Re: hibernate: swap on SSD = not fast
Gaël DONVAL wrote: Le vendredi 24 août 2012 à 16:47 +0200, Sven Hartge a écrit : Maybe the alignment of the swap partition is off? SSD react very badly if you don't align your partitions to their erase block size. This is why I always recommend to align them to 1MB boundaries as most SSDs available today seem to have erase block sizes in the 64KB to 512KB range. So alignment and trimming should be check. Maybe also a look at the logs to spot possible IO/FS errors … Let me look into that. Thanks! Hugo -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/k18s0v$mnu$1...@ger.gmane.org
Re: hibernate: swap on SSD = not fast
Le vendredi 24 août 2012 à 16:47 +0200, Sven Hartge a écrit : > Maybe the alignment of the swap partition is off? SSD react very badly > if you don't align your partitions to their erase block size. This is > why I always recommend to align them to 1MB boundaries as most SSDs > available today seem to have erase block sizes in the 64KB to 512KB > range. So alignment and trimming should be check. Maybe also a look at the logs to spot possible IO/FS errors … -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1345822946.21543.1.ca...@p76-nom-gd.cnrs-imn.fr
Re: hibernate: swap on SSD = not fast
Gary Dale wrote: > So why would switching to an SSD slow hibernation times? Frankly I can't > think of any reasons (comparing apples to apples) why a faster drive > should lead to slower performance. Possibly it's an interface issue - > the SSD's controller is getting swamped - while the HDD the poster had > been using was able to handle a faster continuous write. Maybe the alignment of the swap partition is off? SSD react very badly if you don't align your partitions to their erase block size. This is why I always recommend to align them to 1MB boundaries as most SSDs available today seem to have erase block sizes in the 64KB to 512KB range. Grüße, Sven. -- Sigmentation fault. Core dumped. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/d92ap133n...@mids.svenhartge.de
Re: hibernate: swap on SSD = not fast
On 24/08/12 08:29 AM, Gaël DONVAL wrote: Le vendredi 24 août 2012 à 12:58 +0100, Darac Marjal a écrit : On Fri, Aug 24, 2012 at 01:32:16PM +0200, Gaël DONVAL wrote: Le vendredi 24 août 2012 à 12:04 +0100, Darac Marjal a écrit : I might be wrong here, but isn't the key benefit of SSDs that they have a tiny access time? But that their read speed is about the same as a normal disk (also, I might be wrong, but I understand their write speed is average). This would have been true some years ago: Comparison of average sequential reading rates (HDD of 2012 and SSD of 2011): http://www.tomshardware.com/charts/ssd-charts-2011/AS-SSD-Sequential-Read,2782.html http://www.tomshardware.com/charts/hdd-charts-2012/-01-Read-Throughput-Average-h2benchw-3.16,2901.html Please note that I compared SSD with desktop HDD: mobile HDDs are generally slower. About write speed, the very best HDD gave 164.06MB/s on average while *most* SSDs are above 150MB/s and the best reaches a few MB/s less than 400. Ah, my knowledge was out of date. I'll try and squirrel away that factoid :) :) But you comment still holds true: one should not expect a huge increase of performance here as far as hibernation is concerned. For a fast hibernate, sequential write speed is the key benchmark. The Samsung 830 is supposedly very fast at this. However, the original poster complained that hibernation is actually slower. This is the issue that should be discussed. So why would switching to an SSD slow hibernation times? Frankly I can't think of any reasons (comparing apples to apples) why a faster drive should lead to slower performance. Possibly it's an interface issue - the SSD's controller is getting swamped - while the HDD the poster had been using was able to handle a faster continuous write. However, I can't find any evidence for this in the TomsHardware benchmarks. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/503780e7.80...@rogers.com
Re: hibernate: swap on SSD = not fast
