6

I've found MySQL fairly slow on EC2 and I am sure it is disk speed. My EC2 install is a test environment so I don't need to keep the data if a block should fail which I think allows me to use XFS.

So....What is the fastest filesystem to use under Linux on EC2 for MySQL?

EDIT: I am using EBS, also I am looking to see if anyone has experience outside of EXT3 as that is what I already find slow

1

3 Answers 3

5

ext2 will be the fastest (and cheapest) filesystem. No need for the journaling of ext3, since the database already takes care of that on its own.

Also, what the user "someone" said is correct. The drive will be faster if you initialize it first. However, if you plan to take snapshots, keep in mind that initializing, even with 0s, will make the snapshots bigger (and therefore more expensive), because now all the blocks have changed.

1
  • Just have a plan to deal with the extra FSCK overhead of ext2 over ext3.
    – LapTop006
    Oct 25, 2009 at 13:39
0

EXT3 will work just fine, but you may want to use EBS volumes instead of the space that comes with your EC2 instance, as that may be faster.

Edit: You can create a software RAID on EBS to boost performance (I'd recommend RAID-10) by using mdadm.

2
  • I will edit the question, I am using EBS Oct 23, 2009 at 8:13
  • ext2 is even better. The database already takes care of the consistency; no need to add the overhead of ext3 journaling.
    – jedberg
    Oct 24, 2009 at 6:22
0

I read somewhere in a powerpoint slide evaluating Amazon EC2 for enterprise, that database EBS was slow and writing all 1's first before using sped it up. Hope that helps.

EDIT: It may not have been EBS. Here's the slide: http://www.slideshare.net/cmercier/exploring-amazon-ec2-for-scaleout-applications-366335

1
  • It's not about writing 1s, it's about writing anything at all (according to the slides I linked). Presumably Amazon lazily allocates the space.
    – Jo Liss
    Jul 25, 2011 at 16:53

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .