Server form factors: A guide to rackmount, blade servers and more
A comprehensive collection of articles, videos and more, hand-picked by our editors
Initially, blade servers were considered a management win, shrinking the server footprint and saving space in the data center. These efficiency and management benefits are the reason why blade servers have become a mainstay of modern data centers, but that could be changing.
Server manufacturers have advanced blade technologies to house processing power and memory resources that would have been considered a supercomputer just 10 years ago. There's no doubt that blades have delivered on the "more for less" promise, but now I'm reconsidering.
The answer to why blade servers are losing their appeal is storage -- or virtual storage appliances, to be more precise.
After years of abandoning rackmount servers and local disks for blades and storage area networks (SANs), it may be time to reconsider. Like so many aspects of technology, local storage has become both better and less expensive. At the same time, software advancements have ushered in the concept of virtual storage appliances.
Why blade servers gained traction in virtual data centers
I would still argue that blade servers are more efficient and easier to manage than rackmount servers. However, most blades are limited in the number of hard drives that they can hold, until recently, I would have considered that an advantage. Data center architectures had long been moving away from designs that placed data on local hard drives all over the data center. Storing data in local hard drives or in direct attached storage was inefficient and complicated tasks like data protection and data replication. As storage designs moved toward centralized SANs, many servers shipped with no local hard drives and let's not overlook the impact of server virtualization. The portability of virtual servers is built around the assumption of centralized storage, leaving only a small hypervisor operating system on the physical servers.
So what would cause one to rethink this proven and reliable architecture of diskless servers and centralized storage? Scale-out storage appliances.
About six years ago, LeftHand Networks introduced its Virtual Storage Appliance. This appliance took local storage on VMware ESX hosts and grouped it together to effectively become a centralized iSCSI storage device. This provided the benefits of a centrally managed and redundant storage environment, while avoiding many of the risks associated with storing operating systems on local drives. This generated a lot of buzz in the virtualization community and Hewlett-Packard quickly bought LeftHand Networks. Unfortunately, HP's initial interest appeared to be in protecting its own existing investments in SAN technologies. While HP incorporated the LeftHand OS into its storage products, the Virtual Storage Appliance appeared to go on the back burner.
A new era of virtual storage technology
Fast-forward a few years and Nutanix burst onto the scene. Nutanix incorporated blade servers with PCI-based flash storage and local spinning drives with a virtual storage technology that combines the local storage of each blade into a centralized pool. Much like the LeftHand product, Nutanix based its storage product on a software-driven approach to combine the management and resiliency of centralized storage with the simplicity and low cost of local storage. As you add new hosts to the environment, each new host will contribute its local storage to the centralized pool.
In a related vein, SimpliVity introduced its own product line with a similar storage approach. While their technologies differ in many aspects, both Nutanix and SimpliVity have a software-driven approach to building centralized storage from locally attached disks in each server. Both vendors will tell you that the secret to their storage solution lies in the software and not in the hardware.
VMware has been taking a similar approach with its vSphere Storage Appliance, though not on an enterprise scale -- at least not until now. With vSphere 5.5, VMware introduced Virtual SAN. Not to be outdone, storage giant EMC introduced ScaleIO. Both Virtual SAN and ScaleIO use local storage resources on physical servers to create a centralized storage device.
HP (LeftHand), Nutanix, Simplivity, VMware and EMC all take a software-defined approach to storage. Where Nutanix and Simplivity currently only provide their products on custom hardware stacks, HP's StoreVirtual VSA -- yes, the original LeftHand VSA is still in their portfolio -- VMware's Virtual SAN and EMC's ScaleIO will run on a wide range of hardware that you likely already own. That is, of course, unless you are running blades with no local hard drives.
This is where the blade server status quo is being challenged. Even though many blade servers can meet the requirements for using the software-defined storage products from HP, VMware and EMC, the number of drives they can contribute to the pool will be very limited.
With this technology effectively turning a server into a shelf of storage, it is easy to see a shift away from blades in favor of newer 2U rackmount servers that actually resemble storage shelves, holding up to 24 drives each.
I have not lost my affection for blade servers. They still provide a significant advantage in efficiency and management, especially in architectures like Cisco's UCS.
However, there is a new element to factor into your return on investment equation. If you are responsible for server selection or storage administration, you need to take a look at how these new technologies will affect your long-term data center direction. Even if you do not use them now, you need to keep them on your radar and develop a strategy that can accommodate the addition of these technologies in the future.