Le vendredi 24 août 2012 à 12:58 +0100, Darac Marjal a écrit : > On Fri, Aug 24, 2012 at 01:32:16PM +0200, Gaël DONVAL wrote: > > Le vendredi 24 août 2012 à 12:04 +0100, Darac Marjal a écrit : > > > I might be wrong here, but isn't the key benefit of SSDs that they have > > > a tiny access time? But that their read speed is about the same as a > > > normal disk (also, I might be wrong, but I understand their write speed > > > is average). > > This would have been true some years ago: > > > > Comparison of average sequential reading rates (HDD of 2012 and SSD of > > 2011): > > http://www.tomshardware.com/charts/ssd-charts-2011/AS-SSD-Sequential-Read,2782.html > > http://www.tomshardware.com/charts/hdd-charts-2012/-01-Read-Throughput-Average-h2benchw-3.16,2901.html > > > > Please note that I compared SSD with desktop HDD: mobile HDDs are > > generally slower. > > > > About write speed, the very best HDD gave 164.06MB/s on average while > > *most* SSDs are above 150MB/s and the best reaches a few MB/s less than > > 400. > > Ah, my knowledge was out of date. I'll try and squirrel away that > factoid :) > :) But you comment still holds true: one should not expect a huge increase of performance here as far as hibernation is concerned. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1345811358.16160.2.ca...@p76-nom-gd.cnrs-imn.fr
Re: hibernate: swap on SSD = not fast
On Fri, Aug 24, 2012 at 01:32:16PM +0200, Gaël DONVAL wrote: > Le vendredi 24 août 2012 à 12:04 +0100, Darac Marjal a écrit : > > I might be wrong here, but isn't the key benefit of SSDs that they have > > a tiny access time? But that their read speed is about the same as a > > normal disk (also, I might be wrong, but I understand their write speed > > is average). > This would have been true some years ago: > > Comparison of average sequential reading rates (HDD of 2012 and SSD of > 2011): > http://www.tomshardware.com/charts/ssd-charts-2011/AS-SSD-Sequential-Read,2782.html > http://www.tomshardware.com/charts/hdd-charts-2012/-01-Read-Throughput-Average-h2benchw-3.16,2901.html > > Please note that I compared SSD with desktop HDD: mobile HDDs are > generally slower. > > About write speed, the very best HDD gave 164.06MB/s on average while > *most* SSDs are above 150MB/s and the best reaches a few MB/s less than > 400. Ah, my knowledge was out of date. I'll try and squirrel away that factoid :) signature.asc Description: Digital signature
Re: hibernate: swap on SSD = not fast
Le vendredi 24 août 2012 à 12:04 +0100, Darac Marjal a écrit : > I might be wrong here, but isn't the key benefit of SSDs that they have > a tiny access time? But that their read speed is about the same as a > normal disk (also, I might be wrong, but I understand their write speed > is average). This would have been true some years ago: Comparison of average sequential reading rates (HDD of 2012 and SSD of 2011): http://www.tomshardware.com/charts/ssd-charts-2011/AS-SSD-Sequential-Read,2782.html http://www.tomshardware.com/charts/hdd-charts-2012/-01-Read-Throughput-Average-h2benchw-3.16,2901.html Please note that I compared SSD with desktop HDD: mobile HDDs are generally slower. About write speed, the very best HDD gave 164.06MB/s on average while *most* SSDs are above 150MB/s and the best reaches a few MB/s less than 400. > > Hibernation, in contrast, is about writing out (and reading back) a > linear stream of data. But you are right here: sequential read/write should be fast on HDDs as well. But with and SSD twice as fast as the previous HDD, you would still expect suspend time to be cut off by a factor of 2. > > So, in summary, while SSDs may well help with swap performance, I'd not > expect them to be brilliant at hibernation. Seems right. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1345807936.15003.13.ca...@p76-nom-gd.cnrs-imn.fr
Re: hibernate: swap on SSD = not fast
On Thu, Aug 23, 2012 at 10:00:46AM -0500, hvw59601 wrote: > Hi, > > Mindful of what Stan Hoeppner in various posts has written about SSD > I thought I'd put swap on an SSD I installed (Samsung SSD 830 128GB) > in order to get superfast hibernate. I might be wrong here, but isn't the key benefit of SSDs that they have a tiny access time? But that their read speed is about the same as a normal disk (also, I might be wrong, but I understand their write speed is average). Hibernation, in contrast, is about writing out (and reading back) a linear stream of data. So, in summary, while SSDs may well help with swap performance, I'd not expect them to be brilliant at hibernation. signature.asc Description: Digital signature
