While I agree with bimble's comment about the test not being very scientific, I'm not sure that it's "utter nonsense". It's a person's first impression from using both Linode and Slicehost.
Not only that, but it's also out-of-date; Linode now has one-click backups, in contrast to what this blog post claims (though it was correct when written back in Dec 2009).
Files that have been modified, but are the same length and without any metadata changes (like mtime) will not be considered "changed" during a subsequent incremental backup. Currently, only ext2/3 volumes can be backed up (this limitation may be removed in an upcoming release). ACLs and extended attributes are NOT tracked.
Not a good comparison, I have to point out. You appear to have had one VPS on each service and run the test once on each. No idea how old either was, how crowded the host you wound up on was, or any of that. So interesting, but not scientific.
For all I can tell, the Slicehost VPS may have been very old, and the Linode VPS may have been brand new on a fresh machine (so no one else was using it and eating into the disk access times).
The forums for both companies have customers talking about hdparm results and such, so it's worth looking there for a broader sample. In general, people testing an existing VPS against a new one (for either service) tend to see better results on the new one. It's just a fact of life if you get a VPS - the old hardware stays in service because customers still run on it, and those customers don't want unnecessary downtime. For best results, make a new slice on each, talk to the techs at each to make sure you're in their latest and greatest datacenter, and then see how things stack up.
It's also worth pointing out another caveat with a VPS - shared disk access. So for scientific results you'd also need a way to control for the number of users on the box, and their usage patterns. If some jerk on your machine consistently thrashes the disk with swap access, it's going to be a pain on any service until their admins see it happening and reboot that VPS.
8 comments
[ 3.2 ms ] story [ 18.9 ms ] threadPlease learn at least the very basics about carrying out an i/o benchmark before spreading FUD...
Eivind Uggedal has already beaten scientific comparison to death: http://news.ycombinator.com/item?id=966555
hdparm is not a benchmark. Not even close. He could just as well have posted "the linode green looks faster than the slicehost blue".
from: http://www.linode.com/backups/
Files that have been modified, but are the same length and without any metadata changes (like mtime) will not be considered "changed" during a subsequent incremental backup. Currently, only ext2/3 volumes can be backed up (this limitation may be removed in an upcoming release). ACLs and extended attributes are NOT tracked.
The slice version is: http://www.slicehost.com/questions/#backups
The UI looks like this: http://img16.imageshack.us/img16/3251/linodebackups.png
For all I can tell, the Slicehost VPS may have been very old, and the Linode VPS may have been brand new on a fresh machine (so no one else was using it and eating into the disk access times).
The forums for both companies have customers talking about hdparm results and such, so it's worth looking there for a broader sample. In general, people testing an existing VPS against a new one (for either service) tend to see better results on the new one. It's just a fact of life if you get a VPS - the old hardware stays in service because customers still run on it, and those customers don't want unnecessary downtime. For best results, make a new slice on each, talk to the techs at each to make sure you're in their latest and greatest datacenter, and then see how things stack up.
It's also worth pointing out another caveat with a VPS - shared disk access. So for scientific results you'd also need a way to control for the number of users on the box, and their usage patterns. If some jerk on your machine consistently thrashes the disk with swap access, it's going to be a pain on any service until their admins see it happening and reboot that VPS.
I have been really happy with the rock solid service provided by Slicehost for me over the last year and a half.
I also have an account with Prgmr that has been beyond excellent over the last year for some of my more personal stuff.
Edit :
I just ran the tests on my commercial VM's.
2GB Slicehost --
Timing cached reads: 5370 MB in 1.99 seconds = 2692.83 MB/sec Timing buffered disk reads: 242 MB in 3.04 seconds = 79.51 MB/sec
512MB Slicehost --
Timing cached reads: 2118 MB in 2.00 seconds = 1059.00 MB/sec Timing buffered disk reads: 204 MB in 3.09 seconds = 66.02 MB/sec
2GB Prgmr --
Timing cached reads: 5424 MB in 1.99 seconds = 2720.95 MB/sec Timing buffered disk reads: 196 MB in 3.16 seconds = 62.05 MB/sec