Author
|
Thread
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
LightningCrash
Smile like Bob, order your free LC today
Joined: 03 Apr 2003
Posts: 5020
|
|
|
LightningCrash wrote:
I tried it out in a VM, using loopback disks in /dev/shm and some real disk
You set the real disk up as --write-mostly, set up a write-intent bitmap on the SSDs and enable write-behind (which will enable write-behind only for the --write-mostly disks)
In this configuration the writes are as fast as the SSD until you hit the limit of the write-intent bitmap (however big you configured that to be in write-behind x stripe size)
Reads are almost always SSD speed because of the --write-mostly on the HDD.
Really a pretty smoking setup, it would go great with the appropriate SSDs and HDDs.
Read some blog posts on this very thing, I'll see if I can find them.
Oddly enough, while this is a great deal of wrangling in Linux and nigh impossible under Windows, there is a similar facility that is stupidly simple to use under Solaris with ZFS.
In Solaris you can add disks as disk cache. This way you can load up a server or workstation with SSDs, and have the bulk storage in the zpool be SATA or SAS HDDs. Hell you could just have a Fibre Channel card feeding to your SAN or something, it really doesn't matter.
The SSDs will automatically cache data for reads. Now this doesn't help our your write speed at all, but it does make a big difference on high op/s read loads.
For write loads you can add an SSD as something called the ZIL, which is a write logging system in between the buffers and the disk.
http://blogs.sun.com/brendan/entry/test
It will probably be 5-10 years before Windows ever has anything like this, unfortunately.
Hell it will probably be 5 or more before Linux does anything.
|
Thu Jul 15, 2010 8:57 pm
|
|
|
|
|
|
Forum Rules:
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|
|
|
|
|