Since the very first day this website has been hosted at MediaTemple. Back then it was a simple shared hosting account for a good price and a lot of good and well earned hype. Almost a year ago they started switching to a grid infrastructure which was said to be able to allocate enough computing power for every grid member no matter what the neighboring cells are doing.
The performance of their database server has been worsening ever since. Here are some facts from my own experience of using the Grid Service (GS) hosting package:
- Several times a week the execution time for database queries is equal to the number of queries in seconds (for example, 20 queries take 20 seconds). An average is around 2 seconds for 30 queries.
- When browsing through their testimonials I found a WordPress based blog for which it takes 9.5 seconds to generate the index page (see at the end of
HTMLsource). Another WordPress powered blog listed in their testimonials mentions 14 queries, 12.539 seconds in its source. - According to Pingdom the uptime has been 99.45% (3h 40sec down) during the past month. But why does it take 4 seconds to load only the HTML?
While the uptime is fair and good, the MySQL database performance is beyond acceptable. I have read that support inquiries regarding the database performance are usually returned with a suggestion to buy a dedicated MySQL Container. So I did — in hopes that the problem lies in choosing the cheapest solution for such a complex engineering problem as web hosting. However, it is worth noting that a personal blog with 400 daily visitors may well be applicable for the cheapest of the good solutions.
Having used a dedicated MySQL container for two days, I can attest that the query times are still over 2 seconds, while the long peaks seem to be gone.
Here is their official system incident blog (which still uses WordPress 2.2.3). Having subscribed to its RSS feed (view an HTML version) it seems that not even two days pass without an incident of some sort. Although the outages may not have an impact on all grid instances, they still are a testament to the fact that problems arise more often then they should.
The Solution
If the performance doesn’t improve by the end of March, I will leave MediaTemple for Dreamhost, where I opened a hosting account with a dedicated MySQL server only to test if its really MediaTemple’s fault.
On Dreamhost it is 13 queries, 0.422 seconds, which is so much better.
I feel just the opposite! I was with Dreamhost and had an awful experience. My sites were slow to load and were down every other time I check them. Dreamhost was constantly apologizing for updates and downtime, but it didn’t stop. They only allow email support, which usually took 48 hours to get a (not so informative) response from.
At one point, I noticed ALL my sites had been down for 3 days. I had changed nothing at all. When Dreamhost support finally got back to me, they claimed they “accidentally changed a server setting”. No more details, and a half-ass apology.
Media Temple has been really good to me so far … the support is amazing and can be reached any time by phone. Their support is instant, which is necessary for me and my sites as I can’t afford for them to be down or performing poorly.
I’m curious as to the specific issue you’re talking about – I do wonder if my sites are experiencing the same problem and I just don’t realize it. How would you recommend I check into this further?