Memory Pooling

Requirements and Challenges

• Memory accounts for a large percentage of server costs in the data center
• Memory is stranded behind individual hosts, and leads to inefficient usage of available memory resources and higher costs
• Dynamic allocation of memory resources across multiple hosts from a memory pool can improve utilization to reduce TCO
• Heterogeneous, multi-CPU and CPU/GPU topologies can optimize performance and cost with memory pooling

Leo CXL Memory Connectivity Platform Benefits

• Leo increases CPU efficiency by enabling memory bandwidth expansion up to 89.6 GB/s and capacity expansion up to 2TB for cloud-server applications such as AI and ML
• Plug and play operation for both CXL 1.1 and CXL 2.0 based solutions
• Interoperates seamlessly with all major memory vendors
• Prevents memory stranding through effective utilization of memory resources across multiple-hosts
• Eliminates the need to over-provision memory to reduce overall TCO
• Provides customizable RAS, telemetry and fleet management needed for cloud-scale deployment
• Industry leading security features to ensure end-to-end data integrity and protection