Planning for disaster recovery has long been a prudent business practice. But conceiving and implementing such...
a plan can be an expensive and time-consuming project, so companies tend to drag their heels about it -- and sometimes suffer very costly consequences when disaster does strike. Fortunately, blade servers have matured to the point where they make sense as the cornerstone of a workable, affordable disaster recovery plan. The channel can deliver real value to customers by building and selling blade server solutions for disaster recovery.
Why blade servers? Because there's no better platform for building a disaster recovery site or service. Often called a "data center in a box," blade servers excel in efficient use of space and power distribution. They reduce cable costs, are easy to administer and upgrade, and can easily be accessed for repairs. Best of all, they come in a variety of configurations to suit virtually any customer's budget. All of these factors make blade servers an attractive component of a disaster recovery plan.
One of the biggest selling points for blade servers is their efficient use of space. Blades offer the highest computing density with the lowest footprint available today. For example, a standard 19-inch rack can be populated with 42 1U servers or 64 half-height HP c-Class blades using four HP c7000 BladeSystem chassis, or 84 IBM blades using six IBM BladeCenter E chassis.
In addition to efficiently using space, each blade chassis shares resources -- such as power, cooling, network switches, storage switches, cabling and management resources -- across all the blades within the chassis, so each server doesn't require its own resources.
The shared switches provide the server blades with any-to-any connectivity through the chassis. Most switch modules slide into the back of the chassis, connecting directly to the chassis mid-plane. This eliminates cables between the blade servers and the switches and reduces the number of cables required out of the back of the chassis -- cutting cable costs and the amount of cable management required. The blade solutions that do not use a mid-plane architecture all pre-wire the chassis for ease of cable management. Most blade vendors offer switch modules that are produced by companies such as Cisco and Brocade so that the end-user management interface for their infrastructure remains the same.
The final shared resource is overall chassis management. Each chassis has a management module that allows administrators to monitor and manage all components in the chassis from one interface. These management modules come with a network interface for remote or local management of the chassis and its components, thereby making management of the blade ecosystem effortless.Since blade servers are densely packed, the companies offering blade chassis solutions have put a great deal of time and effort into creating highly efficient power distribution systems and cooling capabilities within the chassis. On average, a blade chassis uses 30% less power than the same number of 1U or 2U servers. Each watt saved in power equals a watt of cooling saved. However, as noted above, many blade servers can be placed in a chassis, requiring a more efficient cooling architecture as well as intelligent disaster recovery site planning to maximize cooling efficiency. The IBM BladeCenter and HP BladeSystem chassis both use front-to-back cooling technologies -- which, depending on the size of the disaster recovery site, can create hot and cold aisles, much like standard rack configurations. Alternatively, the Verari Systems Vertical Cooling Technology moves the hot air out of the top of the box using a high-velocity stream of cool air. It is important that your customer, when planning a disaster recovery site, understands the power and cooling capabilities of blade server offerings; toward that end, you can incorporate disaster recovery site planning into your solutions mix.
Another advantage of blade server solutions is that they are easy to access for maintenance and easy to provision, or add new blades to a chassis. When configuring a disaster recovery site, you can choose to remove all or nearly all state, i.e., identifying information, from a blade server, by using blades without local disks (diskless blades). In this configuration, you would boot the server from a storage network using the PXE protocol. This would allow the disaster recovery site to boot any supported operating system (OS), dependent upon the customer recovery scenario, rather than hardwiring the boot to one OS. This, together with I/O virtualization, eliminates the need to reconfigure the network (both LAN and SAN) any time a change is made to the blade server, making the blades completely interchangeable. This gives the disaster recovery site maximum flexibility to manage and upgrade servers as the business requires.
Blades come in many varieties and capabilities to meet your customers' cost and disaster recovery needs. You could populate your customer's disaster recovery blade chassis with a variety of CPU types, cores and speeds; with or without local disk; fast or slow networking capabilities; or the ability to boot from SAN or not.
If you deploy a disaster recovery system for a client using blade servers, be sure to recommend that they use disaster recovery software that allows for recovery to dissimilar hardware. This way, your customers can be running their applications on either bladed or non-bladed servers but can recover to a blade server. This will give your customers the ability to take the greatest advantage of the blade options available within their price range.
About the author: Anne Skamarock has been involved with computers and associated technology for nearly 30 years. She started her career as a software engineer and a Unix systems administrator. Since then she has worked as a software engineer and in technical sales, marketing and management for SRI International, Sun, Solbourne Computer and StorageTek. She recently coauthored Blades and Virtualization: Transforming Enterprise Computing While Cutting Costs.