Performance Testing: Latency, Load, Stress, or Soak? | HostGator Blog

HostGator Blog

Web Hosting Made Easy!

Blog / Web Hosting Tips

Performance Testing: Latency, Load, Stress, or Soak?

Thursday, October 2, 2014 by

5 heavy load

How well does your web hosting perform? Are you sending web pages in a timely way to visitors to your site? Is your application correctly handling simultaneous requests? Do you actually know what can be measured – and which measures are relevant to your situation? Web hosting performance testing can give you valuable information that can let you keep visitors longer on your site, make sure you can accept the right number of simultaneous visitors, handle overload situations and detect possible design or programming deficiencies.


Latency or How Long It Takes to Get Back to a Visitor

Let’s start with the case of just one visitor (naturally, you’ll probably be aiming for rather more, but we’ll discuss this below.) Normally, you want the response time for that visitor to be as fast as possible. In other words, between the moment when the visitor clicks to send you a request and the moment the visitor sees your response, the least time possible should elapse. This ‘latency’ can however be determined by several different things, including the power of your web hosting platform, the size of your network connection and the power and network speed of your visitor’s computer. You can improve the first two, but the last two are out of your control (although keeping your web pages simple may help.)


Performance Under Load

Ideally, you should have an idea of how many visitors are likely to access your web hosting platform at the same time. If this is not feasible, then you should at least know how many average or typical users can actively work with your site simultaneously, and plan ahead for options to increase capacity if you need to. Different solutions, either free or paying, online or in-server, are available for conducting load tests with up to a few million simulated users or more. Whichever solution you choose, make sure your test is representative of both user numbers and types of activity, including number of pages called per hour, number of requests for database information, ‘think time’ and so on.


When It’s All Just Too Much

If your web site is significantly more popular than you imagined, your web hosting facility may not be able to cope with all the traffic. Then what happens? Does your site send out a polite apology about lower performance while stopping any new connections, or does it just crash without warning? Stress tests are designed to find out what (really) happens under conditions of excessively high loads. How much this affects you will depend on what kind of website you operate. A web site for a bird spotting association that simply crashes may just be an irritation. A web site selling hot new fashion articles that simply crashes could lose you important revenue and customer loyalty.


A More Technical Test

The soak test is done by starting your web site or application and leaving it to run (normally) for an extended period of time to see whether this produces any abnormal conditions either in the application or in the web hosting platform it runs on. One example would be memory leaks, a common enough problem when an application uses some main memory, but fails to return it for general use when it’s finished with it. Testing for these kinds of conditions typically requires technical expertise, for example by the person or team designing the application in the first place.


Finally, Who is the Judge of ‘Good Performance’?

Performance, ultimately, is all about making sure customers or end-users are satisfied with what they experience. Your web hosting platform may be supercharged in processor power and memory, yet they may still complain. Or it may be far more modest and still reply adequately to user expectations. Falling traffic and user comments on your blog (or similar) may indicate a problem, but prevention is always better than cure. There’s only one way to find out what users really want, and that’s to ask them. Armed with this information, you can then do the right performance tests and confirm or tweak afterwards, as appropriate.



Author Bio:
Natalie Lehrer is a senior contributor for CloudWedge. In her spare time, Natalie enjoys exploring all things cloud and is a music enthusiast. Follow Natalie’s daily posts on Google Plus, Twitter @Cloudwedge, or on Facebook.

Image source:

No comment