VMware for CPUs, MEXT for Memory

In 1998, VMware emerged onto the scene and revolutionized computing with its x86-compatible virtualization solution. By abstracting hardware from operating systems, VMware allowed multiple virtual machines (VMs) to run on commodity hardware, democratizing access to virtualization technology. General-compute systems—which previously were not taking full advantage of all of the CPUs and cores they were equipped with—were suddenly bearing witness to the tremendous efficiencies that came from improved CPU utilization.¹

VMware ESX and the Beginnings of Cloud Computing

By the early 2000s, VMware’s ESX server became a market leader, as more and more organizations recognized the potential of virtualization to consolidate servers and reduce costs. They found that VMware’s solution helped optimize CPU utilization, reduce power consumption, and simplify IT management.¹

A couple of years later, virtualization technology became the cornerstone of the modern cloud infrastructure. Cloud service providers leveraged it to offer scalable, on-demand compute resources. Fully virtualized data centers began to emerge, leading to the concept of “Software-Defined Data Centers (SDDC)”.² Today, what VMware pioneered is ubiquitous across enterprise IT environments. They took the CPU utilization challenge and solved it, across the board.

The Memory Utilization Challenge

Studies from major cloud providers have shown that memory utilization for internal and external workloads regularly drops to 50% or below.³This means that over half of the provisioned memory in DRAM isn’t actively being utilized for minutes at a time—an eternity from a microprocessor’s point of view. This “cold memory” is simply sitting in DRAM, and organizations end up consuming a much larger DRAM footprint that they truly need—resulting in potentially billions of dollars in completely wasted spend.

Simply put, while CPU utilization used to be the biggest challenge when it came to data center efficiency, memory utilization is now that crucial issue. This is where MEXT comes into play.

Enter MEXT

MEXT is tackling the memory utilization problem in an unprecedented way. Our patent-pending, software-only solution begins by determining the breakdown of hot vs. cold memory pages in DRAM. It then offloads the cold memory pages to a 20x lower-cost Flash device. The MEXT AI engine continually predicts which memory pages in Flash are likely to be requested by the application in the near future, and transparently pushes them back into DRAM—all before the application even requests them. As a result, the application stays performant because from its perspective, relevant memory pages are contained within DRAM.

This keeps application performance intact within a far smaller DRAM footprint, resulting in substantial cost efficiencies (up to 40% lower costs). Alternatively, customers can double their systems’ effective memory capacity (by leveraging MEXT AI + Flash as memory) while keeping their costs stable.