MemVerge, a global provider of Big Memory Software, has made a significant announcement at the International Supercomputing Conference (ISC) with the introduction of Project Gismo (Global IO-free Shared Memory Objects). This “innovative” technology intends to revolutionize the efficacy of distributed applications by removing network I/O and data copies.
Project Gismo introduces a CXL-based multi-server shared memory architecture, advancing data access and collaboration in distributed environments. By incorporating Compute Express Link (CXL) technology seamlessly, Project Gismo would enable real-time data sharing across multiple servers, thereby eradicating the need for network I/O and decreasing data transfer latencies.
Charles Fan, CEO and co-founder of MemVerge, emphasized the significance of the solution by stating, “Network and storage IO are impediments for distributed, data-intensive applications. Project Gismo utilizes the emerging CXL platform and will effectively eliminate the IO Wall impediment. We foresee a memory-centric future in which memory is both the network and the storage.”
This innovative solution would enable businesses to realize the full potential of their distributed applications by accelerating data-intensive workloads and reducing latency to unheard-of levels. With Project Gismo, businesses can accomplish substantial performance improvements while streamlining their infrastructure and optimizing resource utilization.
Database Systems
AI/ML applications such as large language models, next-generation databases, and financial trading platforms are the primary use cases for Project Gismo. Multiple industry participants, including Timeplus, a developer of a next-generation real-time streaming database, have expressed interest in the solution. Timeplus has used Gismo to enhance the fault-tolerance of their database system, obtaining an extraordinary twentyfold increase in fail-over performance.
The CEO of Timeplus, Ting Wang, shared the following: “With Gismo’s revolutionary CXL-based multi-server shared memory architecture, our database system’s fault-tolerance has significantly increased. The pace of streaming query failover has been increased by a remarkable 20X, allowing us to offer an unmatched user experience for the continuance of data streaming processing.”
Another early adopter of Project Gismo is MatrixOrigin, a developer of a next-generation Hybrid Transactional/Analytical Processing Engine. Gismo substantially streamlines their caching architecture and enhances the efficacy of both transaction processing and analytical processing in the cloud, according to their findings.
MatrixOrigin’s Chief Technology Officer (CTO), Feng Tian, conveyed satisfaction, stating, “Project Gismo has greatly simplified our caching architecture, eradicating the need for complex data copies and reducing data skew. Multiple servers’ access to the global shared memory has enabled seamless data sharing and collaboration, enhancing the overall performance of our engine. As a consequence, we have experienced a remarkable performance increase, allowing us to manage larger workloads and provide clients with quicker query responses.”
MemVerge claims to have a significant impact on the distributed application landscape with the introduction of Project Gismo, which eliminates IO barriers and enables real-time data sharing between servers. This innovation would bear promise for a vast array of industries, empowering them to realize the maximum performance and efficiency potential of their data-intensive duties.
