The Stats - Linode Vs. Slicehost ...... And The Winner Is…...

25 April 2009 | talking about Infrastructure

I long time ago I went to a linux user group (plug) and there was a fresh new company giving a presentation on this way cool new technology called User Mode Linux or UML. They created this company using virtualization to offer dedicated hosting to users. More or less having many different instances of Linux operating systems all running on one server. Each instance is owned and operated by a different customer. I was so taken in by this idea I became a customer right way. I was a customer for many years as a matter of fact. However I was never really that happy with it. Just seemed to be slow and sluggish. So I finally left and went on to a Mac Mini hosting which was interesting but owning the hardware and having \"remove hands\" just annoyed me. So I moved to shared hosting. Well that just sucked!!! So I started to get interested in these VPS dedicated hosting company. Thinking that maybe UML has gotten better with time. I found out the UML is a thing of the past and is now been taken over by Xen and all these VPS hosting companies are using Xen which seems to be way faster! So I thought I would give it a try. I looked at all the companies and concluded that SliceHost and Linode were the two best companies on the market.

I now have an account with both companies. I use them both and over all am very happy with both. However, Linode tends to standout just a little bit more. After reading this blog posting Slicehost Vs. Linode I started to do some performance testing of my own. Bang for your buck Linode wins and performance wise Linode wins. My slice host account is 512 megs and my linode is 360 megs. In my testing I never used more ram that 250megs so this test is valid.

The first thing I wanted to look at was the 32bit vs. 64bit kernel memory issue. The 64bit kernel takes twice as much memory. So Slicehost does not offer a 32bit kernel so you are stuck wasting more memory. Here are my results with both accounts loaded nearly identical.

### Linode ###
root@stowe:/var/www# free -m
total used free shared buffers cached
Mem: 360 348 11 0 5 270
-/+ buffers/cache: 73 286
Swap: 255 0 255

### Slicehost ###
root@slicehost:~# free -m
total used free shared buffers cached
Mem: 512 461 50 0 16 303
-/+ buffers/cache: 141 370
Swap: 1023 0 1023

Both these have a web server with the same configs and about a gig of mysql data loaded into the database. They are running idle.

The next thing I wanted to look at was over all bandwidth and network speeds. So I did a very simple wget test. I was wget on a remote machine 6meg mp3.

### Slice Host
spicer@jam$ wget slicehost/satan.mp3
--08:32:46-- http://slicehost/satan.mp3
=> `satan.mp3.3'
HTTP request sent, awaiting response... 200 OK
Length: 5,721,420 (5.5M) [audio/mpeg]

100%[====================================================>] 5,721,420 1.24M/s ETA 00:00

08:32:51 (1.16 MB/s) - `satan.mp3.3' saved [5721420/5721420]

### Linode ##
spicer@jam$ wget linode/satan.mp3
--08:35:31-- http://linode/satan.mp3
=> `satan.mp3.4'
HTTP request sent, awaiting response... 200 OK
Length: 5,721,420 (5.5M) [audio/mpeg]

100%[======================================================================>] 5,721,420 2.69M/s

I ran the test many times and did conclude that Linode was about 1M/s faster. I am not sure where my slice host node is so this test might be thrown off by the location of the different data centers.

Next and lastly, I did http_load tests. http_load got a php script that loaded a much into memory and then printed it out.

## Linode ##
spicer@jam$ ./http_load -parallel 100 -seconds 10 url.txt
365 fetches, 100 max parallel, 5.57176e+07 bytes, in 10.0029 seconds
152651 mean bytes/connection
36.4896 fetches/sec, 5.57017e+06 bytes/sec
msecs/connect: 270.134 mean, 448.058 max, 71.148 min
msecs/first-response: 353.144 mean, 745.521 max, 76.143 min
HTTP response codes:
code 200 -- 365

## Slice Host
spicer@jam$ ./http_load -parallel 100 -seconds 10 url.txt
42 fetches, 100 max parallel, 6.72844e+06 bytes, in 10.001 seconds
160201 mean bytes/connection
4.19959 fetches/sec, 672779 bytes/sec
msecs/connect: 59.8185 mean, 60.085 max, 59.535 min
msecs/first-response: 234.224 mean, 582.46 max, 68.179 min
HTTP response codes:
code 200 -- 42

Once again Linode was just a little bit better.

Over all Linode is a better bang for your buck, seems to have as good or better stats than Slicehost. Linode also offers more options with a much better control panel. If you liked this blog posting please feel free to use my referral code to signup Sign Up For Linode

Lastly, one of the things that held me back from a dedicated service is I hate being the admin of email and other web services. So I keep a shared hosting account to manage my email. Allowing me only to run the services I really need full control over.

tags: web hosting, vps, linode, slicehost

blog comments powered by Disqus