Think twice about getting that Virtual Private Server

Published on August 7th, 2007 in Web

Virtual Private Servers have become a popular choice over the last couple of years. Products such as Virtuozzo, its open-source sibling OpenVZ and even XEN, have made it relatively easy for web hosting companies to implement virtual hosting technology for their clients at a reasonable cost. Virtual Private Servers are the poor man’s dedicated server, even though this might sound a bit harsh.

With a VPS, you get the ability to maintain and modify almost every aspect of the system (except the kernel). This is great if you have some special needs, like a web application that needs MySQL with InnoDB Storage support enabled (most shared hosting offers do not include InnoDB support) or you just want one for your own needs like backups or source control using Subversion where flexibility is more important than being able to support a huge amount of simultaneous requests. A VPS will serve you just fine in one of these cases.

But beware if you get a lot of visitors on your website. I have conducted benchmarks with several VPS hosts, none of them were as fast as the shared hosting provider I’m currently using. Sure I didn’t optimize the Apache, PHP and MySQL config files as much as my professional web host has, but all VPS’ were at least 10 times slower than my shared host. Even the big ones with 1024 MB of RAM. If you install all of the services your average shared hosting company provides (DNS, Mail, Anti-Spam, Anti-Virus), performance will decline even more.

So, given the fact that most people don’t need the flexibility provided by Virtual Private Servers (let’s be honest, a normal LAMP installation will do for most of your hosting needs), they get less speed and they have to do all of the administrative tasks themselves (using Plesk usually doesn’t get the whole job done). If you want to make sure your LAMP installation is secure, you will have to invest even more time.

I hope I could open your eyes for some issues if you’re going to use a Virtual Private Server. If you still think a VPS is right for you, check out different providers and ask them for a trial period or make sure, they provide short-term contracts so you can bail out after a few months if you’re unhappy with their service.

8 comments

OJ August 9th, 2007

I think this post is fairly misleading, and not really indicative of the plethora of options you have if you decide to invest in a VPS.

There are many providers out there offering different VPS services. There are those who will maintain and support your site for you without you having to invest any time. There are those who give you a bare-bones VPS which you have to build yourself. There are a million other options between these two extremes, and chances are that one of them will fit the scenario you’re looking for.

With a VPS, you get the ability to maintain and modify almost every aspect of the system (except the kernel).
That depends on the provider you go with. I have a VPS, and I have complete control over it, and that includes the kernel I’m running.

But beware if you get a lot of visitors on your website.
Why? I’ve seen many VPS hosts perform better than shared/dedicated hosts purely because the administrator has put effort into it’s configuration. Apache is a bloated beast, and most people don’t require anything more than a php/ruby parser, a static content server and a MySQL DB. Options such as lighttpd and litespeed are not only good options, they’re MUCH lighter and just as easy to configure.

If you run a VPS you don’t need to run DNS, mail, anti-spam and anti-virus. Again, then are many other options that you can utilise (Google Apps is a good example which removes the need for AV/AS/Mail).

Securing your LAMP installation is usually as simple as just making sure that you server is patched on a weekly basis. With most distributions these days, that’s almost a one-step process.

You say you’ve conducted benchmarks on non-optimal VPS servers, and then compared them to dedicated hosts. It’s hardly a surprise that your VPS systems are running slow. Even if you did optimise them, you’re using a stack of software that’s unnecessary, so it’s going to be slow regardless. Only run what you need.

Avoid using Plesk and the like, they’re as ungainly and resource hungry as Apache is.

In short, I think you’re painting a rather awful picture of a perfectly viable web hosting solution. There are many good options for VPS hosts, and they shouldn’t be discounted or avoided just because you haven’t taken the time to set them up properly.

Dennis Martinez August 9th, 2007

Hi,

I recently switched from a regular hosting company to a VPS service (due to poor Rails support in the hosting company, along with me being a control freak sometimes :), and even though it’s been only a week, I’m happy with the results I’m getting, at least for my specific needs. I think the deal with a VPS is that you’re going to have to take some time to tweak it just right. My VPS only has 128 MB of RAM, so I ditched Apache in favor of Lighttpd, which is much lighter on resources. I’ve also ditched Sendmail in favor of Postfix, which is much easier to configure. Right now, my VPS is running a web server, database server (MySQL), mail server (both SMTP and POP3 / IMAP) and DNS server, and it’s only using an average of 60 MB. Granted, my site is just a personal website with a low amount of visitors, but I’m sure it could hold up just fine. I’m planning to release a Rails site which should attract much more visitors, so right there I’ll see if the VPS has enough power for my site.

