The word cloud is getting a lot of attention lately, almost like web 2.0. Marketing gimmick aside, I believe cloud computing does offer a tremendous amount of advantages over traditional computing environment.

At Envato, we use a number of providers for our server needs. We currently have servers with Media Temple, Slicehost and Engine Yard.

Even though Envato was already using Slicehost for some of their applications, I still did my fair share of research before making a suggestion to move away from Media Temple.

I have eventually nailed down to three service providers: Slicehost, Linode and VPS.net.

For Envato, we decided to go with Slicehost, because we already have some servers there, and they offer bandwidth pooling. We would have given Linode a deeper look if they had offered backups (their backups system is still in beta).

Having said that, I was very very impressed by VPS.net, so I decided I should sign up anyway, for my personal projects.

Let’s compare the three services: Slicehost, VPS.net and Media Temple.

Vertical Scalability

As some of you might know, Envato serves over 12 million page views a month, and is growing rapidly. Scalability is becoming more and more important to us. This is the primary reason why we opt to using cloud servers.

Slicehost

Slicehost is one of the first cloud VPS providers out there, they are reputable and reliable. Scalability is based on their VPS plans, which is flexible, but not flexible enough, compared to VPS.net’s offerings.

VPS.net

VPS.net, a new service offered by the UK2 Group. Scalability is offered on a node-basis, a node consists of 0.4Ghz CPU power, 256MB RAM, 10GB storage and 250GB bandwidth. As you can see, this is much more flexible than Slicehost’s plan based approach.

Media Temple

Unfortunately Media Temple is very limited in terms of vertical scalability. (dv) has only three options, the most expensive one costs $150/m. If you need any more power, you’d have to fork out $750/m for a Nitro server.

Virtualisation Technology

Virtualisation technology usually isn’t a big deal, because for the less system-admin-minded people, they all look the same. If you’re keen to finding out the differences, have a read at this post. We much prefer to use Xen over Virtuozzo.

Slicehost

Xen.

VPS.net

Xen.

Media Temple

Virtuozzo.

Automated Backups

Backup is a crucial part of the server architecture. We therefore prefer a service provider that offers automated backups. This way we will have at least two offsite backups, one provided by the VPS and one by ourselves.

Slicehost

Slicehost offers automated backups at a premium cost compared to VPS.net. A 2GB slice costs $30/m, and that’s the most you can get. If you have a slice larger than 2GB, there is no backup service offered.

A maximum of three snapshot slots are offered, you can either have automated ones (daily and weekly) or manual ones.

VPS.net

VPS.net offers a more affordable backup solution, it only costs $5/m, for VPSes of any size.

You get three automated snapshots (daily, weekly and monthly) and one manual snapshot.

Media Temple

Media Temple does not offer any backup solutions.

Control Panel

Control panels help increase our productivity, so we do prefer to use them provided they don’t use up too much system resources.

Plesk is very similar to CPanel in terms of system resource usage, so even though it is powerful, we don’t really like it. ISPmanager on the other hand, claims to only use 10MB of system memory. I like it a lot, so I decided to use it on Slicehost. Be aware that it does have some glitches, I have found a minor issue when using with a remote database server.

Slicehost

No control panels are offered.

VPS.net

Free ISPmanager PRO.

Media Temple

Plesk, for Nitro servers.

System Images

One clear advantage of using a VPS over a dedicated server, is the ability to quickly deploy and redeploy server operating systems.

Slicehost

Offers a wide range of operating systems, in both x86 and x64.

VPS.net

Offers a wide range of operating systems, in both x86 and x64.

Media Temple

Offers 32-bit CentOS 5 by default. I was truly shocked at this decision. The Nitro server has 8GB of RAM, you’d think that they would know people will most likely want to use more than 4GB of RAM for their MySQL server. ‘Switching to 64-bit is available upon request’, according to their FAQ page, but why? Why don’t they make 64-bit the default?

VPS Controls

Same as the operating system, a VPS usually comes with handy tools to start/stop/reboot the server.

Slicehost

Soft reboot / hard reboot / image rebuild.

VPS.net

Graceful shutdown / force power off / graceful reboot. The rebuild feature is estimated to be available in September. I still find it weird that this feature is not built-in in the first place.

Media Temple

Reboot / reinstall (revert to default setup).

DNS Management

All three providers offer DNS management.

Slicehost

To put simply, Slicehost offers an awful awful interface for DNS management. We have a bunch of domains we need to migrate over, and I find it extremely irritating to click back and forth between pages. Terrible, terrible user experience.

VPS.net

VPS.net’s DNS management interface is a delight to use!

Media Temple

Media Temple’s DNS management interface is also very nice!

Billing

Is usage calculated daily or monthly?

Slicehost

Monthly billing with usage calculated daily.

