We explore learnings from first-generation Compute Express Link (CXL) memory. While broad adoption will be a long journey similar to that of RDMA, we show that CXL.mem expansion is viable on today’s hardware and discuss experience with datacenter deployment.
We cover practical guidance on systems engineering, monitoring, failure modes, security, and multi-tenant interference, and we outline a pragmatic adoption pathway: solidify robust expansion first, prototype micro-pooling next, and move to selective sharing as the ecosystem matures.
