In our everyday lives, we rely on technology quite a bit, so it’s shocking when we realize that even with technology advancements happening so regularly, we often hardly notice them. Because of this, it’s even more difficult to keep up with the industry’s best advancements. At PSSC Labs, we find it imperative that we can keep abreast of the ever-changing technology landscape to determine what is really worthwhile and what is just marketing hype.
2020 is going to be an interesting year for the virtualization, HPC, AI and data center server spaces. We’ve outlined five key advancements that will not only provide a step forward in the industry, but can actually deliver business value. This value is so often missed, but we believe it’s critical that companies adopt the best of the best technologies.
Mellanox® BlueField® 2
BlueField System on Chip (SoC) is one of the best solutions for a wide-range of reasons, ranging from performance to cybersecurity. BlueField® 2 is the product of a Mellanox® ConnectX® network adapter paired with an Arm processor. The Arm processor allows the encryption capabilities out of the server itself and onto the network adapter. Organizations can create truly secure data environments without having to pay for expensive hardware or software and instead, just expect that the data on their network will actually be secure. This has immediate relevance for any DOD facility, government agency, financial institution, etc. Expect to see more exciting advancements out of Mellanox® as their acquisition by NVIDIA® plays out.
It has been some time since AMD® was the talk of the industry, but their recent advancements have started to shift them into the spotlight. AMD® has developed the EPYC™ processor line up, which experts say will rival Intel® market dominance. These new processors radically improve processor core count, memory band-width, and bring to market PCI Express® 4.0 (more on this later). Imagine having 512 processor cores in each U of rack space! This changes the game for virtualization and HPC environments (ie. Design & Engineering, Weather Modeling, Life Science, and Artificial Intelligence), where total core count directly relates to program success. AMD® is realizing their investment in the 7 nm technology is finally paying off. Good for AMD®, good for the industry, maybe not so good for Intel®. There’s no doubt that we’ll soon see Intel® rushing to bring new processors to market that can compete head to head with the new AMD® EPYC™ chips.
PCI Express® 4.0
As mentioned above, one of the technologies AMD® brought to market with their new EPYC™ processors is PCI Express® 4.0. With the PCI Express® 4.0, we’re able to see a doubling of bandwidth, from 32 GB/sec total (16 GB/sec unidirectional) to 64 GB/sec total (32 GB/sec unidirectional). Most Host Bus Adapters may not be able to immediately take advantage of this larger pipe, but we don’t expect that to be the case for very long. Over time, more and more network adapters, controllers, etc. will be able to fully take advantage of PCI Express® 4.0. That’s not to say that right now there are no benefits. In fact, the immediate impact is significantly more lanes for PCI devices, such as NVMe devices. This will dramatically improve IO performance and increase the number of devices connected to the server. The products we plan to bring to market in the near future will be built to support up to 36 x NVMe devices, in an effort to help our clients capitalize on the increase in open lanes for data transfer. This will become our standard storage option as we start to see SSDs fade away only to be replaced by NVMe storage options instead.
Everything over Fabric
We foresee GPUoF and NVMEoF becoming most practical for the vast majority of organizations that are ingesting and storing large amounts of data. There are many HBAs available today from companies, including Mellanox®, that allow everything to be run over fabric. This is all part of the movement to offload work from the CPU out to the edge of the server itself. Organizations will reap the benefits of freeing up CPU cycles for other core processing needs while storage and GPU are decoupled from the CPU.
Intel® Optane DC Memory
Jumbo memory, high-speed storage, and memory resilience are all major drivers in determining which products that technology partners like Intel® work to develop. Data center usages are rapidly evolving, creating a need that can only be met with the unique properties of next generation Non Volatile Memory (NVM). As a result, Intel® has developed the Intel® Optane DC Memory. This product brings valuable enhancements through it’s flexibility. Storage is offered right at the memory bus, avoiding traditional slower data paths for storage devices and allowing for maximum speed of processing. High Frequency Trading is one very obvious application but pretty much any real time database could benefit. Intel® Optane DC persistent memory modules come in 128GB, 256GB and 512GB capacities, vastly larger than DRAM sticks, which typically range from 4GB to 32GB.
Our new product line will be coming soon. We are looking forward to utilizing these technologies and others as they come to market in an effort to help government, military, higher education and commercial organizations achieve their business and organizational goals. Keep an eye out for our new releases coming soon!