Recently, we've seen blade computing grip the computer industry. Aggressive vendor marketing pushes and user case examples appear compelling and in many cases they are. In many instances, however, the economics of blade computing have less to do with saving hard dollars and relate more to improving the resiliency and agility of IT environments that rely on commodity operating systems and internal spinning disks.
Specifically, blade computing is a natural evolution of collocated and rack-mounted servers, where instead of including power, cooling, networking, storage and cables in a single box, these components can be shared across a 'farm' of server blades by including them in an enclosure. For sure, blade computing is a 'game of inches' where very high density packaging is the norm and every bit of space counts.
The reality of this approach however is that the workloads which are best suited for blades are also those best suited for standalone servers. These tend to be the ones that are distributed computing-friendly and are not candidates for mega-consolidations and application consolidations where the really big hard dollars are being saved. So the economics of blades as compared to logical alternatives (standalone and rackmounted servers) really relate to packaging, versus a radically new style of computing.
Nonetheless, smart packaging often makes smart business, especially if blade servers are homogenous, meaning IT operations doesn't have to think about what type of blade to add or replace, they just plug it in. Architect blade computing to assume blade replacements use the same cpu and memory configuration from the same vendor. Also, isolate storage from the blade servers. If businesses follow these rules-of-thumb and have a critical mass of blade-friendly applications that can be leveraged, these benefits have been reported by users:
- 5-12% reduction in CAPEX and power/cooling costs over a three year period.
- 15-20% improvement of IT operations productivity due to simplified environments
- 20-50X reduction in cabling
These are nice, but not overwhelming statistics from the standpoint of dollars and cents. The real interesting business metrics we've found are the following:
- Blade computing can contribute to overall application availability (not hardware availability, but what the user sees) improving from 97% to 99.8%, which translates to 100 - 150 hours of reduced application downtime per year.
- Reduced time to deploy new server capacity from hours or days to minutes, meaning accelerated user benefit from new functionality being deployed along with the server capacity.
These factors avoid having to fix problems in real time which affects application availability and allows IT to remediate problems and make changes in a way that users often never see. This is the real economic benefit of blade computing and users should be wary of justifying blades solely on the promise of budget savings.