At HPE Discover this year, the vendor discussions were about composable infrastructure, 25 Gbps networking, VSAN readiness, GPUs, and other new, transformative concepts. These concepts require some significant software and hardware changes. Within the Hewlett Packard Enterprise portfolio, this implies some decisions may need to be made with respect to blades.

HPE Blades & BladeSystem
Courtesy of Hewlett Packard Enterprise

So, let us look at these technologies one by one and see what we can do within the blade world—what choices we have.

Blades: GPUs for Desktops

These require the use of a PCIe extender blade for up to six M6 NVIDIA cards or one GRID K1 or K2 card. Beyond the price, the problem is that you burn a blade slot with the chassis extender but also lose access to all mezzanine cards. In our case, this would mean loss of access to our Fibre Channel array (via a QLogic mezzanine card).
The other option is to use either an NVIDIA M6 mezzanine card or an equivalent AMD device. However, these devices’ heat sinks block the second mezzanine card.
The only solution, therefore, is to combine FlexFabric or Virtual Connect with the GPUs, as the FlexFabric and Virtual Connect devices no longer use mezzanine cards. Yes, to use Virtual Connect, you also need to use a pair of top-of-rack Fibre Channel switches. For some. this could be a bigger investment.
For GPUs within HPE BladeSystems, there are a number of choices to be made, but it is possible without loss of functionality unless both mezzanine cards are in use for non–network related devices. Then, there are no choices.

VSAN

VMware VSAN also has an impact on how blades are used. To get capacity VSAN requires disk slots not just for the cache layer but for the storage layer as well. This implies a need for either more slots or larger disks within a single blade.
Option 1: For most blades, there is a limit of 2 SAS devices (3.84 TB SSD) plus a USB port and an SD Micro port (32 GB). In the Gen9 blades, there is also the ability to add another pair of Solid State PCIe M.2 adapters (dual 120 GB).
Option 2: The alternative is to use a storage blade that can host twelve disks. However, to do a proper VSAN cluster, you would need four such devices with matching storage.
Given the amount of storage available via Option 1, it is possible to build a VSAN-enabled blade with plenty of storage. (HPE did announce 7 and 15 TB SSDs for 3PAR systems at HPE Discover. It will not be long before they are available for other systems.) You can place the OS on SD Micro, USB, or both—perhaps use the Solid State PCIe M.2 adapters for a caching layer and then use the two primary slots for capacity. Choices are required in how you lay out storage, but they do exist. I believe a 1TB SSD and a large capacity drive  plus use of the 32GB microSD is sufficient. However, you loose per-blade redundancy but gain multi-blade redundancy.

Blades: 25 Gbps Networking

At HPE Discover, there was also an announcement for 25 Gbps, with up to four channels bound to make 100 Gbps networking. However, this was limited to the rack server and composable hardware solutions only. The most blades can do is up to four 10 Gbps links per controller. The backplace of an HPE BladeSystem is either 20 Gbps or 40 Gbps depending on chassis type. This is a limitation, but not one most will hit. Even so, 25 Gbps networking is not currently available for any BladeSystem.
The composable hardware has faster backplanes, and these adapters do exist. To get 25 Gbps, we need to either wait for HPE to develop a 25 Gbps FlexFabric switch or upgrade to the Synergy composable hardware platform.

Composable Hardware

The big announcement at HPE Discover addressed composable hardware, aka HPE Synergy. The idea is to create a blade chassis that has built within it an abstraction layer (aka API) that allows for the composition of a blade to be programmatically made based on operating system in use.  While HPE BladeSystem has some of this capability, Synergy takes things to a whole new level.
Unfortunately, to go to Synergy from BladeSystem requires switching not only chassis, but racks, blades, memory, networking, and more. About the only thing you can reuse is storage. However, this is where most of the advancements are being made. The composable blades are longer, allowing for more slots to support things like GPUs. They are about the same size as the HPE sled form factor.

Final Thoughts

HPE Discover announced and showed many new and interesting items. However, I see the future as being one of limited blade use. It will not happen immediately, but it will happen unless blades get more space within for newer hardware and technologies. Currently, blade sales seem to have slowed down, and this pushes other decisions. As a form factor, they are too limited, but within those limits they are extremely powerful and the basis for composable hardware.
Is it time to switch to bigger rack boxes or even HPE Synergy? For me, not yet. However, there will come a time, and there is a need to pay attention to the future and plan for possible changeover and at least the options required to move forward.