You are here

Benchmarking Density & Efficiency

Error message

Note: This blog was published over a year ago. Content may be out of date.

We’re talking about a few new ideas this week here on the NuoDB techblog. We hope you’re enjoying it!

One of those is a new kind of benchmark that we’re working on. Traditionally benchmarks tend to focus on numbers like Transactions Per Second: metrics that measure the speed or size of a database. These are definitely important, and later this spring we’ll be publishing more numbers in this vein.

Today, however, we want to talk about a new kind of benchmark that we think is also important. These are benchmarks that try to measure something about real-world load, and how to optimize a system to ensure that you can meet certain levels of density, availability and unpredictable spikes. In other words, we’re trying to think about how you measure and quantify a system not in terms of how many transactions you can pump through it, but how many resources it consumes to solve a specific problem.

Given modern data centers, public clouds and hosted services we believe it’s critical to be able to talk about how effectively you can solve a class of problem without making it about the specific application or underlying hardware. If you read about how we’re supporting hibernation and bursting you can see some of the justification for this. We want to know that specific problems can be addressed at a minium cost. We also want to be able to talk about how many independent users can be serviced in a minimal server footprint.

In our hibernation example, I cite blogs as a concrete example. We think this is a great example of a real-world, quantifiable application that hosting providers need to support. Blogs have known access patterns, actions and work-load. It may not matter as much how quickly you can respond to a blog request as it does how efficiently you can support a large number of blogs and how easily you can handle spikes in popularity.

I know all that sounds a little vague. Tomorrow we’ll be posting a concrete motivator for this work. Later in the week, look for more details about exactly what we’re doing.

More importantly, later in this week look for our shout-out to possible collaborators! We’re trying to build a new set of benchmarks that quantify real-world use-cases. For that, we want real-world data. If you’re reading our blog, if you want to work with us and if you think you have experience with hosting applications then we want to talk.

Look here tomorrow for an update and links to some concrete examples of what we’re talking about…

Add new comment