Hear how Alfa tested NuoDB to ensure that it was the right database option for their platform.
“How did you determine that NuoDB was the right solution for you?”
Tim Gage: So there’s a Community Edition on the website, which we downloaded -- that was where we got started. We read some of the documentation, had a bit of a play with it, and we decided that it did look like it was going to be a pretty good fit for our product, so I asked one of the guys on my team to have a go at writing a dialect, so we have an abstraction layer between the system and the database, and I got one of the guys on the team to have a go at writing this dialect. That went very well. It took them about 10 days or so to get the majority of the support in, and basically we had the system up and running, and that went very well. We got, basically, pretty good results out of some basic testing, so we decided to move on to doing a full evaluation, and at that point we got in touch and requested an Enterprise trial license.
The Community Edition requires almost no configuration. You could download it, install it, and it just works. It works out of the box. It’s very good.
“How was your experience with the NuoDB Enterprise Trial?”
Tim Gage: Yeah. So deploying the Enterprise version with NuoDB with a minimally redundant configuration does require a little bit more configuration.
So we had a bit of help and assistance from some NuoDB consultants who came in and helped us out, got us started, basically. They essentially taught us how to set up the system and on multiple nodes. And then we went through a process of various kinds of testing that we wanted to do. We basically built up the volume, so we had a sentence -- we put in a realistic scale of database, something moderately large. We then started to push a load at the system using Web services and using the UI, and we’ve also got batch processes that we need to run. So we went -- basically ran through a series of tests. The kind of thing we did, we put use to this stuff, we do performance tests all the time. That gave us pretty good results. One thing, actually, we were particularly impressed by was the fact that the system requires very little configuration to get to the point where it’s running really well, and our deployment model is relatively simple, and almost out of the box, the performance is very good, very impressive.
So then once we’d got through that testing phase, we moved on to some failover testing, in order to test the various features that NuoDB has, like the ability to recover from a node failure. So we start to do things like kill transaction engines and storage managers and check that the system recovered correctly from those scenarios. We did encounter a few issues doing this, which was interesting, but we got good support from NuoDB. We had the consultants available. A couple of things required configuration changes to the system, to NuoDB or to Alfa, and then we retested and the problems went away. We did actually find a bug in NuoDB, and we had pretty good interactions with the support team, and it was great to test that out, actually, and we got some good results. And that bug has now been fixed and now it works.