Skip to content

High performance Distributed Table Cache system designed by JD.com, Offering hotspot data cache from databases in order to reduce pressure of database and improve QPS.

License

Notifications You must be signed in to change notification settings

jd-opensource/dtc

Repository files navigation

DTC

DTC - Distributed Table Cache

ubuntu&gcc-4.9

Overview

DTC is a high performance Distributed Table Cache system designed by JD.com that offering hotspot data cache for databases in order to reduce pressure of database and improve QPS.

The DTC system consists of the following components:

  • Agent - Provides key consistent hash routing in order to reduce connections and improve performance.
  • Dtcd - Provides hot data caching service.
  • Connector - Provides connection and communication between cache and persistent storage database such as MYSQL.

Feature

  • Database Protection
    • protection for null node, prevent cache breakdown.
    • provide long-term data caching, and prevent cache penetration.
    • data source thread available, protect the database with a limited number of connections.
    • Estimated timeout policy to reduce invalid database requests.
  • Data consistency
    • write-through policy, ensure cache and database data consistent.
    • barrier policy to prevent update requests lost while concurrcy.
  • Performance
    • integrated memroy allocation policy to avoid frequent system calls.
    • I/O multiplexing to handle concurrcy requests.
    • multiple data structure models to improve memory performance.
  • Scalability
    • cache node expands horizontally to enhance cache capacity.
    • cache node expands vertically, supports slave reading, and solve the bottleneck of hot keys.
    • provide sharding, supports for persistent storage scalable.

Performance

  • DTC can process 90,000 QPS of query requests at single-core cpu & single dtc instance.
  • DTC can provide above 3,000,000 QPS query capability with above 99.9% hit rate and less than 200 μs response time in actual distributed scenarios.
  • Layered Storage is able to provide about 1,000 QPS write capability with above 99.9% per single instance.

How to Build

DTC provides docker images for quick start:

  • Start server docker:
docker pull dtc8/server:latest
docker run -i -t --name dtc-server -p 127.0.0.1:20015:20015 dtc8/server:latest

Depending on 3rd-party sql parsing engine hsql . For more compile information, click Building.
Trying a demo, visit QuickStart.

License

JD.com © Copyright 2021-2024 JD.com, Inc.
Apache 2.0. Visit LICENSE for more details.

About

High performance Distributed Table Cache system designed by JD.com, Offering hotspot data cache from databases in order to reduce pressure of database and improve QPS.

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages