Tendis

Tendis 是一个与 Redis 协议完全兼容的高性能分布式存储系统。「Tendis is a high-performance distributed storage system fully compatible with the Redis protocol.」Tendis is a high-performance distributed storage system fully compatible with the Redis protocol.

Github星跟踪图

Tendis is a high-performance distributed storage system which is fully compatible with the Redis protocol. 中文文档

Table of Contents

Introduction

Tendis is a high-performance distributed storage system which is fully compatible with the Redis protocol. It uses RocksDB as the storage engine, and all data is stored to disks through RocksDB. Users can access Tendis using a Redis client, and the application hardly needs to be changed. In addition, Tendis supports storage capacity far exceeding memory, which can greatly reduce user storage costs.

Similar to Redis clusters, Tendis uses a decentralized distributed solution. The gossip protocol is used for communication between nodes, and all nodes in a cluster can be routed to the correct node when a user accesses. Cluster nodes support automatic discovery of other nodes, detect faulty nodes, and ensure the application is almost not affected when the master node failed.

Features

  • Redis compatibility

    Redis protocol and commands supported in Tendis are compatible with Redis.

  • Persistent storage

    Using RocksDB as storage engine. All data is stored in RocksDB in a specific format, supporting PB-level storage capacity.

  • Decentralized distributed cluster

    Distributed implementation like Redis clusters, using a gossip protocol to intercommunicate between nodes.

  • Horizontal scalability

    Data migration online between nodes. High performance and linear scalability up to 1,000 nodes.

  • Failover

    Auto-detect non-working nodes, and promote replica nodes to master when a failure occurs.

  • Key component for Tendis Hybrid Storage Edition

    Thanks to the design and internal optimization, Redis and Tendis can work together to be Hybrid Storage Edition. It is suitable for KV storage scenarios, as it balances performance and cost, and greatly reduces your business operating costs by 80% in the scenarios where cold data takes up a lot of storage space.

Get Started

Requirements

  • g++ (required by c++17, version >= 5.5)
  • cmake (version >= 3.13.0)

Build

$ git clone https://github.com/Tencent/tendis.git --recursive
$ git submodule update --init --recursive
$ mkdir bulid
$ cd build & cmake ..
$ make -j12

Run and Play

$ ./build/bin/tendisplus tendisplus.conf

Connect to the server via redis-cli

$ redis-cli -p 51002

TEST

You can make some change to the code and make sure the following test script pass.

$ sh ./testall.sh

Performance

Hardware spec

CPU:2.50 GHz,48 core
DISK:NVMe SSD
MEM:192GB

Benchmarks and QPS numbers

tendisplus: workers = 56

./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="set __key__ __data__" --key-prefix="kv_" --key-minimum=1 --key-maximum=500000000 --random-data --data-size=128 --test-time=1800
./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="get __key__" --key-prefix="kv_" --key-minimum=1 --key-maximum=500000000 --test-time=1800
./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="incr __key__" --key-prefix="int_" --key-minimum=1 --key-maximum=1000000 --test-time=1800
./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="lpush __key__ __data__" --key-prefix="list_" --key-minimum=1 --key-maximum=1000000 --random-data --data-size=128 --test-time=1800
./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="sadd __key__ __data__" --key-prefix="set_" --key-minimum=1 --key-maximum=1000000 --random-data --data-size=128 --test-time=1800
./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="zadd __key__ __key__ __data__" --key-prefix="" --key-minimum=1 --key-maximum=1000000 --random-data --data-size=128 --test-time=1800
./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="hset __key__ __data__ __data__" --key-prefix="hash_" --key-minimum=1 --key-maximum=1000000 --random-data --data-size=128 --test-time=1800

image.png

latency:

SET    1.6ms    99%
SET    2.2ms    99.9%
SET    6.9ms    99.99%
SET    9.4ms    100.00%

QPS on different payload

tendisplus: workers = 56

./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="set __key__ __data__" --key-prefix="kv_" --command-key-pattern=R --random-data --data-size=128 --test-time=1800
./memtier_benchmark -t 20 -c 50 -s 127.0.0.1 -p 51002 --distinct-client-seed --command="get __key__" --key-prefix="kv_" --command-key-pattern=R --test-time=1800

image.png

we test set for half an hour, and then test get key half an hour. because the data is not big enough, most of the data is in memory, so the get qps for different payload is nearly the same.

Contributing

For more information regarding contributing issues or pull requests, checkout CONTRIBUTING

Support

Checkout support for FAQs or join our discussion groups.

License

Tendis is licensed under the GNU General Public License Version 3.0. Copyright and license information can be found in the file LICENSE.txt.

主要指标

概览
名称与所有者Tencent/Tendis
主编程语言C++
编程语言CMake (语言数: 7)
平台Docker, Linux
许可证Other
所有者活动
创建于2020-12-16 03:12:23
推送于2024-06-24 11:57:21
最后一次提交2021-04-29 07:21:21
发布数52
最新版本名称2.7.0-rocksdb-v8.5.3 (发布于 2024-01-24 08:45:54)
第一版名称0.9.0-rocksdb-v5.13.4 (发布于 )
用户参与
星数2.9k
关注者数61
派生数323
提交数1.9k
已启用问题?
问题数266
打开的问题数130
拉请求数8
打开的拉请求数3
关闭的拉请求数7
项目设置
已启用Wiki?
已存档?
是复刻?
已锁定?
是镜像?
是私有?