Re: SSD with firmware upgrade under OpenBSD
On Sun, Nov 28, 2010 at 8:01 PM, Claus Assmann wrote: > I was about to buy an OCZ Vertex 2 SSD when I read that firmware > updates for that kind of SSD require some M$ Windows version. Is > someone using SSDs with a high IOPS rate (the Sandforce controller > claims 45-50 kIOPS) which can be updated under some freely available > software? I would like to try an SSD as mail queue FS etc. To follow up, I bought two OCZ Agility 2 drives recently from newegg, and they came with the latest firmware, so you may not need to update the firmware at all if you're into gambling.
Re: SSD with firmware upgrade under OpenBSD
On 11/30/10 13:56, Robert Bronsdon wrote: On Tue, 30 Nov 2010 20:21:06 -, Brad Tilley wrote: Also, I just noticed that the high-end Intel SSDs claim 2,000,000 hours MTBF. I wonder why they market that number and then say "3 year warranty". There's only roughly 26,280 hours in a three year period. I wonder how many will fail inside the 2,000,000 hours while running your script then how many will fail sat on a shelf for that time period. The disk you've (for any given you) purchased will fall somewhere between these. While not related to the MTBF discussion, quite a lot of testing has been done with SSDs by some of the PostgreSQL folks and as a result there is a Big Warning on the PostgreSQL wiki http://wiki.postgresql.org/wiki/Reliable_Writes That links to two separate articles: "See SSD, XFS, LVM, fsync, write cache, barrier and lost transactions [1] (Vadim Tkachenko) for an introduction. Their consumer drives such as the X-25-M G2 have been reported to be even worse [2] (Evan Jones)." [1] http://www.mysqlperformanceblog.com/2009/03/02/ssd-xfs-lvm-fsync-write-cache-barrier-and-lost-transactions/ [2] http://evanjones.ca/intel-ssd-durability.html You can find much, much more in the postgresql mailing list archives. In short, while there are SSDs appearing on the market that do correctly report safe writes, the Intels are not among them. Jeff Ross
Re: SSD with firmware upgrade under OpenBSD
On Tue, 30 Nov 2010 20:21:06 -, Brad Tilley wrote: Also, I just noticed that the high-end Intel SSDs claim 2,000,000 hours MTBF. I wonder why they market that number and then say "3 year warranty". There's only roughly 26,280 hours in a three year period. I wonder how many will fail inside the 2,000,000 hours while running your script then how many will fail sat on a shelf for that time period. The disk you've (for any given you) purchased will fall somewhere between these. -- Robert Bronsdon
Re: SSD with firmware upgrade under OpenBSD
On Tue, 30 Nov 2010 20:09:14 +0100 Claudio Jeker wrote: > sandforce controller Noted, nice one Claudio.
