# Configuring SingleStore for NUMA

Typically, multi-processor hosts provide for non-uniform memory access (A computer memory design used in multiprocessing. Through NUMA, a processor can access memory that’s considered “local memory” faster than it can access memory that is local to another processor, or memory that is shared between processors.), or NUMA. In a NUMA configuration, the hosts’s motherboard has many physical processors on the same physical board with multiple banks of memory, and memory is designated as being “local” to an assigned processor. From that processor’s standpoint, accessing this local memory is much faster than accessing memory that is considered local to a different processor (hence the "non-uniform"). For hosts that allow NUMA, SingleStore recommends running a node per physical processor in conjunction with CPU pinning. Refer to [Configure CPU Pinning with NUMA](https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux_openstack_platform/7/html/instances_and_images_guide/ch-cpu_pinning) for more information.

When running SingleStore on hosts with multiple NUMA nodes, you should configure SingleStore for NUMA with `numactl` for optimal performance. If you do not configure SingleStore this way, performance will be greatly degraded due to expensive cross-NUMA-node memory access. Configuring for NUMA should be done as part of the installation process; however, you can reconfigure your deployment later, if necessary.

To simplify the configuration process, the SingleStore management tools (`sdb-admin`, `memsqlctl`, etc.) can automatically detect if a host host has multiple NUMA nodes and then configure SingleStore with `numactl` to bind individual SingleStore nodes to NUMA nodes. To download and use the management tools, see the [deployment guides](https://docs.singlestore.com/db/v9.1/deploy.md).

## Configuring a new cluster for NUMA

1. On each host that will contain SingleStore leaf nodes, create one leaf per NUMA node by running `sdb-admin create-node` and then `sdb-admin add-leaf` until you have the same number of SingleStore leaf nodes as NUMA nodes. You will need to provide a different `--port` argument for each SingleStore node on the same host.

   For example, on a host with two NUMA nodes, you would deploy two leaf nodes.
   ```shell
   sdb-admin create-node --host <hostname|IP> --port 3307 --password <secure_pass>

   ```
   ```shell
   sdb-admin create-node --host <hostname|IP> --port 3308 --password <secure_pass>

   ```
   ```shell
   # Assign the nodes the role of "leaf" and add them to the cluster.
   sdb-admin list-nodes -q -r unknown | xargs -I % sdb-admin add-leaf --memsql-id %  -y

   ```

2. Run `sdb-admin optimize` and confirm the prompt.
   ```shell
   sdb-admin optimize

   ```
   This will configure `numactl` to bind SingleStore nodes to NUMA nodes, and configure SingleStore memory settings.

   See [sdb-admin optimize](https://docs.singlestore.com/db/v9.1/reference/singlestore-tools-reference/sdb-admin-commands/optimize.md) for more information.

3. Restart the cluster for the configuration changes to take effect.
   ```shell
   sdb-admin restart-node --all

   ```

## Reconfiguring an existing cluster for NUMA

1. Follow the steps in the previous section to deploy and optimize your leaf nodes.

2. Rebalance all SingleStore databases by running [REBALANCE ALL DATABASES](https://docs.singlestore.com/db/v9.1/reference/sql-reference/cluster-management-commands/rebalance-all-databases.md).

```sql
REBALANCE ALL DATABASES;

```

***

Modified at: October 31, 2023

Source: [/db/v9.1/reference/configuration-reference/cluster-configuration/configuring-singlestore-for-numa/](https://docs.singlestore.com/db/v9.1/reference/configuration-reference/cluster-configuration/configuring-singlestore-for-numa/)

(An index of the documentation is available at /llms.txt)