Today, Cisco introduced the Third-Generation of Fabric computing. The power of unification can be seen in all aspects of the Cisco Unified Computing System (UCS). It unifies physical and virtual compute environments. It integrates the server and network access. It also unifies and simplifies the management of rack and blade servers. The choice between using blade and physical server is not obvious and usually sparks a heated debate. The Cisco UCS Manager with B-Series blades and C-Series Rack-mount servers makes this argument moot.
Without passing judgment on whether rack or blade servers are better, lets begin by listing the benefits of the two form factors. Blade server use is growing and there are many reasons for this:
- Data center space / Power / Cooling constraints – Data center space can get expensive depending on the location. In this CBS 60 minutes segment, proximity to the stock exchange is extolled. This fascinating video shows the collocation of data centers in urban nerve centers. I am sure data center space in these locations commands a premium. Energy efficiencies also become important in the overall cost structure. Power constraints may also favor the use of blades that consume less power than equally powerful rack mount servers.
- Need for homogenous servers with simplified management – A homogenous data center provides a management model that scales well with larger number of servers. The efficiencies gained in managing through a single interface offset any higher initial costs and reduce the total cost of ownership. Homogeneity also provides a predictable gain in load handling (scaling) capability when the workload is horizontally scalable.
- Well understood scaling needs – When the scaling needs are well understood and the rate of server addition is known, blade servers fare well as deployment of additional servers is quicker and predictable.
- When a Storage Area Network (SAN) and Network Attached Store (NAS) infrastructure is already in place- The time to value of blade servers is faster in data centers where the storage infrastructure is in place and ready for use. Blade servers are well suited for diskless boot and the storage infrastructure facilitates faster repurposing of servers.
Rack-mount servers have their advantages, and in some case maybe the only option.
- Workload specific needs such as special hardware cards, local storage, vertically scalable, memory bound – Typically rack servers can pack more hardware such as memory into a single server for the same cost. This may be useful from a server consolidation standpoint. Certain workloads can only be vertically scaled or may need local storage in which case rack servers are a better choice than blades. For memory bound applications, software-licensing costs decrease if it is based on the number of CPUs.
- Real estate, power and cooling are easily available – When there are no constraints on space, power or cooling, rack mount systems may be perfectly suited for the data center.
- Scaling needs not well understood – When scaling needs are unclear and rate of server addition is not known, procuring rack servers as needed is suitable. Servers procured at later times may have different management interfaces and components. Management of this heterogeneous set of servers and components can get progressively complex.
- Capital Expense constraints- If initial cost of hardware is a consideration as opposed to the operational expense, rack mount servers may give a better bang for the buck.
- Form factor freedom to deal with constraints and needs – Imagine a 4u blade server! Treat a 4u rack server with lots of memory as if it were a blade server
- Workload mobility across blade and rack – With the Cisco UCS Manager Service templates and Service profiles, workloads can be moved between servers of different form factors
- Single management interface across blade and rack servers – The Cisco UCS Manager provides a single comprehensive management interface across rack and blade servers