Re: SSD --- The reliability of
On Tue, 11 Aug 2009 23:30:35 -0700 (PDT), Penang <kalambong@gmail.com> wrote: >Please pardon me if this topic has been discussed before, but as far >as I know about the Flash memory, they have a relatively low limit on >re-write, as compared to hard disk > >In situation where database use in concern, read and write operation >takes place very often > >I know that there are pros and cons in this issue, but basically my >question is this --- How reliable is SSD? We cannot reasonably generalize this, it depends on your specific database use, how often and the amount it would write/rewrite data versus the total capacity of the SSD. Consider that if the database is 4GB and you have an 8GB HDD, so it has 4GB free space, the wear-leveling feature of an SSD would write 7X more often to each free area than if you had 4GB database and a 32GB SSD (28GB free space). That 7X could be very significant, suppose the SSD would last 2 years if 8GB total or 14 years as 32GB total. I haven't yet mentioned MLC vs SLC flash chips, chosing the latter typically extends write cycles by roughly 10X. Now in the above context consider a lifespan of 2 years or 10 x 14 = 140 years. When all is said and done, if you pick the right SSD for the job including it having excess capacity, it is far more reliable than a mechanical hard drive is. MUCH MUCH more reliable, nobody in their right mind would use a mechanical HDD for more than a single-digit # of years for any important database use, that single-digit number of years closer to 0 than to 9.
> From: spam@spam.com > Subject: Re: SSD --- The reliability of > Date: Wed, 12 Aug 2009 09:29:09 -0400 > To: informix-list@iiug.org > > On Tue, 11 Aug 2009 23:30:35 -0700 (PDT), Penang > <kalambong@gmail.com> wrote: > > >Please pardon me if this topic has been discussed before, but as far > >as I know about the Flash memory, they have a relatively low limit on > >re-write, as compared to hard disk > > Sigh. This isn't a stupid question and its one that *all* of us should consider. As it has already been pointed out, that not all SSD's are created equal and not all SSDs are destined for laptops. You could probably create a hot swap array of SSD's to limit the effects of losing one drive and be able to maintain uptime. You could choose a manufacturer which will format a drive in such a way that only 50% of the drive is usable and that they swap a failed block with an unused block in their reserve. With respect to hard drives, the hardest thing about hard drives is in a laptop, where they face numerous start up from sleep / power off, and shock from movement. (You do have moving parts.) So hard drive failure on laptops is real. On desktops, always powering on/off has an effect. On servers? power surges on older drives, wear and tear and mechanical failures will happen over time. (Including media failure.) The more drives you have, the lower the average mtbf drops. So 'super fast systems' that rely on a lot of spindles, keep lots of spares. I've asked similar questions on this and there is one poster who has been experimenting and found that the SSDs are reliable enough, and then with a twist, there are PCIe based controllers and disks (on the same card) which are supposed to further improve performance. With respect to reliability, I believe in the right scenario (hot swap raid), they are a viable choice. However its the cost component and the question of necessity. With respect to low latency systems, even an SSD is too slow and you want your data in memory with fast access. If you've got a high volume OLTP system where you have to support a high transaction count per second... You will want the drives in a hot swap raid and a lot of drives in hot active spares and cold spares available. (This is true of any drives. So ask your self what's the cost per advantage over SAS? _________________________________________________________________ Get your vacation photos on your phone! http://windowsliveformobile.com/en-us/photos/default.aspx?&OCID=0809TL-HM