To provide employees with the critical need of interactive querying, we’ve worked with Presto, an open-source distributed SQL query engine, over the years. Operating Presto at Pinterest’s scale has involved resolving quite a few challenges like, supporting deeply nested and huge thrift schemas, slow/ bad worker detection and remediation, auto-scaling cluster, graceful cluster shutdown and impersonation support for ldap authenticator.

Our infrastructure is built on top of Amazon EC2 and we leverage Amazon S3 for storing our data. This separates compute and storage layers, and allows multiple compute clusters to share the S3 data.

We have hundreds of petabytes of data and tens of thousands of Apache Hive tables. Our Presto clusters are comprised of a fleet of 450 r4.8xl EC2 instances. Presto clusters together have over 100 TBs of memory and 14K vcpu cores. Within Pinterest, we have close to more than 1,000 monthly active users (out of total 1,600+ Pinterest employees) using Presto, who run about 400K queries on these clusters per month.

Each query submitted to Presto cluster is logged to a Kafka topic via Singer. Singer is a logging agent built at Pinterest and we talked about it in a previous post. Each query is logged when it is submitted and when it finishes. When a Presto cluster crashes, we will have query submitted events without corresponding query finished events. These events enable us to capture the effect of cluster crashes over time.

Each Presto cluster at Pinterest has workers on a mix of dedicated AWS EC2 instances and Kubernetes pods. Kubernetes platform provides us with the capability to add and remove workers from a Presto cluster very quickly. The best-case latency on bringing up a new worker on Kubernetes is less than a minute. However, when the Kubernetes cluster itself is out of resources and needs to scale up, it can take up to ten minutes. Some other advantages of deploying on Kubernetes platform is that our Presto deployment becomes agnostic of cloud vendor, instance types, OS, etc.

#BigData #AWS #DataScience #DataEngineering

While there's been some very clever techniques that has allowed non-natively supported geo querying to be performed, it is incredibly slow in the long game and error prone at best.

MySQL finally introduced it's own GEO functions and special indexing operations for GIS type data. I prototyped with this, as MySQL is the most familiar database to me. But no matter what I did with it, how much tuning i'd give it, how much I played with it, the results would come back inconsistent.

It was very disappointing.

I figured, at this point, that SQL Server, being an enterprise solution authored by one of the biggest worldwide software developers in the world, Microsoft, might contain some decent GIS in it.

I was very disappointed.

Postgres is a Database solution i'm still getting familiar with, but I noticed it had no built in support for GIS. So I hilariously didn't pay it too much attention. That was until I stumbled upon PostGIS and my world changed forever.

My data was inherently hierarchical, but there was not enough content in each level of the hierarchy to justify a relational DB (SQL) with a one-to-many approach. It was also far easier to share data between the frontend (Angular), backend (Node.js) and DB (MongoDB) as they all pass around JSON natively. This allowed me to skip the translation layer from relational to hierarchical. You do need to think about correct indexes in MongoDB, and make sure the objects have finite size. For instance, an object in your DB shouldn't have a property which is an array that grows over time, without limit. In addition, I did use MySQL for other types of data, such as a catalog of products which (a) has a lot of data, (b) flat and not hierarchical, (c) needed very fast queries.