The bottom line: a VPS is great, but you’ll need to spend time to make it work great for you. If you don’t have the time (or the resources to get a sysadmin to do the work for you), then look for other alternatives. VPS are not for everyone, as you made clear. Thanks for the great post.

Yurii Rashkovskii August 9th, 2007

It seems that dedicated servers are quite cheap these days. I have one with 2GB RAM and 2.something GHZ Celeron for $120/mo. or something like that.

Aaron Saray August 9th, 2007

I think there is a flaw in this ‘investigation’ - and that is the benchmark of a standard VPS vs a configured shared hosting service. You mention 10x slower, but you never even gave any examples or benchmark figures (which is a huge no - no if you want to have any credibility). I also agree with the above posters in some of their statements (about running more software than needed, etc). Finally, a shared host will be faster in most cases against your non-customized vps installation - because they HAVE TO BE. They have tons of clients off of the shared server - so they try as hard as they can to squeak every last little bit of performance out of the installation - in case everyone starts using the resources. Because most shared hosting providers aren’t serving any high traffic sites, you get the benefit in your benchmarks for that configuration and tweaking they have been doing.

All in all man, I’d like to see an article like this backed up with real world examples, benchmark test results and other facts. I think it would help boost your credibility and make your point.

-aaron

Paul Davis August 9th, 2007

You make a good point but, often times there is more to a server than email and a web server.
I’ve run my site on a JVDS (http://jvds.com/) virtual server for about 5 years now.
It is very useful to have a root account on a linux box.
I can run whatever (or multiple) JVM I want, I can run whatever verion of tomcat I choose. I can set-up and destroy databases at a whim, CVS (then change to subversion), what ever. I’ve even ocasionally set up a proxy server to proxy over ssh.

I think you hit the nail of the head with “poor man’s dedicated server”, that’s really what it is. Sometimes, there are just some things you can do in a shared server.
There is another side to it too. I have to do all of my own maintenance, anything that happens is 100% my responsibilty. It won’t perform as well as a “real” server. But, I’m the type that likes it.

However, if somebody just want’s to serve up some pages, you are right, a simple shared LAMP setup will be cheaper and perform better.

Morale of the story, before shopping for hosting, identify your requirements and your budget.

James Coletti August 9th, 2007

When you say “several VPS hosts”, I’m curious to know who these are. We demo’d VPSLink (a subsidiary of Spry) last month, and their service and support has been top-notch, enough for us to sign a 12-month contract. VPSLink provides nearly a self-service solution, versus Spry which specializes in full managed solutions, so when I speak of VPSLink’s “support”, I’m referring to their rapid response time provisioning the server on sign-up (almost instanteous), adding PTRs for reverse DNS, and only a handful of other responsibilities. They’re not going to configure any services for you and in turn, you’re getting better prices. Granted, we’re a low-maintenance customer with special requirements and we’d much rather configure everything to our liking. If you’re not spending time performance tweaking and optimizing, you may not be a VPS candidate. In our case, Apache plays nice whereas MySQL is the beast. Apache can be huge if you compile it that way - thin it down and use only what you need. It all goes back to your statement about “having special needs”; that’s why virtual private servers exist.

Ron August 9th, 2007

I investigated shared vs. VPS solutions as well. For my needs, which includes a Tomcat instance, it was almost imperative that I went with a VPS. All the shared solutions I could find were very weak when it came to Tomcat support.

As far as VPS’s go, I got a real bad one at first. The performance was awful. Near the end of my 30 day trial I switched. The new VPS was almost 1400% faster!

I did some benchmarks on the old and new VPS, take a look if your interested: http://blog.evver.com/2007/08/05/a-faster-virtual-server/

–Ron

chris August 10th, 2007

So, you didn’t name the vendors you tried, nor the host you are using currently, nor any sort of method of how you ‘tested’ these numbers, etc. You didnt even give any sort of raw data either.

How did you perform your tests? When referring to slower, are you referring to html generation speeds, bandwidth, total user load time? What metric?

looks like you are using genotec as a host? Did you try their VPS?

Did you see where the breaking point was for both servers? While you may lose speed, maybe you make it up in # of connections being able to be handled before death.

most shared hosting throttles you significantly. Seen most often in a Digg effect like swarm of traffic usually brings any shared hosting to its knees.

Your claims are frankly just rediculous without any sort of explanation!