Stuart MoncrieffLoad-related website failures happen every day. A small percentage of them make the news, a bigger percentage impact the company’s revenue or business reputation, and 100% of them are preventable. Sure, efficient code is important, configuration settings are important, hardware capacity is important, but modern applications are complicated, and there is no way of knowing your team got all of these things right without testing the whole system under load.

My name is Stuart Moncrieff, and I have found load or performance-related problems on every project I have tested. Over my career, I have led large consulting teams, been performance architect for billion-dollar projects, and have spoken at conferences around the world. For the last few years, I have been building MyLoadTest On-Demand to help more businesses avoid disasters with their mission-critical web applications.

I saw that there was a middle-ground between software companies that wanted to sell their customers a tool and leave them to figure out everything else for themselves, and consulting companies whose business is selling billable days and who want to make their customers dependent on them forever.

There were four principles that were important to me:

No more gatekeeping

You might need an expert to set up a load test, but you don’t need to be an expert to press the GO button once it is ready. Traditional consulting companies want you to pay for their consultant to push the button, but I wanted to give customers the option of doing it themselves with a self-service portal.

This is particularly useful for technical teams who want to spend a week on an iterative tuning exercise. They can re-run their load test after every configuration change and measure the improvement – without having to be dependent on a consultant for something they can do themselves.

Openness and transparency are important too, so I wanted to give customers the ability access the same test results and metrics as my consultants, rather than being dependent on a high-level summary report.

Available on-demand

In a perfect world, performance-related activities would occur throughout an application’s lifecycle – from early load testing during the initial build to ongoing load testing during the maintenance phase. But there are big gaps where there is no need for a performance expert to be involved, or where their involvement is only a few hours of work.

Traditional consulting companies want to park someone on a chair in their client’s office for solid blocks of time. This is partly what drives projects to do a single intensive test cycle just before their first public deployment, then ignore load testing once the system is live despite still making regular changes (that could break performance under load).

Switching to a remotely-delivered on-demand model, where small pieces of work are done when needed makes it easier to safeguard performance over the application’s entire lifespan.

More automation – testing at scale

Imagine that there was a common performance fault that you could check manually in about an hour, but could also be checked automatically if you spent a week of development effort. What do you do?

Clearly it’s a question of scale. You have got to use it at least 40 times to get a return on your investment, but that is going to take a really long time for most performance testing consultants.

One of the bonuses of operating at scale is that when a new type of problem is found with one project, then it makes sense to check all the other projects automatically.

Saving projects from reinventing the wheel

I was recently talking to some of the performance consultants that I had trained and mentored at a previous company. It seemed that they were all busy building performance testing pipelines – duct taping load testing tools to dashboards, Git repositories and CI build tools.

It is great that so many companies are embracing performance testing as part of their continuous integration testing, but it makes much more sense to build something like this once rather than spend months independently building the same thing at multiple companies.

The first thing that I had built at MyLoadTest On-Demand was a web services API for continuous integration, turning continuous performance testing into a simple integration instead of a months-long project for my customers.

It is not possible to build a service that is a perfect fit for everyone. The sweet spot for MyLoadTest On-Demand is public-facing web applications, and companies that have a need for expert load & performance testing, but are too small to have a full-time specialist. Please contact me if you think MyLoadTest On-Demand could help your project.