Zephyr offers a comprehensive suite of test management products which foster global project team collaboration, drives down costs, shortens release cycles, and provides real-time visibility into the status and quality of the software projects. They address the problems of making the transition to agile, challenges around globally distributed teams, slow release cycles, sub-standard quality, business decisions based on stale data, and ever changing toolsets We sat down with Shailesh Mangal, Zephyr’s CTO and VP of Engineering, to discuss the technology challenges his business faces and what led them to select Riak KV to meet these challenges.
This blog post is an extract from a video interview. For the full video interview please visit our Resources page. Here’s a short clip:
Shailesh, thank you for sitting down with us, tell us about Zephyr and the service you provide.
Zephyr offers a comprehensive software testing platform. We started about 8 years ago with a singular vision of providing integrated best of breed solutions in the software testing space. Our product portfolio helps software teams deliver a quality product in time.
We have six thousand customers spread across large enterprises, SMBs and mid-sized businesses and we do this globally – helping customers in approximately 100 countries deliver quality software better and faster.
When you were looking for a database for this service, what were some of the challenges you faced?
So some of the technological challenges that we took into consideration primarily were around scalability. It was very important for us to make sure that our solution remained horizontally scalable, with linear performances. The database needed to enable multi-tenant capabilities and it was essential that data be segregated for each tenant. We also needed to customize the database to our needs as we expanded.
Tell us a little about your process in evaluating databases?
In 2014 we started looking for a suitable solution database solution to support our growth. We looked at many solutions at the time. Our expertise had always been in RDBMS so we started with that and very quickly ruled it out as RDBMS would not scale for this new type of application we were building. We then looked at Cassandra, MongoDB, columnar databases, also document databases like CouchDB and in the end we chose Riak KV as it is the only solution that can be customized and truly scale. Amazon’s Dynamo has its own set of scaling problems for example if you need to dynamically add capacity you need to plan it out. This made it near impossible to auto provision our customers as we’d have to put in a request and wait for AWS approval.
We moved forward with Riak KV after reviewing all the different characteristics that are defined in documentation and everything that we were reading online looked really, really good but we weren’t sure if this is the right solution as we couldn’t find any reference points. So we reached out to peers in the industry I was surprised to learn that many of them were already using Riak in production with multiple node clusters, 30-40 node clusters, and that was kind of eye-opening to us. We found a lot more behind the curtain than we could see from outside.
And how is Riak KV performing for you in production?
We were able to get Riak KV up and running pretty much using all the online help documentation. We did work with Riak to validate our data structures, but going into production was extremely easy. We just followed the procedures provided online. Soon after we went into production we realized it would be good to have a constant support from Riak, so we contracted with Riak and we now have 24 by 7 support which, by the way, has been great.
Our experience with Riak KV has been great. During seven months of operation, we have added approximately 30 times more data than we added in the first month and we’ve been able to scale the entire cluster. The self-managing aspect of Riak has been great, the cluster stabilizes itself, and the entire management is completely automated. Without incurring any additional DevOps cost we’ve been able to scale the system.
Any advice for those embarking on similar journeys?
Ah yes, we moved from RDBMS and it was a difficult pill to swallow as we didn’t have any experience but after we have done it we definitely are glad that we did and we would like to encourage everybody to consider NoSQL especially Riak. When you are developing your revenue impacting business critical apps, continuous growth and innovation is very important and we very strongly feel that we got all of that with Riak.