Re: SSD with firmware upgrade under OpenBSD
On Tue, Nov 30, 2010 at 3:21 PM, Brad Tilley wrote: > Kevin Chadwick wrote: > >> I almost completely agree, but also disagree and yes I'd say it's not >> worth getting into again. I would have to check the latest developments >> as I can imagine an algorithm which solved the problem during idle >> periods or didn't use it's full capacity but currently I don't agree >> fully with "huge amounts of data". The problem was reduced immensely by >> spreading writes across all free sectors rather than sequentially but I >> believe? the problem re-appears on a busy nearly full disk. I would also >> hope/imagine the only affect would be getting bad sectors in that area >> but I haven't looked into it very far as I currently have no need to >> and so maybe I should shut up untill I do. However, I for one will not >> be treating SSDs like HDDs in all applications of disks untill after I >> learn more. > > One thing you might consider... buy a SSD and do some testing. Attach it > to an OpenBSD box, put a file system on it, then write a script similar > to this to repeatedly fill and empty the file system: > > while : > do >dd if=/dev/arandom of=big_un.bin bs=64k >sync >sleep 1 >rm -P big_un.bin > done > > Let that run for a few years and see how long the disk actually lasts. > You could put up a website with live results. You'd become famous too... > especially if you hit the decade mark and the thing still works :) > > Also, I just noticed that the high-end Intel SSDs claim 2,000,000 hours > MTBF. I wonder why they market that number and then say "3 year > warranty". There's only roughly 26,280 hours in a three year period. > > Brad MTBF numbers are typically cumulative, i.e. for the population of devices as a whole. If you had 2,000,000 Intel SSDs, you could expect (statistically) for one of them to fail every hour. Paul. -- -- Paul D. Ouderkirk Senior UNIX System Administrator p...@ouderkirk.ca -- laughing, in the mechanism -- William Gibson
Re: SSD with firmware upgrade under OpenBSD
Kevin Chadwick wrote: > I almost completely agree, but also disagree and yes I'd say it's not > worth getting into again. I would have to check the latest developments > as I can imagine an algorithm which solved the problem during idle > periods or didn't use it's full capacity but currently I don't agree > fully with "huge amounts of data". The problem was reduced immensely by > spreading writes across all free sectors rather than sequentially but I > believe? the problem re-appears on a busy nearly full disk. I would also > hope/imagine the only affect would be getting bad sectors in that area > but I haven't looked into it very far as I currently have no need to > and so maybe I should shut up untill I do. However, I for one will not > be treating SSDs like HDDs in all applications of disks untill after I > learn more. One thing you might consider... buy a SSD and do some testing. Attach it to an OpenBSD box, put a file system on it, then write a script similar to this to repeatedly fill and empty the file system: while : do dd if=/dev/arandom of=big_un.bin bs=64k sync sleep 1 rm -P big_un.bin done Let that run for a few years and see how long the disk actually lasts. You could put up a website with live results. You'd become famous too... especially if you hit the decade mark and the thing still works :) Also, I just noticed that the high-end Intel SSDs claim 2,000,000 hours MTBF. I wonder why they market that number and then say "3 year warranty". There's only roughly 26,280 hours in a three year period. Brad
Re: SSD with firmware upgrade under OpenBSD
On Tue, Nov 30, 2010 at 05:42:51PM +, Kevin Chadwick wrote: > On Tue, 30 Nov 2010 16:44:51 +0100 > Jan Stary wrote: > > > On Nov 30 12:32:16, Kevin Chadwick wrote: > > > On Mon, 29 Nov 2010 21:17:17 -0500 > > > Brad Tilley wrote: > > > > > > > Do they really fail that often? > > > > > > My current understanding is that a mostly empty SSDS electronics will > > > fail before it forgets what it's written but a mostly full and busy SSD > > > may start forgeting fairly soon, unless it shuffles data which would > > > slow it down considerably. > > > > My current understanding is that you treat a SSD as any other disk and > > never even notice that your wd0/sd0 is not a piece of metal rotating > > at 7200RPM, unless you read/write huge amounts of data, which you don't. > > > > Let's not get into that again. > > > > I almost completely agree, but also disagree and yes I'd say it's not > worth getting into again. I would have to check the latest developments > as I can imagine an algorithm which solved the problem during idle > periods or didn't use it's full capacity but currently I don't agree > fully with "huge amounts of data". The problem was reduced immensely by > spreading writes across all free sectors rather