MemVerge

MemVerge is a company focused on optimizing the cost and performance of AI and other data-intensive workloads. Its flagship product, Memory Machine X, intelligently manages the memory-storage hierarchy and fabric-attached memory. Memory Machine X includes features like a memory object store API called GISMO, which allows applications to create and access memory objects across multiple nodes using memory semantics. It also supports intelligent tiering, enabling the use of lower-cost, higher-capacity CXL memory, and accelerates shuffles with shared memory objects in the Ray AI Framework. The platform improves performance for MySQL databases, delivers higher QPS and lower latency with the AI-native Weaviate AI vector database, and makes Ray clusters IO-free by eliminating object serialization and transfers over the network for remote object access. Additionally, Memory Machine X allows multiple servers to access shared memory with software-enabled cache coherence, employs intelligent placement policies and memory page movement to optimize latency or bandwidth, and provides observability into server resource usage to help IT organizations understand the benefits of CXL memory. It can scale memory in a server to 32TB at half the cost without sacrificing application performance, and its latency tiering and bandwidth policy engine optimize performance based on memory page ‘temperature’ and available bandwidth, respectively. The GISMO feature reduces or eliminates data transfers over the network by allowing direct access to data in the shared memory pool and maintaining cache coherence between processors in different servers. The Server Memory Expansion software advises on the appropriateness of CXL memory for application environments and automatically places data in the optimal memory tier for performance, adapting to varying workloads with intelligent placement policies and memory page movement. MemVerge also offers Memory Machine Cloud, a powerful and intuitive container orchestration platform for running data-intensive pipelines and interactive computing applications in the cloud. Memory Machine Cloud significantly reduces BLAST Compute costs by 91% and boosts the performance of Nextflow workflows on AWS. Memory Machine X notably boosts NVIDIA GPU utilization by 77%.