L0->L0 compaction; RocksDB for the Cloud; Similarity Search; RocksDB in TiKV

Agenda:

6:30 - 7:00 Food, drinks

7:00 - 7:20 Andrew Kryczka - L0->L0 compaction in RocksDB

7:20 - 7:40 Dhruba Borthakur - RocksDB for the Cloud

7:40 - 8:00 Pete Hunt - Streaming Similarity Search for Fraud Detection

8:00 - 8:20 Siddon Tang - RocksDB in TiKV

8:20 - 9:00 Drinks, hangout

Big thanks to Greylock Partners for welcoming us in their space and Facebook for sponsoring food and drinks!

Andrew Kryczka - L0->L0 compaction in RocksDB

Abstract: Compaction is a fundamental process in RocksDB to reduce read- and space-amp. Recently, we modified our default compaction algorithm (level-based) to improve its performance for write-heavy workloads. This change significantly improved write latency jitter, a common production concern. We will talk about the feature’s design and its performance impact.

Bio: Andrew is a software engineer on Facebook’s database team. He has worked on RocksDB features related to compaction, statistics, range deletion, backup, and compression. Prior to RocksDB, he built Facebook’s configuration monitoring system for network devices.

Dhruba Borthakur - RocksDB for the Cloud

Abstract: RocksDB is used extensively by applications on the cloud. The stock RocksDB library does not provide for durability of data in the case of machine failures. This means that applications typically have to implement their own mechanisms for replicating data. On the other-hand, the AWS cloud environment provides services that allow elegant durability and replication of data. This talk describes how RocksDB-cloud can leverage these cloud-services to achieve data durability in the case of machine failures.

Bio: Dhruba is the CTO and Co-founder of Rockset, a stealth-mode startup. Prior to this, he was an engineer in the database team at Facebook where he was the founding engineer of the RocksDB datastore. Earlier at Yahoo, he was the founding engineer of Hadoop Distributed File System. He is a contributor to the open source Apache HBase project. Earlier, he held various roles at Veritas Software, founded an e-commerce startup Oreceipt.com (http://oreceipt.com/) and contributed to Andrew File System (AFS) at IBM-Transarc Labs. Longer version at: https://www.linkedin.com/in/dhruba

Pete Hunt - Streaming Similarity Search for Fraud Detection

Abstract: https://medium.com/smyte/streaming-similarity-search-for-fraud-detection-64d7476953d4

Bio: Pete is a co-founder of Smyte, a YC-backed startup fighting against online fraud. Previously at Facebook and Instagram, Pete is one of the early people that worked on React.

Siddon Tang - RocksDB in TiKV

Abstract: TiKV (pingcap/tikv) is a distributed, strong consistency, transactional key-value database, which uses RocksDB as underlying storage. This talk is about how we use, how we tune and how we test RocksDB in our project.
Agenda: 1. Why we choose RocksDB; 2. How we use and tune RocksDB; 3. How we contribute to RocksDB.

Bio: Siddon Tang, 10 years experience in distributed system, chief engineer of PingCAP, before joining PingCAP, he worked for wps.com (http://wps.com/) as the tech lead of backend infrastructure.