Architecting a vSphere Compute Platform : Scalability and Designing Physical Resources : 5.3 Blade Servers Compared with Rackmount Servers
   
5.3 Blade Servers Compared with Rackmount Servers
The long-standing discussion regarding blade servers compared with rackmount server systems for use with cloud platforms has no definitive correct answer because every use case is different, and no one solution has superiority over the other.
The likelihood is that each service provider has their own preference when it comes to their cloud platform compute hardware. However, it is important to be able to work through the advantages and drawbacks of each option when it comes to server form factor.
Blades have the density advantage, and with rack space often provided at a significant cost, higher density is a good thing. Centralized management also provides a major advantage. A single interface where you can see and work with all of the blades, creating server profiles, adding networks, adding SAN connectivity, configuring VLANs, and managing power provides a serious advantage over rackmount servers. The ability to perform all of these tasks, and many more, on hundreds of blades from a single pane of glass has many operational advantages.
On the downside, blades are expensive and require special power connections. You are also locked into a single vendor when you buy a chassis or enclosure. The initial setup of blade systems can be time consuming and require additional skills and experience. Their setup is often a manual process although once done, adding additional blades to an existing chassis is simple and is often extremely simple, cloning the configuration for an existing system.
Though blade system density is higher, there is often a maximum of two enclosures per cabinet because of the weight of a full enclosure. Two full enclosures can weigh in at over 900 pounds so it is necessary to spread the load across additional floor space as well as spread the power consumption across multiple 32 amp connectors.
References to rackmount systems typically describe 19-inch rack-mounted 1U, 2U, or 4U servers with their own power connections, network connections, local disks (possibly), and HBA cards.
Rackmount servers are independent, which means that you can insert them into any 19-inch server cabinet without the need for special wiring, power, or enclosure. Rackmount systems are often more affordable than their blade counterparts and typically come equipped with KVM access. Older rack systems can also generate much more heat and pull a great deal of power compared to modern blade systems. Therefore, rackmount systems can require much cooling and high volumes of airflow to keep their components at a working temperature, although these factors are improving all of the time.
So, when balancing the advantages and drawbacks of blades compared with rackmount servers, typically employ blade systems when:
You need higher density and can support the weight.
You have skilled, trained blade administrators.
Centralized management provides operational benefits.
Rapid and simple deployment of new compute nodes on the existing cloud platform is required.
You want to continue with an established vendor relationship and are amenable to lock-ins.
The numbers of servers required justifies the cost of the chassis.
Employ rackmount systems when:
CapEx budget is tight.
You do not want to be locked into a specific hardware vendor.
You do not mind standard (“old school”) management procedures.
You have sufficient cooling, power, and airflow to accommodate standard systems.
No service provider data center has just one compute form factor in it. Most large data centers have every kind of system imaginable, with systems from every vendor, often dating back decades, and as discussed earlier, there are advantages and disadvantages to both. There is no single correct answer or a single ultimate architecture for a VMware Cloud Provider Program cloud platform.
Whichever compute architecture type you choose, the aim is to create a building block approach for the platform that will enable both horizontal and vertical scaling of resources. Standardizing the configuration and using a consistent hardware platform helps to provide a manageable and supportable infrastructure by eliminating variability. Maintaining this consistent and standardized building block approach to compute resources:
Simplifies capacity planning and scaling-out
Simplifies automated installation, patching, and configuration of hosts
Simplifies troubleshooting, fault finding, and configuration management
Preserves VMware vSphere vMotion® compatibility across the entire platform
Minimizes unused resources in VMware vSphere High Availability (vSphere HA) configurations
With the rapid adoption of next generation hyper-converged VMware vSAN based storage, a new dimension has been added to the blade versus rackmount debate. Due to their small form factor, blade systems do not typically provide adequate local SFF disk spaces to deliver sufficient capacity, and as such, make vSAN a viable option. For this reason, purchasing a new blade system for cloud platform projects typically removes the option of either deploying vSAN as part of the initial implementation, or later repurposing hardware to support vSAN distributed datastores as part of, or as a complete, storage solution.
In addition, if you consider the commodity-based hardware concept of the software-defined data center, the vendor tie-ins associated with blade systems are not typically conducive to this model.