than sequentially but I > believe? the problem re-appears on a busy nearly full disk. I would also > hope/imagine the only affect would be getting bad sectors in that area > but I haven't looked into it very far as I currently have no need to > and so maybe I should shut up untill I do. However, I for one will not > be treating SSDs like HDDs in all applications of disks untill after I > learn more. > Just for the record, the sandforce controller used by the mentioned SSD drives keeps 6% or 25% (depending on version) for internal consumption and write wear leveling logic. So a 100GB OCZ drive is internally a 128GB one and the drive will do a lot of magic to limit the number of writes to the SSD. The sandforce controller is very different form any other device I know. Most controllers like the samsung and intel one use a small SDRAM as cache and when the cache (normaly 64 or 128MB) is full the disk starts to behave badly. Now the sandforce does not have a cache but uses the unallocated memory for such tasks by doing that and a few other magical tricks they claim a write amplification below 1. Modern SSDs are fairly reliable like HDDs and in the end both will die and they will die in the most unpleasent way and at the worst time you can imagine. So be prepared :) -- :wq Claudio
Re: SSD with firmware upgrade under OpenBSD
Kevin Chadwick wrote: > On Tue, 30 Nov 2010 16:44:51 +0100 > Jan Stary wrote: > >> On Nov 30 12:32:16, Kevin Chadwick wrote: >>> On Mon, 29 Nov 2010 21:17:17 -0500 >>> Brad Tilley wrote: >>> Do they really fail that often? >>> My current understanding is that a mostly empty SSDS electronics will >>> fail before it forgets what it's written but a mostly full and busy SSD >>> may start forgeting fairly soon, unless it shuffles data which would >>> slow it down considerably. >> My current understanding is that you treat a SSD as any other disk and >> never even notice that your wd0/sd0 is not a piece of metal rotating >> at 7200RPM, unless you read/write huge amounts of data, which you don't. >> >> Let's not get into that again. >> > > I almost completely agree, but also disagree and yes I'd say it's not > worth getting into again. I would have to check the latest developments > as I can imagine an algorithm which solved the problem during idle > periods or didn't use it's full capacity but currently I don't agree > fully with "huge amounts of data". The problem was reduced immensely by > spreading writes across all free sectors rather than sequentially but I > believe? the problem re-appears on a busy nearly full disk. I would also > hope/imagine the only affect would be getting bad sectors in that area > but I haven't looked into it very far as I currently have no need to > and so maybe I should shut up untill I do. However, I for one will not > be treating SSDs like HDDs in all applications of disks untill after I > learn more. I've been treating my SSD like any other hard disk during the last year. It is still working fine. The specs say it has a MTBF of 1,000,000 hours and I've only used it for about 10,000 hours so far. I've been at 60% capacity since day one. If it fails before meeting the MTBF, I'll send it back for a refund. If it lasts as long as they claim it will (about a hundred years), then I'll be dead before it stops working. :) Brad
Re: SSD with firmware upgrade under OpenBSD
On Tue, 30 Nov 2010 16:44:51 +0100 Jan Stary wrote: > On Nov 30 12:32:16, Kevin Chadwick wrote: > > On Mon, 29 Nov 2010 21:17:17 -0500 > > Brad Tilley wrote: > > > > > Do they really fail that often? > > > > My current understanding is that a mostly empty SSDS electronics will > > fail before it forgets what it's written but a mostly full and busy SSD > > may start forgeting fairly soon, unless it shuffles data which would > > slow it down considerably. > > My current understanding is that you treat a SSD as any other disk and > never even notice that your wd0/sd0 is not a piece of metal rotating > at 7200RPM, unless you read/write huge amounts of data, which you don't. > > Let's not get into that again. > I almost completely agree, but also disagree and yes I'd say it's not worth getting into again. I would have to check the latest developments as I can imagine an algorithm which solved the problem during idle periods or didn't use it's full capacity but currently I don't agree fully with "huge amounts of data". The problem was reduced immensely by spreading writes across all free sectors rather than sequentially but I believe? the problem re-appears on a busy nearly full disk. I would also hope/imagine the only affect would be getting bad sectors in that area but I haven't looked into it very far as I currently have no need to and so maybe I should shut up untill I do. However, I for one will not be treating SSDs like HDDs in all applications of disks untill after I learn more.
