Media Temple vs Mosso vs Dreamhost Speed Test
Update 3/24/2009: As Chad mentions in the comments below, the comparison here is not really fair. We’ve been building some really high-traffic sites lately, and any virtual server will hit their limits under enough load. In other words, what’s the use of having The Fastest Server in Town if it crashes when it gets super-duper-boy-band-popular?
Update 3/24/2009: Because Mosso is missing some features that we desperately need for an upcoming job, we have been forced to leave them. I wasn’t happy about this, because their service is excellent, and the speed was great. We moved to a Dedicated-Virtual 3.5 - Base plan at Media Temple. For half the price, we get less disk space, and I was thinking it would be a bit slower. I was wrong. I’ve updated the chart below to reflect Media Temple’s speed test results for the same site.
This week was a bad one to be hosting with Dreamhost. I didn’t do any real tests to count downtime, but sites were slow and even inaccessible many times throughout the week. Sometimes, these fits would only last 10 minutes, but when you host ~30 sites on the same account, you will be notified if a site is down or painfully slow.
Let me make this clear: I know that Dreamhost is a value host, and it does give a great value for the money. They basically offer no limits on anything for $10/mo. That’s crazy cheap. Therefore, I wouldn’t expect them to offer the same speed or the reliability of a $100/mo host such as Mosso. This comparison is simply to share our site load time change when I switched one very large and demanding site from Dreamhost to Mosso.
(The site I tested is a Drupal site with caching turned off. (I tested it both ways on Dreamhost, but I can’t test Mosso with caching until it builds up a bit. I’ll post those results here as well.) The site is very heavy on images, and moderately heavy on CSS and Javascript. Nothing was changed on the site between the switch. The total page load size was 408.4 KB. Load time was checked using Pingdom Tools. I tested each host 5 times and these are the averages.)
16.4 seconds
7.97 seconds
5.9 seconds
As you can see, Mosso Media Temple loads the site in roughly less than half the time as Dreamhost. For 10X 5X the cost, I’ll take it. As a web developer, I must be able to depend on our host. Obviously, this test only shows speed improvement, but I am speculating that downtime will also be greatly reduced (Update: Now, the downtime is only a result of the configuration of LAMP, which is frustrating as well).
Have you had a similar or not-so-similar experience? Leave a comment below.
Printer-friendly version
Email to a friendAddToAny
Similar
- Having trouble targeting page titles with CSS when using Views?
- Which Open-Source CMS Has the Most Active Development Community? [Part Two]
- The University of Washington's Latest Drupal Project, Assisted by ENGINE
- New Drupal Nonprofit Project -- Redesigning the D.C. Open Government Coalition's Web Presence
- California Nonprofit Using Drupal: ENGINE Contributes to New GCIR.org
Look at this testimonial:
ENGINE is the first and only choice when it comes to SEO. Their efforts doubled my hit count nearly over night.
– Zachary Brown
Atlanta Artist
Click to view more about this project.
Subscribe to the ENGINE Blog
Hi! Click right here to subscribe via RSS to the ENGINE Blog.
It’s the rare small-company blog good enough to be shared by big-time publications like the Wall Street Journal, Slate, the Consumerist, and Kottke (not to mention tweeted by Drupal).
Or subscribe by email below!
Comments
No Rails
Mosso dropped their Rails support. FAIL.
Java
Yeah, I ultimately left because I needed Java.
Not a proper comparison
Hey Ben,
I noticed that you ran your comparison between the Mosso Cloud Sites platform and the Media Temple DV platform. Keep in mind that these are two entirely different infrastructures (true platform cloud vs a dedicated VPS). If you want to do a more accurate performance benchmark you should compare your MT DV server to a 256 Cloud Server from Mosso, which is actually about 1/5 the cost of your DV solution.
Cloud Sites segments its services into different clusters (web cloud / database cloud), so when testing a database driven site with no real load, such as a Drupal install you will always see slightly better performance when MySQL is running on the same local machine. However, when this site comes under any significant load, this is where you will see the true benefit of platforms such as Cloud Sites. It will be able to disperse the web and database load across multiple physical machines and automatically scale to handle any spike in traffic. A MT DV solution will simply get slower and slower or even crash when you hit the resource limitations of your virtual server.
There are other aspects to highlight as well such as greater redundancy since there are no single points of failure, but that is a whole different story.
True Chad (a year later)
You’re right about MT DV servers crashing under heavy load. You have to really know how to tune LAMP to get it super-stable. Even then, it will eventually run out of resources with enough load. Now, I’m just dreaming about a cloud system like Mosso with the speed and flexibility of VPS.
hmnn
wondering about download speeds for content. I run a site with huge traffic peaks for a bi-weekly podcast, anyone have any experience with any of these hosts and dl speeds, throttle nightmares, etc…
Speed Comparison
Don’t forget, it also matters WHERE you are and WHERE the colo is located as well. Latency will impact your test. So if your on the East Coast, that will hurt your test against dreamhost since they are on the west coast. Media temple is nice because they are in the central US (Texas I think???).
Anyway, just my 2 cents.
Since (mt) Media Temple is a subject of this conversation...
(mt) Media Temple has one data center in Los Angeles, and another in northern Virginia (near Washington D.C.). (dv) Dedicated Virtual servers are available in either data center - default provisioning is in VA. The new blazing (ve) Server is available in the VA data center and is extremely cost effective. It will also be available in CA data center later this year.
Load times from either data center are outstanding. You will get all of the resources that you subscribed to, and usually more. Still, you should definitely monitor your system because it will perform very well right up to the hard resource limit.
Stop relying on performance degradation to tell you when to upgrade - expect high performance all the way to the limit. Just remember, if you let it hit the limit, it will crash. You wouldn’t drive a high-performance race car all the way to “empty” then complain about it while sitting in the middle of the racetrack - so why would you do that with a high performance virtual server?
The joy of “fast” is amplified or lost according to skill - it doesn’t take that much to learn and the wiki will help - so go ahead, give it a shot and compare for yourself.
Post new comment