We have chosen Tibero over Oracle because we want to offer a PL/SQL-as-a-Service that the users can deploy in any Cloud without concerns from our website at some standard cost. With Oracle Database, developers would have to worry about what they implement and the related costs of each feature but the licensing model from Tibero is just 1 price and we have all features included, so we don't have to worry and developers using our SQLaaS neither. PostgreSQL would be open source. We have chosen Tibero over Oracle because we want to offer a PL/SQL that you can deploy in any Cloud without concerns. PostgreSQL would be the open source option but we need to offer an SQLaaS with encryption and more enterprise features in the background and best value option we have found, it was Tibero Database for PL/SQL-based applications.

We wanted a JSON datastore that could save the state of our bioinformatics visualizations without destructive normalization. As a leading NoSQL data storage technology, MongoDB has been a perfect fit for our needs. Plus it's open source, and has an enterprise SLA scale-out path, with support of hosted solutions like Atlas. Mongo has been an absolute champ. So much so that SQL and Oracle have begun shipping JSON column types as a new feature for their databases. And when Fast Healthcare Interoperability Resources (FHIR) announced support for JSON, we basically had our FHIR datalake technology.

In the field of bioinformatics, we regularly work with hierarchical and unstructured document data. Unstructured text data from PDFs, image data from radiographs, phylogenetic trees and cladograms, network graphs, streaming ECG data... none of it fits into a traditional SQL database particularly well. As such, we prefer to use document oriented databases.

MongoDB is probably the oldest component in our stack besides Javascript, having been in it for over 5 years. At the time, we were looking for a technology that could simply cache our data visualization state (stored in JSON) in a database as-is without any destructive normalization. MongoDB was the perfect tool; and has been exceeding expectations ever since.

Trivia fact: some of the earliest electronic medical records (EMRs) used a document oriented database called MUMPS as early as the 1960s, prior to the invention of SQL. MUMPS is still in use today in systems like Epic and VistA, and stores upwards of 40% of all medical records at hospitals. So, we saw MongoDB as something as a 21st century version of the MUMPS database.

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

The platform deals with time series data from sensors aggregated against things( event data that originates at periodic intervals). We use Cassandra as our distributed database to store time series data. Aggregated data insights from Cassandra is delivered as web API for consumption from other applications. Presto as a distributed sql querying engine, can provide a faster execution time provided the queries are tuned for proper distribution across the cluster. Another objective that we had was to combine Cassandra table data with other business data from RDBMS or other big data systems where presto through its connector architecture would have opened up a whole lot of options for us.