Re: hibernate: swap on SSD = not fast
On 8/24/2012 3:43 AM, Jon Dowland wrote: > On Fri, Aug 24, 2012 at 01:42:45AM -0500, Stan Hoeppner wrote: >> This email mentions nothing of any troubleshooting performed by you up >> to this point. It's sole purpose seems to be to blame me for what you >> apparently believe is a bad recommendation. > > I read it very differently: veiled admiration and thanks for your expert > advice, and genuine surprise that things weren't working properly, with > a request for help in figuring out why. I'm in a very good mood right now > though, YMMV ☺ Bah, maybe I was being unnecessarily defensive. If so my apologies. When I see my name specified on a worldwide mailing list, by someone whose alias I don't recognize, followed by a complaint that something I recommended isn't working, I guess I tend to assume the worst... -- Stan -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/5037552f.2090...@hardwarefreak.com
Re: hibernate: swap on SSD = not fast
Le jeudi 23 août 2012 à 10:00 -0500, hvw59601 a écrit : > Hi, > > Mindful of what Stan Hoeppner in various posts has written about SSD I > thought I'd put swap on an SSD I installed (Samsung SSD 830 128GB) in > order to get superfast hibernate. > > Surprise: it is slower than usual and the disk light is on. >From my very limited knowledge of how SSDs actually work, I wonder if swapon actually trims your swap partition. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1345801314.4875.31.ca...@p76-nom-gd.cnrs-imn.fr
Re: hibernate: swap on SSD = not fast
On Fri, Aug 24, 2012 at 01:42:45AM -0500, Stan Hoeppner wrote: > This email mentions nothing of any troubleshooting performed by you up > to this point. It's sole purpose seems to be to blame me for what you > apparently believe is a bad recommendation. I read it very differently: veiled admiration and thanks for your expert advice, and genuine surprise that things weren't working properly, with a request for help in figuring out why. I'm in a very good mood right now though, YMMV ☺ -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20120824084332.GB19780@debian
Re: hibernate: swap on SSD = not fast
On 8/23/2012 10:00 AM, hvw59601 wrote: > Mindful of what Stan Hoeppner in various posts has written about SSD I > thought I'd put swap on an SSD I installed (Samsung SSD 830 128GB) in > order to get superfast hibernate. > > Surprise: it is slower than usual and the disk light is on. [snip] > Is there an explanation of this? It takes about 20s. to hibernate when > swap is on SSD. SSDs aren't magical white unicorns. They're just another SATA storage device, albeit a fast one, and when problems arise you troubleshoot, as with anything else. This email mentions nothing of any troubleshooting performed by you up to this point. It's sole purpose seems to be to blame me for what you apparently believe is a bad recommendation. Until you do basic troubleshooting we simply won't know what the problem is. It could be any number of things that are fixable, hardware or software related, or you could have a bad SSD. But I assure you the problem isn't that hibernate to SSD is universally slow. -- Stan -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/50372265.2060...@hardwarefreak.com
hibernate: swap on SSD = not fast
Hi, Mindful of what Stan Hoeppner in various posts has written about SSD I thought I'd put swap on an SSD I installed (Samsung SSD 830 128GB) in order to get superfast hibernate. Surprise: it is slower than usual and the disk light is on. swap is here: + cat /proc/swaps FilenameTypeSizeUsed Priority /dev/sdb3 partition 19535036 0 -1 which is on SSD: ... /dev/sdb3: LABEL="SSD830.03" UUID="ede71620-d8fa-47fd-8aee-0e8b37f9e982" TYPE="swap" ... and this is the kernel cmdline: BOOT_IMAGE=/SDB2.boot/vmlinuz-3.4.7-nodeb-amd64 root=LABEL=SSD830.02 ro vga=791 nouveau.modeset=0 resume=LABEL=SSD830.03 Is there an explanation of this? It takes about 20s. to hibernate when swap is on SSD. Hugo -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/k15giu$8tp$1...@ger.gmane.org