MemVerge is a company that is delivering the world’s first Memory-Converged Infrastructure system, built with proprietary software technology. Pulse 2.0 interviewed MemVerge co-founder and CEO Charles Fan to learn more about the company.
Charles Fan’s Background
Fan has been working in the data center infrastructure space for over 25 years. Fan said:
“I co-founded Rainfinity with my PhD advisor Shuki Bruck, while I was a PhD student at Caltech. At Rainfinity, we created a file virtualization product called RainStorage. Rainfinity was subsequently acquired by EMC. I joined VMware in 2009 and started the VMware storage business unit, and led the development of VSAN, a leading software-defined storage product that became a billion-dollar product. The most recent company that I co-founded with Shuki, MemVerge, is a leading provider of what we call Big Memory Computing, which is designed to accelerate and optimize data-intensive applications.”
Formation Of MemVerge
How did the idea for MemVerge come together? Fan shared:
“For decades, computing had been bottlenecked by the ‘memory wall,’ where the capacity and the bandwidth of memory are significant performance bottlenecks of applications.”
“In 2017, as new memory hardware technologies started to emerge, we decided to start MemVerge. We knew that with every new hardware substrate, a new software stack must be developed to allow the applications to take full advantage of the new hardware. We developed a Big Memory Computing stack that automatically pools and manages tiers of memory.”
“Our decision was validated in 2019 when the industry announced the first open CXL specifications for composable memory. The latest megatrend of Generative AI and Large Language Models further fuel the growth of data-intensive workloads. We believe these events will make Big Memory Computing into a multi-billion-dollar market.”
Core Products
What are MemVerge’s core products and features? Fan explained:
“Generative AI has taken the world by storm, creating a boom for GPUs, High Bandwidth Memory (HBM) and Compute Express Link (CXL) memory. At this pivotal moment, MemVerge has built a Big Memory Computing Platform that continuously rightsizes computing and memory resources for AI workloads. The platform consists of software suites that optimize performance, cost and availability of applications running in the public cloud and hybrid cloud.”
“For the public cloud, Memory Machine Cloud is a new cloud automation platform that intelligently manages workloads on the public cloud. By dynamically adapting cloud resources to real-time application needs and enabling all applications to run reliably on spot instances, Memory Machine Cloud has delivered an order-of-magnitude higher resource efficiency for users such as Translational Genomics Research Institute (TGen), European Bioinformatics Institute (EBI), and Seekgene. The software also features real-time tracking of cloud resources and carbon footprint needed for the future of sustainable IT.”
“The advent of CXL is a giant step towards removing the fundamental “memory wall” that is bottlenecking AI applications today. CXL technology gives AI apps desperately needed memory bandwidth and capacity while driving down costs with composable memory that can be shared. Partnering closely with CXL leaders such as Samsung, SK hynix and Micron, award-winning MemVerge software for CXL enables applications to access pools of memory without modifying applications while providing optimal performance and cost. MemVerge software for CXL also offers data services, such as an application checkpointing service, ensuring uninterrupted application availability, with full integration into Kubernetes.”
Favorite Memory
What has been your favorite memory working for MemVerge so far? Fan reflected:
“We shipped our new product, Memory Machine Cloud, at the beginning of this year, and quickly customers started to adopt it to save cloud costs. A couple of months later, a blog of a founder of a biotech startup circulated over the internet, saying how a product demo fundamentally changed his outlook of the industry for the better. We then found out that product demo was actually done by us. It was a pleasant surprise to see the positive impact our product is making on our customers.”
Challenges Faced
What are some of the challenges you face in building the company? Fan acknowledged:
“We started our company in 2017 because new memory hardware started to emerge. The first such new hardware was Intel Optane Persistent Memory, and we formed a fantastic partnership with Intel to support Optane customers with our software. It became the core of our business.”
“Unfortunately in July of 2022, Intel announced the plan to end-of-life Optane products for its own strategic considerations. It essentially wiped out almost all of our revenue stream. Thanks to the determination and hard work by the MemVerge team, the company successfully pivoted over the following 12 months, applying the Big Memory Computing technology to the cloud and to CXL. We shipped the Memory Machine Cloud product at the beginning of 2023 and are looking to ship our CXL product by the end of 2023, and we are on our way to rebuilding our business, even stronger than before.”
Evolution Of MemVerge’s Technology
How has MemVerge’s technology evolved since launching? Fan noted:
“When we first launched MemVerge, our focus was on pooling and provisioning tiers of DRAM and Optane Persistent Memory in an Intel-centric ecosystem. Our focus has now expanded to pooling and sharing tiers of DRAM and CXL memory in a much broader partner ecosystem. We have also applied the technology to the cloud.”
Significant Milestones
What have been some of MemVerge’s most significant milestones? Fan cited:
“In July we unveiled a new version of Memory Machine Cloud for sustainability, as well as a 3-Step blueprint for slashing carbon footprint. We also introduced the latest version of Memory Machine Cloud Essentials, which is free forever, and includes capabilities that are ‘essential’ to getting started with cloud automation and optimization.”
“In August, at Flash Memory Summit, we demonstrated ‘Endless Memory’ along with Samsung, H3 Platform and XConn Technologies. This 2TB pooled CXL memory system allocates memory on-demand to replace swap-to-storage. The solution proved that organizations could eliminate dreaded Out-of-Memory errors and accelerate AI/ML workloads.”
Customer Success Stories
After asking Fan about customer success stories, he highlighted:
“Tgen – ‘I was getting up to 80% batch failure rates with Spot EC2. Now with SpotSurfer we have already brought failure rates due to spot reclaims to below 1%, and we are just getting started.’
Honsun – WaveRider provides the perfect balance between performance and cost, with a cost reduction of 93% when compared to the large, 32vCPU, 32GB RAM, OnDemand option, but with nearly the same performance of 12h 23m execution time vs 11 h 44m. For organizations running 1,000 BLAST applications per day in the cloud, using Memory Machine with WaveRider would result in a savings of $16,065 per day, or $4,173,000 per year.
Total Addressable Market
What total addressable market (TAM) size is MemVerge pursuing? Fan assessed:
“The pooled memory market is expected to grow to $20 billion in the next decade.”
Differentiation From The Competition
What differentiates MemVerge from its competition? Fan affirmed:
“MemVerge is pioneering software for Big Memory Computing. While a few competitive offerings exist that provide pieces of the stack, MemVerge is the only Big Memory software provider with a complete portfolio for hybrid cloud and public cloud.”
Future Company Goals
What are some of MemVerge’s future company goals? Fan concluded:
“We want to be there on day 1 when system vendors start shipping CXL servers and pooling systems. CXL hardware products, which were introduced in 2022, are now beginning to be integrated into systems in 2023. MemVerge is the software that is powering these systems. We are currently identifying the ‘killer apps’ for CXL and look forward to working with CXL hardware providers to drive those killer CXL apps in 2024.”
“The cloud will continue to be our other primary focus. Our Memory Machine Cloud product is finding new ways to help customers reduce their skyrocketing cloud cost.”
