IT Management Reference Guide

An old adage states, 'Be careful what you wish for.' Never has that saying been more apt than when applied to the management of today's data centers. For years, data center managers lamented the fact that ever-increasing demands from customers for more services and applications – and the equipment they would require – resulted in ever-increasing demands for precious raised floor space. The recent advent of high density storage systems, communication equipment and especially blade servers has temporarily eased the demand for expanding data centers. But this new generation of equipment brings with it its own perplexing set of challenges: how to effectively power and cool these new devices. Blade servers, with their extremely high density and large amounts of heat generation and electricity needs, pose an especially daunting obstacle.

This section discusses how we got to where we are today with the heating challenges of modern data centers, and what we can do to address these issues. Included are ten specific suggestions for 'Keeping It Cool'.

How We Got To Where We Are Today

In today's environments, data centers, network operations centers and facilities for hosting web operations are all very different from what they were just a few years ago. Smaller, faster and more powerful blade servers, communication switches and other electronics are being packed into more densely populated spaces. Devices that used to fill several rooms can now be housed in a single cabinet. This results in the concentrating of power demand, and the subsequent heat these devices generate, into small specific areas of a computer room. These isolated sources of heat complicate the task of providing the proper quantities of cooling to the proper locations.

It wasn't always this complicated. When mainframe computers first became prevalent for businesses and universities in the mid 1950s, the generally accepted method of heat removable was by ambient air cooling which remained through the mid 1960s. The introduction of the IBM System/360 in April 1964 ushered in the era of large, general purpose computers with significantly higher demands for cooling. The accepted practice for cooling from this point in time through the 1970s was to install dedicated air conditioning units in the computer room and to use raised flooring to force cold air underneath the equipment. During the 1980s, a combination of air cooled and water cooled techniques were used to remove heat from mainframes.

In 1995, with client server systems flooding computer rooms with cabinet housed equipment, the first dedicated, air-cooled, rack-based systems came on the scene. In 1999, the air-cooled, hot aisle/cold aisle configuration of racks was introduced. By 2002, high density blade servers brought the issue of hot spots and spot cooling to the forefront. This brings us to today where ever-increasing equipment densities are bringing more challenges to data center managers to keep their computer rooms adequately cooled.

Ten Suggestions for Keeping It Cool

So what are some of the methods used today? The following is a list of ten suggestions that data center mangers can use to plan and implement effective methods for keeping their computer rooms cool.

1. Hot Aisles/cold aisles – Install rows of cabinets so that each row has alternating face-to-face cabinets and back-to-back cabinets. This results in alternating hot aisles and aisles which prevents the warm backside of one row of cabinets from feeding hot air into the front of another row. Perorated floor tiles are obviously positioned only in the hot aisles.
2. Minimum clearances – Establish minimum clearances for underneath the floor and above cabinets. Raised flooring should be two to three feet deep for effective cooling. The minimum clearance from the top of cabinets to the ceiling should be a minimum of three feet.
3. Floor Tiles – The overuse of perforated floor tiles can be a leading cause of inefficient cooling. Use a minimum number of perforated floor tiles to cool the heat load in the area immediately adjacent to the tiles. Replace perforated floor tiles with solid ones whenever equipment re-locations eliminate the need for these tiles.
4. Seal Cutouts – Seal all cable cutouts and openings in the raised floor, the subfloor, walls, and above the dropped ceiling.
5. Filler plates – Use filler plates to seal unused openings in cabinets.
6. Power cables – Spread power cables out on the subfloor to eliminate tangling and to reduce hot spots. Preferably, run the power cables underneath the cold aisles.
7. Data cables – Place data cables overhead trays at the stringer level in the hot aisles.
8. Cooling measurements – Periodically measure the output of the cooling units to ensure they are generating the specified levels of cooling.
9. Throw distance – Minimize cooling unit throw distances as much as possible. Keeping this to less than 40 feet is a good rule of thumb.
10. Extreme cooling – Consider extreme cooling schemes if large numbers of high density equipment needs to be cooled. Vendors are now offering ceiling mounted extreme cooling systems for these environments. In this case the hot air from the hot aisles is drawn into the system and carried away, while cold air is discharged into the cold aisles. An added benefit of these systems is that they are very scalable which helps offset some of their expense which may be significant but cost effective.

These are some of the methods that data centers managers can use to reduce the challenge of keeping their computer rooms cool, and their critical systems up and running.