Everybody understands that it’s important to monitor a website for uptime, after all if your website is down you’re down. But it’s not as simple as that. You also need to monitor your server’s response times. That is the time it takes from the time you request a webpage to the time loads in your web browser. The longer the response time the worse the user experience. No one wants to wait for a page to load anymore, we’re no longer living in the dial-up modem era.
And for those of you who use Google AdWords, you’ll really want to pay careful attention to your website’s response times. Google has recently indicated that a website’s response time will be incorporated into the Quality Score. And any affects in the Quality Score will therefore affect the eligibility and minimum prices for your ads!
The key to monitoring a website’s response time is to get an average from multiple sources on a continual basis. You need to take the average response time throughout the day to compensate for when your server is handling more requests, and thereby responding slower. You also need to measure the response times from different sources because some paths on the net may get bogged down for reasons unrelated to your web server’s performance.
So how do you this cost effectively? I recommend using a website monitoring service. For my company LandlordMax we use Pingdom, and we’re extremely happy with them. What initially attracted us to their service was Pingdom’s blog, which I recommend reading if you have a chance. There’s lots of great material in there.
In any case a good average response time should be well under 500 ms per request. You can find our uptime and response times graphs for LandlordMax below:
As you can see there are always some issues. In the first graph, the uptime graph, our website was down for a total of 5 minutes for the month of March. Although we want 100% uptime, it rarely happens, not without excessive costs. To be honest, until you really monitor your website you probably won’t realize the real uptime. A lot of people are surprised when they see these graphs! They never really thought that their websites were ever down. But if you think about it, OS patches often cause a 5 minute outage. Even with load balancers you can go down for a bit as you tweak the load balancers themselves.
Another great benefit with these types of service is that they will contact you when your website does go down. They’ll contact you, depending on your preferences, by email, by text message, and so on. For us we’ve got it checking our server every 5 minutes, so I personally know within 5 minutes by text message when our server goes down. It’s great peace of mind!
But going back to our discussion about response times, on average our (LandlordMax‘s) response times are about 200 ms. That’s great! We’re very happy with this. We had a couple of glitches were the response times exploded to 1 second, but overall it’s a very good looking graph. And even 1 second is better than many many websites out there.
On Pingdom’s blog, they’re kind enough to provide some graphs to show the dramatic differences in response times some websites have experienced. The following shows a very significant difference in response time starting at 4pm. Based on the blog’s comment, the difference in response time is because they switched hosting services.
And here’s another graph from Pingdom which shows what a dramatic difference optimizing your webpage code can have on your response times:
Obviously there’s a lot you can do to optimize your website’s response times. It’s worth monitoring, especially if you use Google AdWords!