoretical maximum calculation, most of the good 5400 RPM drives will be within 15% of the speeds I reference above. So (in theory) you should exceed 20MB/s on 5400 RPM drives as well.
February 10, 201016 yr No problem. I have been doing a lot of reading regarding performance specifications on the various drives and trying to determine the best drives to purchase. Some of today's green drives perform as well as my 7200 RPM 1TB Caviar Black from 2 years ago. However, looking at the current crop of drives, there is an obvious tradeoff: Performance or Efficiency. 5400-5900 RPM drives have the advantage of consuming about 1/2 the power of the 7200 RPM drives at about 3-5 idle Watts vs 6-9 for the 7200 drives. 7200 RPM drives on the other hand are delivering performance in the 100-110 MB/s range where the 5400 RPM drives only hit 80-90 MB/s. The difference in idle power consumption from the best performing to best power consuming drive is only 38.5 kWh per year = $4.24 USD per drive per year at $.11 per kWh in the midwest. This number then drops down quite a bit further since most people are spinning down their disks when inactive. Unless your drives are active a lot of the time, I don't see low power hard drives being a big factor for anyone financially. The bigger issue would be heat and reliability. But, depending on % active time, this could be a bigger deal for someone with a 20 drive array. How the above performance differences translate to real-world over-the-wire transfer performance is yet to be determined in my testing. Stay tuned.
February 14, 201016 yr I completed some testing with a Western Digital WD10EARS (Green, 5400RPM, 64MB Cache) substituted for the parity drive in a test rig. The results were not pretty. It absolutely crushed the write performance of the system. See detials here: http://lime-technology.com/forum/index.php?topic=5076.135 I didn't include it on the chart, but I have results from 3 year old (lower platter density) 7200 RPM drives on the same host controller which provide over 2x the write throughput of the EARS. The VRaptors I used as a benchmark are intended to illustrate a "best case scenario", but other more practical drives like the 2TB WD Black and 2TB WD RE4 post comparable read/write throughput so it may not be all that impractical as a benchmark performance level.
Archived
This topic is now archived and is closed to further replies.