Dell’s VRTX is one of those devices that is just sexy, as IT hardware goes. It strikes a chord and drives IT professionals nearly wild. It looks cool, it has an incredible amount of power, it can be rack mounted or placed under a desk, it is quiet – so quiet that it can be run right in the middle of an open office space. It’s just really cool, and nearly every IT professional wants one – even if they have no idea why.
The problem with the VRTX is that it is generally misunderstood and the misunderstandings around the device itself and the architecture used within it have led to a lot of proposals, nearly continuous, to use the device where it is least suited. The device itself truly is awesome and has excellent use cases, but it is very important to understand what they are and what they are not as this is a very specialty piece of hardware.
First, we need to determine what the VRTX “is”. The Dell VRTX is primarily a blade enclosure, more or less like any blade system. But unlike traditional blade enclosure that typically hold six to ten blades per enclosure, the VRTX only holds four. So it is a “baby” blade enclosure. Because it is a true blade system, the Dell VRTX carriers the normal caveats of any blade enclosure. However, due to its small size the probability of it being able to be used, and retired, effectively make it quite a bit more reasonable to consider than traditional, larger blade enclosures. So an understanding of its blade nature is important in evaluating it for your organization’s needs.
Along with the included blade component, the VRTX also has a DAS (Direct Attached Storage) system attached via SAS to the blades. This storage array offers either twelve large form factor (3.5”) or twenty five small form factor (2.5”) hard drives attached by way of either one or two PERC8 hardware RAID controllers. This included, large scale, shared external storage array inside of the VRTX blade enclosure is what makes the VRTX unit truly unique.
So all four blades share the single DAS unit for storage. The four blades constitute 2U of the VRTX enclosure and the DAS unit another 2U for a total enclosure size of 4U.
Of course, as with any blade system, there is no requirement that you fully populate the VRTX initially, or ever. The system can be used with any number of blades from one to four, as needed. But the value of a blade enclosure, especially a small one such as this, depends heavily on being completely populated or nearly so, to be cost viable.
Architecturally what the VRTX represents is a highly compact, single chassis, Inverted Pyramid of Doom (the “traditional” 3-2-1 architectural design) built following what are, more or less, the best approaches for that type of system. The biggest advantages here are that the use of a solid DAS is mandated and cannot be altered and all connections between the DAS and the compute nodes are hard wired internally for the highest level of potential reliability for a shared external storage system with the least opportunity for human failure. By using DAS instead of SAN in this example, our 3-2-1 has its “2” layer removed resulting in a far better inverted pyramid structure. What we are left with is a 4-1 inverted pyramid design.
The overall profile of the VRTX is one of massive compute capabilities, far outstripping the computational needs of a normal SMB business, all in a single chassis. The smallest blade option is a dual processor module and the biggest are quad processor meaning that when populated we have a minimum of eight Intel Xeon processors over four nodes and a maximum of sixteen Intel Xeon processors over four nodes. This is truly a mammoth computational system in a small package. But it is critical to understand that all of this horsepower shares a single storage array and is not highly available and cannot be made so. This is a system designed for processing power, not as a reliable infrastructure component.
It should also be noted that Dell experienced reliability issues with the redundant PERC8 hardware RAID controller setup and had to pull it from the market for some time. As with nearly all storage systems in this category, which includes many DAS and SAN devices, redundant controllers are commonly the cause of storage outages rather than the preventers of such. Redundancy of RAID controllers is rarely a valuable addition and should never be looked on as a panacea to storage reliability concerns.
Given that fact that the VRTX is compute heavy and reliability weak, what are its designated use cases? Where does it make the most sense to consider deploying this unit?
There are three extremely common deployment scenarios today where large compute and shared “fragile” storage often fit. Of course there may be many special cases and those should be evaluated individually based on the power, cost and reliability profiles of the VRTX relative to other options. But by and large the big three use cases where we would want to see the VRTX deployed would be:
Enterprise Remote Office and Branch Office (ROBO): This use case is based around the concept of the VRTX being a single device, easily deployable with nothing to do but to “plug it in” delivering a “reliable enough” but very powerful platform for remote offices. Not every remote or branch office would need the kind of horsepower than a VRTX can provide and some would require high availability which it does not have, but large ROBOs are often ideally suited to this architectural profile due to the ease of remote management and the common ability to use remote access to a central office or datacenter as a means of providing failover and reliability in the event of a major disaster either to IT itself (such as a total failure of the VRTX) or to the ROBO itself (fire, flood, etc.)
A VRTX in this scenario can easily be the sole IT device, outside of networking equipment, powering an entire ROBO of hundreds or potentially even thousands of users. And the ability to do nearly all maintenance in a non-disruptive way, which if properly designed is trivial to provide with a VRTX, can be quite significant to a ROBO.
The concept of this being solely for the “enterprise” ROBO rather than SMB ROBOs is simply because of the total scale of the VRTX being larger than the typical needs of an SMB as a whole let alone the needs of just one remote office. The VRTX is just too “big” for the typical needs of an SMB without being specifically focused on the needs of SMB.
Virtual Desktop Infrastructure (VDI): VDI generally requires a large amount of compute power, non-disruptive updates and shared storage which is perfect for the VRTX. Of course this only makes sense in shops that need at least three nodes, if not four nodes, of compute power to leverage the blade chassis natural of the VRTX. But for companies looking for eight to sixteen CPUs worth of VDI power the VRTX can be a slam dunk. Possibly no use case is more appropriate for the VRTX than as a single, modular VDI system.
Big Data: Not many SMBs look to do big data processing today (Hadoop, Apache Spark, etc.) but a VRTX can be an ideal platform for doing huge processing in a small business that does not need to scale its data processing beyond this point. For larger enterprises needing a much larger scale of processing the VRTX would not be well suited, what makes it exceptionally valuable is in matching the size to the organization’s need. Of course other kinds of computationally heavy processing, such as Monte Carlo simulations, would also work well on this platform.
Now that we know where the VRTX is well suited, where does it not fit well?
The VRTX is very poorly suited to general computing use, in both the SMB and the enterprise sectors. In the enterprise the VRTX represents a fully contained, but non-scaling, stack which would be unwieldy and expensive in a large infrastructure.
In the SMB the VRTX is dramatic overkill on the computational size while underkill, generally in reliability, on the storage side. Most SMBs, when scaling past a single computation node, are seeking both flexible scalability as well as higher than typical reliability. Often it is a desire for high availability alone that drives SMBs past a single computation node considering the incredible capacity of a single node that is available today. So moving to an inverted pyramid architecture would be counter-productive to the needs of the typical SMB. The VRTX is simply too big, too rigid and lacks the reliability profile desired by SMBs. The SMB is really the last market where I would expect the VRTX to be deployed as general computing needs that drive SMB needs simply is the farthest appropriate use case for this device.
The VRTX is an amazing piece of equipment and well designed for several niche use cases, but is not designed to replace or be used in typical scenarios where standard servers, such as the Dell PowerEdge R730, have been designed to be the ideal equipment. General use equipment exists as the industry standards and best sellers for a reason, niche equipment also exists for a reason. Be sure to understand why the equipment you are considering makes sense for your environment, new and interesting is not enough to justify moving to special case gear.