What are you looking for ?
Infinidat
Articles_top

Network Entropy Decomposing Memory

Memory advances reshaping storage and data center architectures

This report was written on September 30, 2021 by Ken Clipperton, lead analyst, storage, DCIG LLC.

 

 

 

The article title, Network Entropy is Decomposing Memory, may call up ghoulish images from a B-Horror movie, but it shouldn’t. It refers to a positive dynamic described by Rob Gingell, the former chief engineer at Sun Microsystems.

He said: “Over time networks have this property of wanting to decompose things into constituent parts, have those parts get specialized and then reintegrated.”

Dcig Network Entropy F1

This network entropy process has occurred many times. Some have described it as a pendulum swinging between centralized computing and decentralized computing. But Greg Papadopoulos, former CTO at Sun Microsystems, sees it differently. “I think it’s a spiral upwards and we come to a place and we look down and it looks familiar. You know, where you’ll say, oh I see, here’s a 3270 connected to a mainframe. Well, that looks like a browser connected to a web server.”

The Network Is the Computer
This current research into composable infrastructure has taken me back to the early days of my IT career. I managed DEC VAX minicomputers. At that time, the phrase The Network is the Computer described the vision of Sun Microcomputers. Today, it is a trademark of Cloudflare, Inc. Whoever owns the trademark, it was – and is – a powerful concept that continues to shapes computer architectures and IT careers.

Network Entropy Has Decomposed Servers, Storage, Compute and Networking
The most common terminology used for network entropy is virtualization. We have experienced server virtualization, storage virtualization, network virtualization, and even memory virtualization for decades now. Indeed, the initials of the OS used by the first computer systems I managed stood for Virtual Memory System.

More recently, the rise of GPUs, SmartNICs, ASICs, FPGAs, IPUs and other specialized processors demonstrates the network entropy effect on computing resources. When applied to networking, the industry appears to be coalescing around the term DPU. A DPU is a programmable processor with acceleration engines for moving data around the data center. DPUs handle various security, communication, and storage functions.

Network Entropy Is Decomposing Memory
Network entropy is decomposing memory. Though we are still in the early phases of a persistent-memory-enabled revolution in performance, cost, and capacity, multiple vendors are now shipping storage class memory in their enterprise servers and storage systems. The revolution has begun.

Memory is the last critical infrastructure component that needs to become virtualized and composable to better enable the data-intensive processing demands of modern applications,” said Bernie Wu, VP of business development, MemVerge.

Specialization – The Decades-Long Search for Next-Gen Memory
Key infrastructure providers such as Intel, IBM, Micron, and Samsung have long recognized that DRAM technology is approaching limits in both capacity and bandwidth. A significant aspect of that next-gen is that it is nonvolatile. It retains data in the absence of power.

These companies have for decades been exploring alternate memory technologies. These specialized technologies include MRAM, FRAM, ReRAM, and PCM. For a long time, these research efforts did not make it out of the lab and into availability. That has changed.

Storage Class Memory Is Here in Production Volumes
Today multiple providers are shipping what is popularly called storage class memory in production volumes. These memories come from multiple major chip foundries, including TSMC, Samsung, Global Foundries.

In terms of shipped capacity, the initial SCM leaders are 3D XPoint (Intel Optane) and MRAM. In 2021, MRAM shipments crossed the 100TB threshold, and 3D XPoint shipments are estimated at 1,000PB.

Reintegration – Memory Advances Are Reshaping Storage and Data Center Architectures
Making big memory broadly available to a wide range of standard workloads requires an ecosystem beyond the storage class memory itself. Standards, plugfests, and hackathons all play a role in the reintegration process.

Standards such as DDR5, PCIe 5.0, CXL, Gen-Z, and NVMe are essential enablers of the next-gen data center. All these technologies are now available for integration into systems.

Call It What You Will, This Is a Big Deal
These standards enable large amounts of DRAM and SCM to be disaggregated, pooled, and accessed via either CXL or even more widely via Gen-Z. This is new.

Dcig Network Entropy

 

The industry has coalesced around these standards, but not around what to call this new memory entity.

Candidates include:
• Big memory
• Fabric-attached memory
• Memory virtualization
• Shared far memory
• Software-defined memory

Whatever terminology we use, these technologies are about to reshape storage and data center architectures. For the first time, memory is becoming a networked resource. Within the data center, this gives new meaning to the phrase, The Network Is the Computer.

What to Do Now
Today, we are at the turning of the tide for storage and data center architectures. The time has come for enterprise technologists to look for high-value tactical wins. They should also begin planning infrastructure refreshes around a composable, data-centric, big memory architecture.

Resources:
The Network is the Computer
The Network is the Computer: A Conversation with Greg Papadopoulos
Experienced Business Development Leader Bernie Wu Joins MemVerge
Dynamic Trends in Nonvolatile Memory Technologies
Four Top Use Cases for Big Memory Today and Tomorrow

 

Articles_bottom
AIC
ATTO
OPEN-E