Software and Hardware Recommendations

As an open-source distributed SQL database with high performance, TiDB can be deployed in the Intel architecture server, ARM architecture server, and major virtualization environments and runs well. TiDB supports most of the major hardware networks and Linux operating systems.

OS and platform requirements

In v7.1 LTS, TiDB provides multi-level support for different quality standards on the combination of operating systems and CPU architectures.

• For the following combinations of operating systems and CPU architectures, TiDB provides enterprise-level production quality, and the product features have been comprehensively and systematically verified:

  Operating systems	Supported CPU architectures
  Red Hat Enterprise Linux 8.4 or a later 8.x version	x86_64 ARM 64
  Red Hat Enterprise Linux 7.3 or a later 7.x version CentOS 7.3 or a later 7.x version	x86_64 ARM 64
  Amazon Linux 2	x86_64 ARM 64
  Kylin Euler V10 SP1/SP2	x86_64 ARM 64
  UnionTech OS (UOS) V20	x86_64 ARM 64
  openEuler 22.03 LTS SP1/SP3	x86_64 ARM 64

  

  Note

  According to CentOS Linux EOL, the upstream support for CentOS Linux 8 ended on December 31, 2021. CentOS Stream 8 continues to be supported by the CentOS organization.

• For the following combinations of operating systems and CPU architectures, you can compile, build, and deploy TiDB. In addition, you can also use the basic features of OLTP, OLAP, and the data tools. However, TiDB does not guarantee enterprise-level production quality:

  Operating systems	Supported CPU architectures
  macOS 12 (Monterey) or later	x86_64 ARM 64
  Oracle Enterprise Linux 7.3 or a later 7.x version	x86_64
  Ubuntu LTS 18.04 or later	x86_64
  CentOS 8 Stream	x86_64 ARM 64
  Debian 9 (Stretch) or later	x86_64
  Fedora 35 or later	x86_64
  openSUSE Leap later than v15.3 (not including Tumbleweed)	x86_64
  SUSE Linux Enterprise Server 15	x86_64

  

  Note

  • For Oracle Enterprise Linux, TiDB supports the Red Hat Compatible Kernel (RHCK) and does not support the Unbreakable Enterprise Kernel provided by Oracle Enterprise Linux.
  • Support for Ubuntu 16.04 will be removed in future versions of TiDB. Upgrading to Ubuntu 18.04 or later is strongly recommended.

• If you are using the 32-bit version of an operating system listed in the preceding two tables, TiDB is not guaranteed to be compilable, buildable or deployable on the 32-bit operating system and the corresponding CPU architecture, or TiDB does not actively adapt to the 32-bit operating system.

• Other operating system versions not mentioned above might work but are not officially supported.

Libraries required for compiling and running TiDB

Library required for running TiDB: glibc (2.28-151.el8 version)

Docker image dependencies

The following CPU architectures are supported:

• x86_64. Starting from TiDB v6.6.0, the x86-64-v2 instruction set is required.
• ARM 64

Software recommendations

Control machine

Note

It is required that you deploy TiUP on the control machine to operate and manage TiDB clusters.

Target machines

Server recommendations

You can deploy and run TiDB on the 64-bit generic hardware server platform in the Intel x86-64 architecture or on the hardware server platform in the ARM architecture. The requirements and recommendations about server hardware configuration (ignoring the resources occupied by the operating system itself) for development, test, and production environments are as follows:

Development and test environments

Note

• In the test environment, the TiDB and PD instances can be deployed on the same server.
• For performance-related test, do not use low-performance storage and network hardware configuration, in order to guarantee the correctness of the test result.
• For the TiKV server, it is recommended to use NVMe SSDs to ensure faster reads and writes.
• If you only want to test and verify the features, follow Quick Start Guide for TiDB to deploy TiDB on a single machine.
• The TiDB server uses the disk to store server logs, so there are no special requirements for the disk type and capacity in the test environment.
• Starting from v6.3.0, to deploy TiFlash under the Linux AMD64 architecture, the CPU must support the AVX2 instruction set. Ensure that cat /proc/cpuinfo | grep avx2 has output. To deploy TiFlash under the Linux ARM64 architecture, the CPU must support the ARMv8 instruction set architecture. Ensure that cat /proc/cpuinfo | grep 'crc32' | grep 'asimd' has output. By using the instruction set extensions, TiFlash’s vectorization engine can deliver better performance.

Production environment

Note

• In the production environment, the TiDB and PD instances can be deployed on the same server. If you have a higher requirement for performance and reliability, try to deploy them separately.
• It is strongly recommended to configure TiDB, TiKV, and TiFlash with at least 8 CPU cores each in the production environment. To get better performance, a higher configuration is recommended.
• It is recommended to keep the size of TiKV hard disk within 4 TB if you are using PCIe SSDs or within 1.5 TB if you are using regular SSDs.

Before you deploy TiFlash, note the following items:

• TiFlash can be deployed on multiple disks.
• It is recommended to use a high-performance SSD as the first disk of the TiFlash data directory to buffer the real-time replication of TiKV data. The performance of this disk should not be lower than that of TiKV, such as PCIe SSD. The disk capacity should be no less than 10% of the total capacity; otherwise, it might become the bottleneck of this node. You can deploy ordinary SSDs for other disks, but note that a better PCIe SSD brings better performance.
• It is recommended to deploy TiFlash on different nodes from TiKV. If you must deploy TiFlash and TiKV on the same node, increase the number of CPU cores and memory, and try to deploy TiFlash and TiKV on different disks to avoid interfering each other.
• The total capacity of the TiFlash disks is calculated in this way: the data volume of the entire TiKV cluster to be replicated / the number of TiKV replicas * the number of TiFlash replicas. For example, if the overall planned capacity of TiKV is 1 TB, the number of TiKV replicas is 3, and the number of TiFlash replicas is 2, then the recommended total capacity of TiFlash is 1024 GB / 3 * 2. You can replicate only the data of some tables. In such case, determine the TiFlash capacity according to the data volume of the tables to be replicated.

Before you deploy TiCDC, note that it is recommended to deploy TiCDC on PCIe SSD disks larger than 500 GB.

Network requirements

As an open-source distributed SQL database, TiDB requires the following network port configuration to run. Based on the TiDB deployment in actual environments, the administrator can open relevant ports in the network side and host side.

Disk space requirements

Web browser requirements

TiDB relies on Grafana to provide visualization of database metrics. A recent version of Internet Explorer, Chrome or Firefox with Javascript enabled is sufficient.

Hardware and software requirements for TiFlash disaggregated storage and compute architecture

The preceding TiFlash software and hardware requirements are for the coupled storage and compute architecture. Starting from v7.0.0, TiFlash supports the disaggregated storage and compute architecture. In this architecture, TiFlash is divided into two types of nodes: the Write Node and the Compute Node. The requirements for these nodes are as follows:

• Software: remain the same as the coupled storage and compute architecture, see OS and platform requirements.
• Network port: remain the same as the coupled storage and compute architecture, see Network.
• Disk space:
  • TiFlash Write Node: it is recommended to configure at least 200 GB of disk space, which is used as a local buffer when adding TiFlash replicas and migrating Region replicas before uploading data to Amazon S3. In addition, an object storage compatible with Amazon S3 is required.
  • TiFlash Compute Node: it is recommended to configure at least 100 GB of disk space, which is mainly used to cache the data read from the Write Node to improve performance. The cache of the Compute Node might be fully used, which is normal.
• CPU and memory requirements are described in the following sections.

Development and test environments

Production environment

Note

You can use deployment tools such as TiUP to quickly scale in or out the TiFlash Compute Node, within the range of [0, +inf].