VPS.net

Monthly billing with usage calculated daily.

Media Temple

Monthly billing with usage calculated monthly.

Account

Are we allowed to have multiple logins (i.e. one person can handle the billing, another can handle the VPS management)?

Slicehost

Yes.

VPS.net

No.

Media Temple

Yes.

Extra Features

Let’s see what extra features they have. :)

Slicehost

Web-based console for troubleshooting, simple stats (CPU %, CPU time, Disk read/write, Network in/out), slice cloning, rescue mode, root password reset.

VPS.net

Web-based console for troubleshooting, CPU usage charts, bandwidth charts, scheduled upgrades (temporary resource burst), additional IP addresses, dotDefender WebApp Firewall, Cannybill, one-time managed service.

Media Temple

Bandwidth report (just a simple list), root password change, installation of developer tools (gcc, make, etc).

—

Overall VPS.net has more features and is well priced, I am so far very impressed. Some of the areas they could improve are:

• Image rebuild
• Multiple logins under the same account
• Root password reset

Let’s hope competition drives better products. :)

*Disclaimer* I am not affiliated with any of the providers mentioned in this post.

Related posts

• Benchmark: LiteSpeed vs Apache (PHP and Plain HTML) (11)
• [How-To] Guide to Install LiteSpeed with or without Apache on LxAdmin Host-In-A-Box (1)
• Notes on Choosing a PHP Framework: A Quick Comparison of CodeIgniter and Kohana (49)
• Choosing a PHP Framework Round 2: Yii vs Kohana vs CodeIgniter (48)

Dramatic improvements in performance and SMP scalability shown by various database and other benchmarks, in some cases showing peak performance improvements as high as 350% over FreeBSD 6.X under normal loads and 1500% at high loads. When compared with the best performing Linux kernel (2.6.22 or 2.6.24) performance is 15% better.

Ah I wish I had enough time to play with it. Perhaps when my VPS providers upgrade their offerings I’ll give it a try. ;)

Related posts

• Why should Linux triumph over Windows? (14)

Tool used: ApacheBench, Version 2.0.40-dev <$Revision: 1.146 $> apache-2.0

Benchmark object: My WordPress blog (homepage), of course I made an exact copy of the current blog you’re seeing to my VPS box (which has both Apache and LiteSpeed installed).

Both Apache and LiteSpeed use default configuration parameters. Since the benchmark was done on the same server, hardware is the same, however there are some software differences.

Server hardware information: Intel(R) Core(TM)2 Quad CPU @ 2.40GHz, 256MB RAM (remember it’s a VPS, not a dedicated server).

Server software information: Apache 2.2.3 + PHP 5.2.1 (mod_php), LiteSpeed 3.3.5 + PHP 5.2.5 (LSAPI 4.5) + XCache 1.2.1 + Suhosin 0.9.23

Yes the comparison isn’t totally fair since the LiteSpeed installation has all the bells and whistles (XCache in particular), but it gives you an idea of how the two web servers perform.

Firstly, let’s take a look at dynamic content.

Benchmark command: ab -n50 http://url/to/my/vps/wordpress

Apache

Server Software:        Apache/2.2.3
Server Hostname:        Not tellin' ya! :P
Server Port:            80

Document Path:          /wordpress/
Document Length:        69626 bytes

Concurrency Level:      1
Time taken for tests:   46.295309 seconds
Complete requests:      50
Failed requests:        0
Write errors:           0
Total transferred:      3495800 bytes
HTML transferred:       3481300 bytes
Requests per second:    1.08 [#/sec] (mean)
Time per request:       925.906 [ms] (mean)
Time per request:       925.906 [ms] (mean, across all concurrent requests)
Transfer rate:          73.72 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    0   0.0      0       0
Processing:   794  925 163.7    875    1474
Waiting:      765  910 160.9    869    1447
Total:        794  925 163.7    875    1474

Percentage of the requests served within a certain time (ms)
  50%    875
  66%    901
  75%    919
  80%    945
  90%   1029
  95%   1430
  98%   1474
  99%   1474
 100%   1474 (longest request)

LiteSpeed

Server Software:        LiteSpeed
Server Hostname:        Not tellin' ya! :P
Server Port:            80

Document Path:          /wordpress/
Document Length:        69626 bytes

Concurrency Level:      1
Time taken for tests:   46.565238 seconds
Complete requests:      50
Failed requests:        0
Write errors:           0
Total transferred:      3495500 bytes
HTML transferred:       3481300 bytes
Requests per second:    1.07 [#/sec] (mean)
Time per request:       931.305 [ms] (mean)
Time per request:       931.305 [ms] (mean, across all concurrent requests)
Transfer rate:          73.30 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    0   0.0      0       0
Processing:   701  930 171.0    882    1598
Waiting:      407  639 120.7    597    1153
Total:        701  930 171.0    882    1598

Percentage of the requests served within a certain time (ms)
  50%    882
  66%    917
  75%    962
  80%   1052
  90%   1158
  95%   1264
  98%   1598
  99%   1598
 100%   1598 (longest request)

Interestingly enough, the performances are on par! I am a little surprised actually, since LiteSpeed has XCache applied. It perhaps has something to do with the VPS (it’s a little slow), and of course the Standard version of LiteSpeed does not perform as well as the Enterprise version.