Re: SSD with firmware upgrade under OpenBSD
On Nov 30 12:32:16, Kevin Chadwick wrote: > On Mon, 29 Nov 2010 21:17:17 -0500 > Brad Tilley wrote: > > > Do they really fail that often? > > My current understanding is that a mostly empty SSDS electronics will > fail before it forgets what it's written but a mostly full and busy SSD > may start forgeting fairly soon, unless it shuffles data which would > slow it down considerably. My current understanding is that you treat a SSD as any other disk and never even notice that your wd0/sd0 is not a piece of metal rotating at 7200RPM, unless you read/write huge amounts of data, which you don't. Let's not get into that again.
Re: SSD with firmware upgrade under OpenBSD
On Mon, 29 Nov 2010 21:17:17 -0500 Brad Tilley wrote: > Do they really fail that often? My current understanding is that a mostly empty SSDS electronics will fail before it forgets what it's written but a mostly full and busy SSD may start forgeting fairly soon, unless it shuffles data which would slow it down considerably.
Re: SSD with firmware upgrade under OpenBSD
On Sun, Nov 28, 2010 at 8:01 PM, Claus Assmann wrote: > I was about to buy an OCZ Vertex 2 SSD when I read that firmware > updates for that kind of SSD require some M$ Windows version. Don't know how the Intel SSD's compare performance wise but you can upgrade their firmware via DOS (FreeDOS is what they use) and it's easy to make a bootable FreeDOS USB stick.
Re: SSD with firmware upgrade under OpenBSD
> On 11/29/2010 02:56 PM, Ted Unangst wrote: > > buying a new SSD to replace your "burned out" one every year is still > > cheaper than building a 15k sas drive raid set with equivalent > > performance. > > I've been using an inexpensive Kingston SSD for more than a year now in > a 4.6 box. It works fine and I've never thought about flashing its > firmware. Its MTBF is astronomical. Do they really fail that often? Reports are that when you re-flash their firmware, they don't fail, but sometimes the data you have on them does..
Re: SSD with firmware upgrade under OpenBSD
On 11/29/2010 02:56 PM, Ted Unangst wrote: > buying a new SSD to replace your "burned out" one every year is still > cheaper than building a 15k sas drive raid set with equivalent > performance. I've been using an inexpensive Kingston SSD for more than a year now in a 4.6 box. It works fine and I've never thought about flashing its firmware. Its MTBF is astronomical. Do they really fail that often? Brad
Re: SSD with firmware upgrade under OpenBSD
On 11/29/10 18:42, Ted Unangst wrote: > On Mon, Nov 29, 2010 at 5:25 PM, Nick Holland > wrote: >> Assuming your firmware update utility works through the USB interface (I >> suspect it would, they have to be doing some kind of command abstraction, >> since they probably don't wish to deal with all the potential different > > ocz recommends against flashing a drive in legacy sata mode instead of > ahci, so i really don't think they'd recommend a usb flash. "never mind..." Nick.
Re: SSD with firmware upgrade under OpenBSD
On Mon, Nov 29, 2010 at 5:25 PM, Nick Holland wrote: > Assuming your firmware update utility works through the USB interface (I > suspect it would, they have to be doing some kind of command abstraction, > since they probably don't wish to deal with all the potential different ocz recommends against flashing a drive in legacy sata mode instead of ahci, so i really don't think they'd recommend a usb flash.
