CoreWeave’s Local Object Transport Accelerator (LOTA) now runs on CPU Nodes in addition to GPU Nodes. CPU Nodes participate in the distributed LOTA cache alongside GPU Nodes, and LOTA is now available in clusters with any mix of Node types, including CPU-only clusters.Documentation Index
Fetch the complete documentation index at: https://docs.coreweave.com/llms.txt
Use this file to discover all available pages before exploring further.
Overview
Previously, LOTA was deployed exclusively on GPU Nodes. Workloads running on CPU Nodes in a mixed cluster could reach the LOTA endpoint (http://cwlota.com) but relied on routing through GPU Nodes to access the cache.
With this release:
- LOTA runs as a local proxy on every Node in a CKS cluster, regardless of Node type.
- CPU-only clusters can now use the LOTA endpoint for accelerated object storage access.
- Cache capacity scales with cluster size across all Node types. The amount each Node contributes to the distributed cache varies by instance type.
What’s unchanged
- The LOTA endpoint remains
http://cwlota.comfor workloads running inside a CoreWeave cluster. - The primary endpoint,
https://cwobject.com, continues to be the correct endpoint for access from outside a CoreWeave cluster. - All existing LOTA features (pre-stage cache, LRU eviction, cross-region reads, SSE-C support) work the same way on all Node types.