Uber started using docker containers at scale in 2015, and has gone through a few generations of cluster management and service discovery technologies. In early 2019, Uber started working on migration from Mesos to Kubernetes to support secure service mesh and machine learning workloads.
This talk highlights the following:
- Overview of Uber Compute Infra
- API server benchmark and tweaks
- Custom controller and scheduler logic
- CRI: resource, health check, logging, isolation
- SPIRE and service discovery setup at Uber
eBay has been using Analytical DBMS (ADBMS) data warehouse solution for over a decade, there are millions of batch queries running every day against 6000+ key DW tables, which contains over 22PB data (compressed) and still keeps booming every year. Based upon that, data services and products enables eBay business decisions and site features, so it has to be always available and accurate.
Lipeng Zhu and Edward Zhang from Ebay discuss how eBay has been working on migrating ADBMS batch workload to Spark.
OLX produces about 50 millions messages daily to be delivered to 300+ millions users across the globe; via email, sms or push. The majority of these notifications relies on the processing of the billions of events generated by their web and mobile platforms to understand the users behaviour and to craft relevant messages designed to influence the customer journey positively.
In this presentation Emanuele Bardelli discusses the approach, challenges and learnings of migrating OLX's notification platform from a monolithic, batch system based on AWS Redshift, SQL and ETL pipelines to a micro-service, real-time system developed with Apache Spark and Python.
Learn how Jet’s Kafka clusters squeeze every drop of disk performance out of Azure, all completely transparent to Kafka.
-Striping cheap disks to maximize instance IOPS
-Block compression to reduce disk usage by ~80% (JSON data)
-Instance SSD as the secondary read cache (storing compressed data), eliminating >99% of disk reads and safe across host redeployments
-Upcoming features: Compressed blocks in memory, potentially quadrupling your page cache (RAM) for free
These are the areas covered;
-Adapting ZFS for cloud instances (gotchas)
-Performance tuning for Kafka
In this talk, Thanh Tran a Director of Data Science for Upwork presents their modernization efforts in moving towards a
1) holistic data processing infrastructure for batch and stream data processing using S3, Kinesis, Spark and Spark Structured Streaming
2) model development using Spark MLlib and other ML libraries for Spark
3) model serving using Databricks Model Scoring, Scoring over Structured Streams and microservices and
4) how they orchestrate and streamline all these processes using Apache Airflow and a CI/CD workflow customized to our Data Science product engineering needs.
Learn how eBay moved their ETL computation from conventional RDBMS environment over to Spark. This was a journey which led to an implementation of a 1000+ node Spark Cluster running 10,000+ ETL jobs daily, all done in a span of less than 6 months, by a team with limited Spark experience.