The server room

Middle East enterprises are busy building or consolidating data centres - so busy they might miss out on the latest products. In the first of a two part series, NME presents the technologies and practices that enterprises should consider to get more out of the data centre.

  • E-Mail
By  Administrator Published  May 1, 2007

Whether building a new data centre or consolidating existing ones, infrastructure and IT managers across the region are busy like never before.

Increased communication and data needs, coupled with international standards and the appearance of new vendors, has led to a deepening awareness of the importance of constructing and maintaining a data centre that maximises a business's IT investments and adds efficiency to its data processes.

This sudden growth, welcome as it may be, comes with its own set of problems - such as the sheer volume of new products, breakthrough technologies and international standards available in the region.

"The Middle East is an amalgamation of cultures. It has influences from Europe, Asia and the Americas. Technology choices in the area reflect these various regions in terms of standards and product availability. This can create a lot of confusion in the minds of customers and IT managers and requires from the vendor a bit more of an educative role," points out Gautier Humbert, technical and training manager for Ortronics Middle East.

Some argue that such a mixture has led to a higher failure rate of implementations in the region than anywhere else in the world.Others say, the Middle East has a singular advantage in that it can draw lessons from countries which have made mistakes earlier.

Failure in the data centre can be a costly affair. Middle East enterprises need to learn fast to avoid slip ups, leapfrog technologies and ensure the data centre lasts long into the future.

And such learning starts with the infrastructure.

Setting the connection

Cabling has the dubious fame of being the one infrastructure that tends to last the longest time in any data centre. All the more reason that extra caution should be paid when you choose and put it in place. Talk to any cabling vendor and there are two constant mantras that you would likely hear - 10 Gigabit and Cat 6a.

"There is no doubt about it. 10 Gigabit on the server side is coming soon. If you are building a data centre today, you should certainly put in place Cat 6a cabling. With Cat 6a they can be sure to get optimal benefits when 10 Gigabit hits," says Wes Tweedley, regional technical manager, Middle East, Pakistan and Africa for Systimax Solutions.

Most other vendors agree with him. General trends in the region indicate that businesses are more likely to go for Cat 6a in green field data centre environments. There is also a steady shift to shielded cables from UTP (Unshielded Twisted Pair) across new and consolidating data centres.

Fibre in the backbone continues to hold sway among businesses.

Laurent Amestoy, regional manager for R&M, the Swiss cabling vendor says, "There will be an increasing move towards higher end fibre cabling systems - from multimode OM2 and OM3 to the single mode OS1. The required network support, the application bandwidth requirement and the distance to be covered will be factors in determining which fibre is used."

However, copper and fibre will continue to share floor space inside the data centre.

The battle on what percentage of floor space each would occupy rages on with both sides claiming eventual victory. All advice, though, is to pick either solution based on the application you need it for and the bandwidth requirements that go along with it.

"There is a definite interest among Middle East companies to go for the latest technologies. But the market is also a highly price sensitive one. Even the largest enterprises tend to make decisions based on price. Lack of specific knowledge on what the higher-priced solution can provide leads to this choice," says Humbert.

While a niggardly look into apparent cost might give you a product that will not sustain beyond a few years, ignorance of installation procedures could mean you lose out on even the best products bought at the highest cost.

According to industry observers, nearly 70% of all failed cabling implementations can be traced back to improper installation. Often users ignore important landing and grounding parameters which leads to disastrous results.

One likely solution is for businesses to request a list of trained installation partners from cabling vendors. Most often, vendors provide a guarantee of success with these partners, which is a major worry off an IT manager's shoulders.

Cooling off

It is surprising how many organisations building data centres in the region tend to ignore the importance of power and cooling solutions.

"Power and cooling are crucial to extract optimal performance of physical infrastructure inside a data centre. They are elements that should be considered early in the design phase and a data centre should be built with them in mind," says Gary Highton, MD of Mayflex Middle East.

Power redundancy is essential in a data centre. This ought to be planned based on the data centre level desired - Tier I having no redundancy and Tier IV having provision for multiple distribution paths. Often companies depend on battery and generators for power redundancy.

"In countries where power is provided by a single supplier, getting lines from two sub-stations might not be possible. Some thought to location can help overcome this. For example, a company can be placed on the border between Dubai and Abu Dhabi and draw lines from both cities in order to build power redundancy," says Tweedley.

As the average heat produced by servers increase, cooling them for performance optimisation is becoming increasingly important.

"There is a lot happening with cooling solutions. While traditional air cooling systems are getting more sophisticated there are the newer liquid cooling systems. These pump water through the racks and tests prove that they are more effective than air cooling alone," says Dr Saeed Al Barwani, CEO of eHosting Datafort (EHDF).

"Using a combination of air and liquid cooling will give maximum cooling capacity and provide fall back options in case one or the other fails."

The boxes in the room

Some would say that servers are the most stable elements of the data centre infrastructure. Over the last few years, there has been minimal, if any, changes in the architecture of servers. This still favours RISC and x86 boxes. However, performance metrics and the physical state of servers have changed quite a bit.

