Deployment Scenarios for the Dell PowerEdge FX
The Dell PowerEdge FX enclosure might, at first, be a bit of a mystery. For all intents and purposes, this is a 2U rack mount blade chassis and, in many ways, it is the baby sibling of the 5U VRTX enclosure. But at its core, while the general theory of the two units are related, the FX has a different mission entirely and some interesting use cases.
At 2U the Dell FX enclosure is targeted at high density installations of all sizes. It can accommodate as few as two physical servers within its chassis or as many as sixteen micro-blades. That is a lot of flexibility. In an enterprise these might be used to squeeze very high density deployments into a datacenter, but a far more interesting use case is combining these units with industry standard colocation offerings.
Because so many colocation facilities will accept single 2U server installations, the Dell FX offers the potential to build and deploy an entire corporate infrastructure from just two standard rack units! Of course these are not big storage devices, so if that is the intended use case the VRTX is going to fit the bill far better. But for a large range of compute resources in a small space the FX is hard to beat.
Of course the high density options are perfect for those shops needing a large physical deployment of web servers or similar workloads. It also can be an answer for IT shops needing to address a management mandate or niche workload case where virtualization is not allowed with getting up to sixteen individual physical machines into a small footprint at reasonable cost. By supplying many smaller, physically isolated workload containers, but with a single physical footprint and single hardware management interface we can leverage some of the benefits that we would otherwise get from virtualization’s system segmentation without relying on a hypervisor.
There is no question that the FX is not designed for the average use case, it is a high density form factor designed to squeeze extreme amounts of compute power and flexibility into an extremely small chassis. One of the more difficult challenges with any limited form factor of this nature is always squeezing necessary storage into the chassis. The FX does this surprising well, but obviously cannot fully overcome the limitations of the form factor. The FX offers drive modules of up to sixteen small form factor drives, but this uses much valuable compute space making the trade off rather extreme.
Many configurations of the FX are designed to be used for stateless workloads or workloads with external storage. Most compute node options do not offer RAID storage, but just a single disk. This means that our use cases here will either be heavily dependent on some other system for data protection or for us to implement a RAIN approach (node redundancy) to overcome this fragility.
Because of this, the most obvious use cases for the FX center primarily around heavy, stateless compute needs such as are used for high performance computing (super computing clusters), decision systems (such as Hadoop), or stateless workloads such as web servers, application servers, caches, load balancers and such.
Some larger node options for the FX do include small array capabilities with numerous fast drives and RAID for protection. Because of the extremely small form factor, this is mostly focused on the super small 1.8” form factor drives and, therefore, mostly focused on SSD. While this adds considerably to the storage options on the FX, it does leave general storage use cases behind. It, instead, is focused on smaller capacity, but very high performance drive systems. This could include small to mid-sized databases or smaller general storage needs at high performance. This, in a mixed node chassis, can be effective for combining a database node with application nodes for a complete application stack in a single enclosure.
The Dell FX is certainly an interesting device, but not one that is going to immediately replace most traditional workloads. It’s really sweet spots are reserved for extremely high density, computationally heavy clustered environments, for self contained large web-style application hosting, and for small deployment scenario colocation facility deployments. The former being almost exclusive to the medium and large business markets and the later almost exclusively to the SMB market. For larger SMBs, the FX can represent a really interesting offering allowing them to deploy a single physical platform cost effectively to a colocation facility without giving up the flexibility of running several, possibly disparate, physical hosts.
Probably the biggest slam dunk deployment option for the FX is as a fully contained, single cluster web-based application location. Because of the ability to mix a high performance database on one side of the chassis and a range of application nodes on the other, it is reasonable to build an incredibly powerful, high performance application “cluster node” from the platform and deploy this, most likely, to a shared hosting facility. Then deploy matching nodes to other geographically disparate locations for global load balancing.
At first glance the FX might seem like a great candidate for a remote branch office solution, and while in very specific cases it may indeed fit this bill, it is far more often that the big sibling VRTX is going to shine far better in that role with its whisper quiet cooling, convertible for factor and enormous storage options. Branch offices are rarely space constrained in such a way as to be conducive to the FX.
The FX is an interesting device to watch and consider how it might fit into your own infrastructure.
When considering our next IPOD, we had considered the FX with 3 blade servers, but decided to pass on it since the chassis can still be the SPOF and moved on back to the basic 2-1U rack servers. Its a nice concept, but unless you can purchase more than 2 chassis for ultimate redundancies, your eggs are still in one basket, shared with the storage.
Its a nice concept, but unless you can purchase more than 2 chassis for ultimate redundancies, your eggs are still in one basket, shared with the storage.
But in IT, eggs in one basket is a good thing, not a bad one. It's better to have your storage in the same basket as your servers, even when doing an IPOD. The worst option is to separate them.