In-Memory Database Performance on the OpenMetal XXL v4

Q: How does the OpenMetal XXL v4 support large in-memory database workloads?

The OpenMetal XXL v4 supports large in-memory database workloads through its 2048GB DDR5 base configuration, 320MB total L3 cache, approximately 614 GB/s peak memory bandwidth, and 64 dedicated physical cores on single-tenant hardware.

Note: the XXL v4 is no longer in the OpenMetal catalog for new orders. The guidance below still applies to deployed units; for a new build, the XL v5 is the current equivalent.

See bare metal pricing

In-memory databases perform best when the working dataset fits entirely in RAM, eliminating disk reads from the query path. The XXL v4’s 2TB handles most large-scale OLTP and analytical datasets in-memory, and 2TB is the maximum published memory configuration on the v4 generation — all 32 DIMM slots are populated at that capacity.

Memory bandwidth matters as much as capacity for vectorized and analytical queries. DDR5-4800 across 16 channels (8 per socket) delivers approximately 614 GB/s theoretical peak — relevant for columnar databases like ClickHouse and analytical engines like Apache Ignite where sequential memory throughput is the binding constraint.

ECC memory is enabled across all 32 DIMM slots, a requirement for financial and healthcare data where silent memory errors carry compliance implications. The dedicated single-tenant model means the full 2TB is available to the workload without memory balloon drivers or hypervisor reservations common on cloud instances.