Re: SSD with firmware upgrade under OpenBSD
On 11/28/10 20:01, Claus Assmann wrote: > I was about to buy an OCZ Vertex 2 SSD when I read that firmware > updates for that kind of SSD require some M$ Windows version. Is > someone using SSDs with a high IOPS rate (the Sandforce controller > claims 45-50 kIOPS) which can be updated under some freely available > software? I would like to try an SSD as mail queue FS etc. I'm using a Vertex 2 as my boot disk on my OpenBSD desktop machine. The firmware update utility does need to be run from Windows, but you can use a freely-available Windows recovery disc to do the firmware flashing if you don't dual-boot or have a Windows install disc handy. Look here for the recovery disc: http://neosmart.net/blog/2009/windows-7-system-repair-discs/ BTW, the Vertex 2 is FAST! I'm very, very happy with mine. -- Joe Gidi j...@entropicblur.com
Re: SSD with firmware upgrade under OpenBSD
On 11/28/10 20:01, Claus Assmann wrote: I was about to buy an OCZ Vertex 2 SSD when I read that firmware updates for that kind of SSD require some M$ Windows version. Is someone using SSDs with a high IOPS rate (the Sandforce controller claims 45-50 kIOPS) which can be updated under some freely available software? I would like to try an SSD as mail queue FS etc. Yet another "I've not done this, but here's what I'd do" answer: Your local very well stocked computer store might have one of these USB to IDE/SATA adapters, designed to plug into the back of a bare drive. I have one sitting here on my desk that I paid $15US for, including a power supply for the drive. Assuming your firmware update utility works through the USB interface (I suspect it would, they have to be doing some kind of command abstraction, since they probably don't wish to deal with all the potential different drive interface hardware out there!) and you have a dual-booting laptop handy, take the laptop with Windows and the update utility to your machine, shut down the machine, unplug the drive, plug it into your USB->SATA adapter, do your update, unplug, hook back up to host, enjoy. Probably don't even need to remove the drive from the machine, just unplug the data cable, maybe the power cable. Sure, less convenient than booting off a CD with an update utility on it, but probably far safer than doing the update on a windows machine (I got to believe you really don't want any activity on the device when you are updating its flash, and that's hard to promise under Windows). Nick.
Re: SSD with firmware upgrade under OpenBSD
On 29-Nov-10 11:56, Ted Unangst wrote: buying a new SSD to replace your "burned out" one every year is still cheaper than building a 15k sas drive raid set with equivalent performance. Yes, but I that kind of performance is over-kill for a mail server. Unless you are pushing well over 1 gb/s for mail, you don't need anything more than just a cheap 80 Gig 7.2k SATA II disk, or a couple in RAID. And if you are pushing that much mail, you should just CARP a couple MTAs together. The only time I've seen SSDs worth the price is in rough service equipment and database servers getting hit with over a hundred thousand (read) queries a minute. Most SSDs are slower for writing than their whirly counter-parts. -Christopher Ahrens
Re: SSD with firmware upgrade under OpenBSD
buying a new SSD to replace your "burned out" one every year is still cheaper than building a 15k sas drive raid set with equivalent performance. On Mon, Nov 29, 2010 at 2:48 PM, Robert wrote: > On Sun, 28 Nov 2010 17:01:51 -0800 > Claus Assmann wrote: >> software? I would like to try an SSD as mail queue FS etc. > > I'm not sure if that is a good idea. SSDs are Flash memory and have a > limited read/write cycle. The firmware tries to optimize this by not > always writing to the same address. But since it seems that you want to > use it for a high-volume mail server this will have an impact. Better > check the Web for some wear&tear tests of this drive... > > kind regards, > Robert
Re: SSD with firmware upgrade under OpenBSD
On Sun, 28 Nov 2010 17:01:51 -0800 Claus Assmann wrote: > software? I would like to try an SSD as mail queue FS etc. I'm not sure if that is a good idea. SSDs are Flash memory and have a limited read/write cycle. The firmware tries to optimize this by not always writing to the same address. But since it seems that you want to use it for a high-volume mail server this will have an impact. Better check the Web for some wear&tear tests of this drive... kind regards, Robert
SSD with firmware upgrade under OpenBSD
I was about to buy an OCZ Vertex 2 SSD when I read that firmware updates for that kind of SSD require some M$ Windows version. Is someone using SSDs with a high IOPS rate (the Sandforce controller claims 45-50 kIOPS) which can be updated under some freely available software? I would like to try an SSD as mail queue FS etc.