Take a look at the result of one of my other shared hosting accounts (just for reference):

Server hardware information: 8-core Intel(R) Xeon(R) CPU L5320 @ 1.86GHz, 12GB RAM

Server software information: Apache 2.2.6 + PHP 5.2.5 (mod_php) + eAccelerator 0.9.5.2 + Suhosin 0.9.23 + Zend Optimizer 3.3.3

Apache only (don’t have LiteSpeed on this server)

Server Software:        Apache/2.2.6
Server Hostname:        Not tellin' ya! :P
Server Port:            80

Document Path:          /wordpress/
Document Length:        60370 bytes

Concurrency Level:      1
Time taken for tests:   10.540660 seconds
Complete requests:      50
Failed requests:        0
Write errors:           0
Total transferred:      3038000 bytes
HTML transferred:       3018500 bytes
Requests per second:    4.74 [#/sec] (mean)
Time per request:       210.813 [ms] (mean)
Time per request:       210.813 [ms] (mean, across all concurrent requests)
Transfer rate:          281.39 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    0   0.0      0       0
Processing:   191  210  46.4    203     514
Waiting:      124  140   6.7    141     160
Total:        191  210  46.4    203     514

Percentage of the requests served within a certain time (ms)
  50%    203
  66%    204
  75%    205
  80%    206
  90%    215
  95%    219
  98%    514
  99%    514
 100%    514 (longest request)

Now let’s take a look at how static content compete. The target object is still my blog homepage, but this time with WP Super Cache enabled!

Benchmark command: ab -n10000 -c50 http://url/to/my/vps/wordpress

Apache

Server Software:        Apache/2.2.3
Server Hostname:        Not tellin' ya! :P
Server Port:            80

Document Path:          /wordpress/
Document Length:        69700 bytes

Concurrency Level:      50
Time taken for tests:   37.19618 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      700449648 bytes
HTML transferred:       697766700 bytes
Requests per second:    270.13 [#/sec] (mean)
Time per request:       185.098 [ms] (mean)
Time per request:       3.702 [ms] (mean, across all concurrent requests)
Transfer rate:          18477.55 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    0   4.0      0     101
Processing:     0  183 197.9    145    2693
Waiting:        0  134  90.7    128    1086
Total:          0  183 197.8    146    2693

Percentage of the requests served within a certain time (ms)
  50%    146
  66%    156
  75%    164
  80%    170
  90%    215
  95%    485
  98%    852
  99%   1159
 100%   2693 (longest request)

LiteSpeed

Server Software:        LiteSpeed
Server Hostname:        Not tellin' ya! :P
Server Port:            80

Document Path:          /wordpress/
Document Length:        69700 bytes

Concurrency Level:      50
Time taken for tests:   12.495997 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      700889164 bytes
HTML transferred:       698324300 bytes
Requests per second:    800.26 [#/sec] (mean)
Time per request:       62.480 [ms] (mean)
Time per request:       1.250 [ms] (mean, across all concurrent requests)
Transfer rate:          54774.50 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    2   6.2      1      55
Processing:     2   59  27.4     59     401
Waiting:        1   55  27.1     56     399
Total:          2   61  27.0     60     401

Percentage of the requests served within a certain time (ms)
  50%     60
  66%     61
  75%     62
  80%     64
  90%     76
  95%     86
  98%     94
  99%     97
 100%    401 (longest request)

Now that’s better! :) LiteSpeed is three times faster than Apache at serving static pages! Not bad huh?

In general, LiteSpeed is much better at serving static content than Apache. Dynamic content however, I don’t see much performance difference. Although it will be interesting to see benchmarks done on the Enterprise version on higher end hardware.

Some Lighttpd users let Apache to serve PHP content because it’s more stable and has better application compatibility. LiteSpeed on the other hand, is very Apache-compatible and very stable. I would recommend it to any high profile website owners. :)

Related posts

• [How-To] Guide to Install LiteSpeed with or without Apache on LxAdmin Host-In-A-Box (1)
• Ruby on Rails, Passenger (ModRails) vs CodeIgniter and Kohana (15)
• Cloud Hosting: Slicehost, VPS.net and Media Temple Comparison Review (14)
• WordPress Plugins Recommendation List (2)
• Release: [Kohana Module] Authlite, for User Authentication (21)