Coauthors: Tao Feng, John Nicol, Chen Li, Peinan Chen, Hari Ramachandra LinkedIn has built hundreds of application services, with thousands of instances running in data centers. Optimizing the performance of these services can dramatically improve user experience and reduce operational costs, and profilers are commonly used to help achieve this. LinkedIn’s...
Posts by Tao Feng
-
- Topics:
- optimization
-
Update Apr 13, 2016: There are numerous improvement to Samza cachestore (SAMZA-658, SAMZA-812, SAMZA-873 etc.) since our last test runs. Those fixes improve two cases' (rocksdb only and rocksdb with changelog) performance dramatically. We reran those two cases and are reposting the numbers here. Per node throughput is about 1.1 million messages / sec with...
- Topics:
- Stream Processing,
- Apache Samza,
- Performance,
- performance analysis,
- Kafka,
- Data
-
Apache Samza has been run in production and is used by many LinkedIn services to solve a variety of stream processing scenarios. For example, we use it for application and system monitoring, or to track user behavior for improving feed relevance. The objective of this study was to measure Samza's performance in terms of the message-processing rate for a single...
- Topics:
- Performance,
- Apache Samza,
- Stream Processing,
- performance analysis,
- Kafka,
- Data
-
Update Apr 13, 2016: There are numerous improvement to Samza cachestore (SAMZA-658, SAMZA-812, SAMZA-873 etc.) since our last test...
- Topics:
- Performance,
- Apache Samza,
- Stream Processing,
- performance analysis,
- Kafka,
- Data