Those of you who have been working on building dynamic request intensive web services, know the importance of load testing and the challenges that come with it. You also know that somehow the test environment hardly ever is “really” an exact mirrowed version of the live environment. That said, load testing of ones services can be a tricky one. There’s also so many variables that can make the service to jam up. One can test successfully almost every function separately but when having all working together things start to happen. Those knots are usually not as easy to spot.
So, when I first run into Load Impact, a Swedish startup running its easy online load testing service for six months now, I must admit I was a bit sceptical at first.
I sat down with Ragnar Lönn, CEO and owner of Load Impact, to hear his story and the elevator pitch. (Btw. I always do that, it’s a great excercise and makes a good conversation, often ending up the startup having a better pitch and have thought through regarding killer features, customer focus etc.) Ragnar knew what he was talking about. He has a profound background when it comes to Internet traffic and system architecture. Back in 1994, he founded Algonet, Swedish ISP, and has since then been working with complex systems and performance testing.
Load Impact offering no barriers to entry for testing
Load Impact wants to be the easiest way to test web sites. No barriers to entry, just type an URL and push start. It’s free to try out and use without needing to register. To prevent unauthorized anonymous requests with the free account the service has an inbuilt locking system, so it’s no use to try any funny business. One can only run four tests per one particular URL with maximum 50 concurrent users. Since start they have delivered over 40 000 test results, of which 75% are run by anonymous users.
They recently launched Web Page Analyzer, a new feature to help web developers to go even deeper debugging their web pages. A kind of like Pingdom’s Full Page Test, but with some added features. It lets one choose between 15 different browser versions, and one can also hoover over the objects listed in the results to see them. With all the premium accounts one can also set a client bandwith limit, handy when testing sites and functions, e.g. for mobile device users who usually have less bandwith to play with. It also lets one filter out external URLs/objects.
The service doesn’t support executing javascript, but it’s possible to record a user session in which the javascript is being executed. Also any downloads made by javascript will be recorded. It’s only in certain cases, e.g. a javascript generates dynamic URL that the simulated load will differ from the load generated by a real browser.
There are three different premium accounts, primarly based on the number of concurrent users and test configurations.
Since it’s a web-based tool I was wondering how well that would work for all the developers with test environments behind the firewalls. Load Impact provides a dedicated IP address so as long as one’s willing to open up their firewall it can be done.
They’re also trying to reach out and be accessible to the developer community by blogging and with a new knowledge bank wiki.
Is there place for Load Impact
As it doesn’t seem too far when the cloud becomes the ideal testing infrastructure of the future, it strikes me how relatively few players there are to get a piece of the cake.
Loac Impact joins the Testing-as-a-service load testing space with companies like Loadstom, Browsermob, and venture capital funded Sosta. Keynote offers Kite, a free desktop-based test environment, only available for Windows users though. Besides that the prices vary, if even displayed, it’s clear there’s place at least in the more accessible and less costly segment. Sosta and Keynote also seem to be more specialized on testing applications and services prior to migrating to the cloud.
Again, easy is very hard to do. That’s what Load Impact is aiming to succeed on, enabling to do tech stuff almost without being a techie. With more services becoming cloud based, as well as easier to develop usings mashups, I believe the need to load test ones service, without having a major in systems sciences or being able to set up ones own test server, is to grow.
