By Anil Patrick R, Chief Editor, SearchDataCenter.in
Indian organizations are at present trying to come to grips
with the post-liberalization era's aftereffects, when rapid organizational growth created an exponential rise in applications (as well as in server farms). Sanjay Sharma, the managing director and CEO of IDBI Intech Ltd., identifies with the issues that accompany such growth. After all, in his dual role as IT advisor at IDBI Ltd., Sharma has spearheaded consolidation and data center relocation projects necessitated due to such unexpected business growth.
According to Sharma, all IT departments face a typical challenge when the number of applications increases. "Each application vendor asks for things like separate infrastructure and platforms, so you keep on stacking a lot of servers. In the end, you face real estate problems, since a lot of space is occupied. You need a more powerful technology which offers the option of expansion, and is less power hungry in the server space. So consolidation around a lesser number of servers is where blade server architectures score," Sharma says.
Now the question arises as to what constitutes a blade server. Technically, the blade server architecture consists of just two basic components — chassis (also known as the enclosure) and the blades (modular, stripped-down servers). The blade chassis is a full-fledged ecosystem for the individual blades with its shared inbuilt power, networking, storage and cooling capabilities, as opposed to a normal rack-mount server design.
Blade system architectures yield many benefits like small, energy-efficient server blades, a significant reduction in cabling, simpler configuration and the ease of management of an all-in-one design, according to James Staten, a principal analyst
at Forrester Research Inc. "Blade server architectures also provide significant configuration flexibility, offering a choice among myriad servers, I/O options and other internal components. The chassis can accommodate a mix of x86 (Intel or AMD CPUs) and Unix RISC servers, storage blades, workstations and PC blades, as well as multiple I/O connections per blade," Staten says.
Adding to these are recent blade server architecture developments such as integrated thermal management options that optimally adjust power consumption and performance based on temperature changes. The stripped-down nature of individual blades also ensures that hot-swappability is never a concern. Such benefits have endeared blade server architectures to Indian businesses.
Apart from occupying less space, power consumption is a major consideration driving the demand for blade servers in an energy-deficit country like India. Ease of management is the next critical aspect favoring blade server adoption, since it's possible to manage a large number of servers from a single console. "Operational cost management and ease of manageability seem to be the primary objectives behind blade server architecture's adoption in India. Blade servers can bring about reduction in power consumption, as the energy supply-to-demand ratio is still skewed in India," observes S Vijay, a director with the IT advisory practice at KPMG in India. Cooling concerns surrounding rack-mount servers can also be alleviated, since it's already incorporated in the blade chassis.
The power of virtualization
M V Sreeram, vice president of IT at Reliance Power Ltd., is a firm believer in the benefits offered by blade server technology. "In a single blade chassis, you can have different operating systems, different memory capacities, a mix and match of 32-bit or 64-bit CPUs, and so on. Once you have these, you can always run virtualization software on top. Today, users demand dedicated servers for their applications. So virtualization software really saves your IT team's day," Sreeram says.
Blade architectures are a good fit for server virtualization since they complement the server consolidation value, as well as support much higher shared aggregate network bandwidth. "Plus, certain blade software solutions let you pre-assign blade servers' network and storage configurations, and easily move these assignments to another blade server when needed. In essence, blade architectures let you pair your dynamic software with dynamic hardware, making deployment and management of virtual servers that much easier," Staten says.
Selection of blade server architectures
Once you have decided to go in for blade server architecture, networking capabilities are an essential consideration. The backplane throughput should be measured to find its maximum output as the first step. Always opt for a blade system that provides support for technology such as 10 GbE, fiber channel and 8x InfiniBand. "If you plan to deploy high-performance applications or high-density server virtualization, then I/O throughput should matter to you greatly," Staten says. "Most VMware servers (especially those protected by VMware HA or Fault Tolerance) need a minimum of six I/O connections: redundant IP, redundant storage and redundant management paths. These could be greatly consolidated if you are using 10 GbE to the chassis; otherwise, you should be ready to invest in a lot of adapters."
Supported storage subsystems should be checked at this point. It's also a good idea to check if the blade system being considered is interoperable with your existing storage area network, preferably with a demo setup.
Automated blade failover is the next critical parameter. "The system should allow the WWN and MAC address of a blade server to be reassigned to another blade either in the same chassis or in another chassis from the same vendor. This feature can also be used to pre-assign network and storage attributes to particular server bays. This way, when you pop in a new server blade, it will automatically come up into a given role (assuming you network-boot the blade)," Staten explains.
Power consumption comes next on the agenda. Intelligent power management features should be paramount on your checklist at this point. Power-saving capabilities should be available to the CPU, various motherboard components, power supplies and fans. The blade server should also have intelligent service processors to keep systems dormant during low demand.
Implementation of blade server architecture
Consolidation ensures optimal outputs, so this should be the first measure to undertake before blade server deployment. After the consolidation phase, implementing server virtualization can be yet another optimization measure to derive maximum benefits from your blade server rollout.
Due to the blade server architecture's inherently higher server density, it's best to refrain from populating the entire chassis. "Normally, it's not required to fill up all the slots. It's best if you utilize only 80% of the total capacity," Sreeram says. In this context, airflow between the blades is a critical concern, which has to be evaluated in-house. Once this has been ascertained, it's always safer to suitably tailor your data center's cooling capabilities. "Blade server architectures are designed for optimum air flow. Therefore, it is essential that data centers housing blade servers should have an adequately designed hot aisle and cold aisle," Vijay recommends.
Apart from the occasional firmware updates and patches, blade server architecture is usually low maintenance in nature. However, one aspect to remember is that the blades' availability is completely dependent on the chassis' well-being, so the required recovery and continuity procedures should be in place to account for chassis failure.
Since most blade server deployments are based on (or become) virtualized environments, staff training is an important parameter. "Adequate training should be provided to personnel to develop skills to manage a highly virtualized and service-oriented environment," Vijay opines. This will ensure that you continue to reap optimal benefits from your blade system architecture.
This was first published in January 2010