Geraldo Thomas Co-CEO at VTEX
Thanks Geraldo for talking with us today. Can you please explain what VTEX does?
We are a software as a service e-Commerce company. We provide a common platform to a variety of retailers in Brazil and Latin America. We generate a GMV, general merchandise value, of almost half a billion dollars per year that flows into our product. We serve more than 300 clients, including a lot of multinational companies – basically anyone who wants to conduct e-commerce in Brazil.
What does your technical infrastructure look like, and how do you support your customers with the platform? I’m assuming you run everything in the cloud.
Yes, our entire infrastructure is hosted there. All of our system exists in an application that we host by ourselves. We have more than 400 instances running in Amazon.
Do you run across multiple cloud availability zones so you can ensure disaster avoidance and constant uptime?
Yes. We actually run multiple availability zones for that very reason.
What kind of business challenges caused you to move toward NoSQL technology and specifically Cassandra?
We needed to ensure that we could scale our application very easily and maintain strong availability. We had some big concerns about the availability of our database, and that’s why we invested so much in your NoSQL technology.
Did you try initially to use relational database technology like either Oracle or MySQL?
Sure, for a portion of our use case, but not all user cases. We started very lean with a Microsoft SQL server and then we grew. SQL Server is not as reliable as we want it to be for availability. That’s why we’re investing so much in NoSQL.
Did you also look at other NoSQL technologies like, MongoDB or HBase or any of those?
We tried DynamoDB.
Tell me then how you use Cassandra and or DataStax Enterprise today? What portion of your application does it power?
We have a metric aggregator that we use for taking metrics over all of our environment, both optional metrics and business metrics as well. We have to cache conversional metrics for our clients and basic metrics which we aggregate and adopt into our company installation. We also use Cassandra to streamline all of our client requests. We have maybe 1 billion events per year and all these stations use—most of the information that is needed after analytic is installed in the Cassandra collective as well. We call this part a required cache.
We are also going to use Cassandra for our log aggregation.
What caused you to move to DataStax Enterprise instead of just open-source Cassandra? It sounds like the log aggregation and the search capabilities were the key thing there.
The search capability was key, yes.
In terms of your administration, how many people do you have working on Cassandra and do you find it pretty easy to take care of?
That’s a great aspect of Cassandra. It runs by itself. We don’t have anyone who is dedicated to managing the Cassandra cluster.
Do you feel you’ve saved money using DataStax Enterprise over some of the other databases that you looked at?
Frankly, for us it’s not about the cost; I am much more concerned that everything remains safe and available. So other than that, cost is not as big of a concern for me compared to availability.