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

When adding a new feature to Checkly rearchitecting some older piece, I tend to pick Heroku for rolling it out. But not always, because sometimes I pick AWS Lambda . The short story:

• Developer Experience trumps everything.
• AWS Lambda is cheap. Up to a limit though. This impact not only your wallet.
• If you need geographic spread, AWS is lonely at the top.

The setup

Recently, I was doing a brainstorm at a startup here in Berlin on the future of their infrastructure. They were ready to move on from their initial, almost 100% Ec2 + Chef based setup. Everything was on the table. But we crossed out a lot quite quickly:

• Pure, uncut, self hosted Kubernetes — way too much complexity
• Managed Kubernetes in various flavors — still too much complexity
• Zeit — Maybe, but no Docker support
• Elastic Beanstalk — Maybe, bit old but does the job
• Heroku
• Lambda

It became clear a mix of PaaS and FaaS was the way to go. What a surprise! That is exactly what I use for Checkly! But when do you pick which model?

I chopped that question up into the following categories:

• Developer Experience / DX 🤓
• Ops Experience / OX 🐂 (?)
• Cost 💵
• Lock in 🔐

Read the full post linked below for all details

Need advice on what platform, systems and tools to use.

Evaluating whether to start a new digital business for which we will need to build a website that handles all traffic. Website only right now. May add smartphone apps later. No desktop app will ever be added. Website to serve various countries and languages. B2B and B2C type customers. Need to handle heavy traffic, be low cost, and scale well.

We are open to either build it on AWS or on Microsoft Azure.

Apologies if I'm leaving out some info. My first post. :) Thanks in advance!