"Blade servers and virtualisation are the two biggest trends in the region. The former allows a lot more flexibility in usage terms to customers, especially ones who are starting out on data centres. The latter allows for a comprehensive usage of resources especially when you are consolidating infrastructure," says Yasser Ragaei, business manager of enterprise servers for HP Middle East.

Every other vendor swears by the increasing importance and sales of these two in the Middle East market. The other benefit of the two trends is their ability to save on power.

"As cores increase within a server, it is essential that servers are tuned to consume less power and produce less heat than others in order to survive longer and prove more efficient. Our servers do exactly that," says Iain Jardin, SPARC product sales manager, South Eastern EMEA, Sun Microsystems.

Dell, which recently launched Energy Smart server models, claims almost the same.

"Our servers use a lot less power that other similar servers and produce less heat. The entire Energy Smart program is built on these metrics," says Jul Johnson, solutions manager for Dell ME.

One likely solution would be to conduct pilots with several server vendors to get a clear picture on how the products function in your specific environment.

From beginning to end

Building or consolidating a data centre in the region can be a confusing job and one in which it is easy to make mistakes. There are multiple technology choices and it is easy to get carried away or make impulsive decisions which can potentially harm the network.

However, an IT manager can achieve success with a data centre simply by following processes rigidly. And the first step towards that involves understanding business requirements.

"Business needs are the key. Technology is just an implementation tool. IT managers and top management has to work from day one keeping in mind what the business needs. This includes what kind of mission critical applications are required, what kind of downtime can be afforded and what the end goal is. Only then can the process towards a data centre begin," says Jardin.

Getting a set of requirements in place is easier said than done. As the user plays the most critical role at this stage, he can draw upon expertise to ensure the process runs smoothly and produces more effective results. This help can come in the form of consultants or even vendors.

"We play a huge consultative role with customers. We work with the end user and advise them on ways to extract more from the data centre ," says Humbert.

Dell's Johnson says, "There are examples of customers worldwide who have requested Dell to help them in designing and putting in place their data centres. And Dell has done it for them. We have not done anything similar here in the region but the truth is we can draw upon such capabilities to help customers."

Businesses need to put similar effort into the design phases of the data centre. This is where the entire architecture of the room is planned out and decisions are made on how each element would fit into another. Improper design can result in bad cable management, ineffective cooling or other mishaps.

"Ideally users should plan their data centres in such a manner that they use 60% to 70% of the capacity of the centre initially and rely on the remainder to provide scalability in the future. Also, they should plan for infrastructure that lasts around three years. After that a review and redeployment might be needed," says Tweedley.

The truth lies in management

All the effort that goes into data centre planning and construction can go to waste if it is not monitored and managed properly.

"Vendors have a role to play in educating users on the best ways to manage and get the most from their data centre. These training sessions might even have to be rather regular to keep them updated on the latest in the industry," says Johnson.

"Human error is the biggest reason for any failings in a functional data centre. Users are critical to the success of any project. It is essential that users put in place stringent management processes and stick by it to avoid a complete mess in the data centre in the future," says Tweedley.

Effective management comes with its own hassles and is a story in itself. The next part of the series will look into the various aspects of data centre management, security considerations, consolidated storage and business continuity.

The driving forces

Gartner research defines eight critical forces affecting enterprise data centres today. They are:

1. Processor/System Design/Power - embraces watts, core, threading, server, storage and networks

2. Architectural Topology/RTI - includes server density, types of servers and virtualisation

3. Operational Processes/Tools - covers remote management, utilisation, configuration changes and human resources/skill sets

4. Facilities Modification - includes the supply and cost of power, design of cooling, structure of floors and external suppliers

5. Consolidation rationalisation - consists of floor space and new technologies

6. Disaster Recovery/ Business Continuity - includes data management, telecommunications and data centre distribution

7. Capacity Growth - covers increased floor space, collocation and outsourcing requirements

8. O/S, application changes - comprises of platforms such as Linux and Windows, SOA and legacy applications


Cool planning

NME presents Gartners estimates for power and space requirements for a data centre. The planning model below is a simplified version for a base case example.

A key objective of this planning model is to maintain a power rating of approximately 4kW per rack. Research has determined that densities higher than this increase power demands disproportionately because of added heat gain, resulting in an increase in overall operating costs.

Space per Rack: Use 30 square feet per rack, which includes space for the rack as well as aisles, ramps and clearances between racks. The planning model assumes a traditional cooling scheme - that is, a front-to-back, bottom-to-top cooling distribution.

Servers per Rack: This varies by vendor product. Here, we've assumed a server configuration of 10 blade servers per enclosure and a total rack configuration of five enclosures, equaling a total rack configuration of 50 blade servers.

Power per Rack: Using the vendor's power specification, we can calculate the total wattage per server enclosure as 505 watts.

We now have a tool to estimate total facility capacity from a space and power perspective. Assume that computing capacity requires the equivalent of 200 racks of servers.

Expansion capability: The outer limit (100 watts per square foot) would permit the addition of 97 racks (at five enclosures per rack), requiring 2,900 square feet of the 6,000 square foot expansion space. Anything above this level would require additional power and air conditioning capacity, assuming a design rating of 100 watts per square foot.

Add a Comment

Your display name This field is mandatory

Your e-mail address This field is mandatory (Your e-mail address won't be published)

Security code