Obviously doesn't work for extremely large datasets, but nothing stops you
from stuffing a server full of memory, assigning a huge block to ramfs, and
using that as the second leg of a mirror, with the first leg a real disk
device set to write-mostly.
Obviously you'll need to create an init script to hot-add the ramfs block
device at boot time, as all mirror volume headers are lost at power down :-)
Also, note that I specifically said ramfs, NOT tmpfs - the former only takes
sizing parameters at boot time, but creates block devices in /dev, while the
latter may be resized on-the-fly, but creates a *filesystem*, not a block
device, so you a) have to mess with loopback devices and b) are stuck with
the filesize limits of tmpfs, whatever they are. Additionally, ramfs
allocates the full amount of memory as soon as you use the block device, so
you won't unexpectedly run into memory issues months after starting your
On Sun, Feb 8, 2009 at 8:09 AM, mos <mos99@stripped> wrote:
> At 04:53 PM 2/6/2009, you wrote:
>> While SSD's (Solid State Disks) have traditionally not been the best
>> hardware to use for rewrite-intensive operations like databases, over
>> the last few months, some leading Linux kernel engineers have been
>> raving about next generation Intel SSD's that are close to 20x faster
>> than the fastest disk drives for random access. If robust enough, these
>> next generation SSD's may greatly improve relational database
>> Is anyone using SSD drives currently and can share their experiences?
>> Also, even if they are currently poor for writing, they would make a
>> fantastic slave drive for mega-fast access I would think right?
>> Our current DB is about 80GB and over 1/2 billion rows in each of two of
>> the tables. Reports are starting to take as long as 20 seconds to
>> generate. ...and don't get me started on export/import (it can take DAYS
>> to import).
> There are Flash SSD's and DDR SSD's. I assume you mean DDR SSD's.
> You may want to try something like HyperDrive and use Raid to increase the
> size of the volume. See http://www.hyperossystems.co.uk/. It was reviewed
> at http://www.techreport.com/articles.x/16255/9. For random access, these
> solid state devices are extremely fast. But sequential access they are
> slightly faster than the fastest hard drive. These devices are also getting
> quite cheap compared to what they cost 5 years ago. I haven't tried it, but
> would love to get my hands on a few of these.
> MySQL General Mailing List
> For list archives: http://lists.mysql.com/mysql
> To unsubscribe: http://lists.mysql.com/mysql?unsub=1
Celsius is based on water temperature.
Fahrenheit is based on alcohol temperature.
Ergo, Fahrenheit is better than Celsius. QED.