YDB system requirements and recommendations
This section provides recommendations for deploying YDB clusters that are relevant regardless of the chosen deployment method (Ansible, Kubernetes, or manual).
Hardware configuration
The fault-tolerance requirements determine the necessary number of servers and disks. For more information, see Topology.
- Processor (CPU)
A YDB server can only run on x86-64 processors with AVX2 instruction support: Intel Haswell (4th generation) and later, AMD EPYC and later.
The ARM architecture is currently not supported.
- RAM
We recommend using error-correcting code (ECC) memory to protect against hardware failures.
- Disk subsystem
A YDB server can run on servers with any disk type (HDD/SSD/NVMe). However, we recommend using SSD/NVMe disks for better performance.
For YDB to work efficiently, we recommend using physical (not virtual) disks larger than 800 GB as block devices.
The minimum disk size is 80 GB, otherwise the YDB node won't be able to use the device. Correct and uninterrupted operation with minimum-size disks is not guaranteed. We recommend using such disks exclusively for informational purposes.
Warning
Configurations with disks less than 800 GB or any types of storage system virtualization cannot be used for production services or system performance testing.
We don't recommend storing YDB data on disks shared with other processes (for example, the operating system).
YDB works with disk drives directly and does not use any filesystem to store data. Don't mount a file system or perform other operations with partitions used by YDB. Also, avoid sharing the YDB's block device with the operating system and different processes, which can lead to significant performance degradation.
Prefer to use physical local disk drives for YDB instead of virtual or network storage devices.
Remember that YDB uses some disk space for internal needs when planning disk capacity. For example, on a medium-sized cluster of 8 nodes, you can expect approximately 100 GB to be consumed for a static group on the whole cluster. On a large cluster with more than 1500 nodes, this will be about 200 GB. There are also 25.6 GB of logs on each Pdisk and a system area on each Pdisk. Its size depends on the size of the Pdisk but is no less than 0.2 GB.
Software configuration
A YDB server can be run on servers with a Linux operating system, kernel 4.19 and higher, and libc 2.30. For example, Ubuntu 20.04, Debian 11, Fedora 34, or newer releases. YDB uses the TCMalloc memory allocator. To make it efficient, enable Transparent Huge Pages and Memory overcommitment.
If the server has more than 32 CPU cores, to increase YDB performance, run each dynamic node in a separate taskset/cpuset of 10 to 32 cores. For example, in the case of 128 CPU cores a viable approach would be to run four 32-CPU dynamic nodes, each in a dedicated taskset.
MacOS and Windows operating systems are currently unsupported for running production YDB servers. However, running YDB in a Docker container on them is acceptable for development and functional testing.