Server Hardware Explained (Part 2)

• Published: Sep 15, 2011
• Author: Brien M. Posey

Introduction

In my first article in this series, I talked about the various form factors used by network servers. Although it might at first seem a bit odd, I want to turn my attention to server cases. My reason for doing so is that server cases often have a number of features that are designed to support the server’s overall wellbeing. Since my goal in writing this series is to educate novices on server hardware I wanted to start at the beginning.

 

Cooling Features

Computers must be kept cool in order to function properly. All kinds of bizarre errors can occur if a system's processor begins to overheat. Servers use devices such as heat sinks and CPU fans to keep components cool, but the case design also plays a vital role in regulating the server's overall temperature. Blade servers however, are the exception. The blade server chassis is responsible for cooling the individual blade servers. For rack mount and tower servers however, the case is specifically designed to keep the server cool.

The actual cooling features that are found in a server's case vary from one server to the next. At a minimum, a server typically has at least a couple of case fans. The fans are arranged in a way so that cool air is pulled into the server through vents. This cool air passes over the individual system components and is then exhausted from the system's case. The size of the fans and the number of fans used can vary considerably depending on the case design.

Some server cases rely solely on case fans that really aren’t much different from those found in PCs. Others use a more creative design. For example, I recently saw one server that was equipped with exhaust fans located between each expansion slot.

Servers are also typically equipped with temperature sensors. Sometimes these sensors are integrated into the system board or into individual components, but they may also be built into the case. If the temperature inside the case begins to rise then the fans will spin faster in an effort to keep the server operating at a safe temperature. The system BIOS usually also has a safety mechanism that can shut down the server before damage occurs if the temperature gets to be too high.

 

Filtering

Sometimes fans alone are not enough to keep a server cool. In industrial areas a server’s fans can suck in dust, dirt, soot, and other contaminants. These particles can clog air vents and may also prevent fans from operating efficiently. In order to prevent this type of contamination, some (but certainly not all) server cases are equipped with built in air filters.

Air filters do a good job of keeping stray particles of contaminants out of a server case, but they require routine maintenance. Air filters (which are usually washable) must be cleaned on a regular basis. A dirty air filter can actually restrict air flow just as badly as if the server’s vents were completely obstructed.

 

Physical Security

Physical security doesn’t seem to be a major priority when it comes to server case design. Many server manufacturers seem to assume that servers will be placed in physically secure datacenters, so there is little need for physically secure server cases. However, there are aftermarket server cases that are designed to provide good physically security. Such cases are usually key lockable and are often made of heavy gauge steel.

 

Status Indicators

Servers almost always have various forms of status indicators. The status indicators are typically LED lights. Status LEDs are not unique to servers. PCs also use status LEDs. For example, almost every PC has a power indicator and a hard drive activity indicator. Servers offer such indicators too, but often include additional indicators that are not usually found on PCs.

Each server manufacturer uses their own proprietary set of indicator lights. However, there is usually a global indicator light that can convey the server’s status through colors and / or flash patterns.

Some servers also have a locate light. The idea behind the locate light is that when you are managing a server in a large datacenter it can sometimes be difficult to figure out which physical machine you are connecting to. The locate light helps you to spot the machine that you are managing. For example, if you were to connect to a server using a management tool and then discovered that the server needed to have a hardware component replaced. You could illuminate the server’s locate light and use it to figure out which physical server needs to be repaired. Not all servers have locate lights, and on some servers the simple act of connecting to the server with a management tool causes the locate light to illuminate.

In addition to the status indicator LEDs, there are a few server cases that have integrated thermometers that visually display the temperature inside the case.

 

Ports

In most datacenters, servers are only accessed remotely. Even so, servers typically include keyboard, video, and mouse ports. These ports can be used during the initial setup or in those rare situations in which it is necessary to perform diagnostics locally.

In the case of blade servers, individual blade servers do not include keyboard, video, and mouse ports. These ports are provided through the blade server chassis.

In addition to the keyboard, video, and mouse ports, servers also typically include USB ports. It is becoming more common for servers to provide USB ports in place of keyboard and mouse ports since USB ports take up less space and can be used as a keyboard or mouse connection.

 

Drive Bays

Most servers contain at least one or two drive bays. The biggest determining factor in the number of drive bays that a server can accommodate is its form factor. For example, a 1U server might only have one or two drive bays, while some 4U servers offer dozens of drive bays.

Servers with smaller form factors often use local hard drives solely for the operating system and connect to remote storage (such as a SAN) for everything else.

 

Swing Out Bezels

Another set of features that are often found on rack mount servers are swing out bezels. To help you to understand what swing out bezels are and why they are important, imagine for a moment that you have a rack with half a dozen or so servers installed in it. Now, imagine that you need to add some memory to one of the servers that is located in the middle of the vertical stack. How do you do it? The top, bottom, and sides of the server are all blocked. You could remove the server from the rack, but that’s a lot of work for a simple memory upgrade.

This is where swing out bezels come into play. Swing out bezels allow you to slide the entire server forward so that you can open the server’s case without having to remove the server from the rack. The There are various designs for swing out bezels, but they are usually integrated into a server’s side rails.

Server Hardware Explained (Part 1)

Information shared from http://www.windowsnetworking.com

Author: Brien M. Posey

Introduction

A few years ago I wrote an article series for this site called Networking for Beginners. To date this series remains immensely popular and I routinely receive E-mail messages from readers asking me to continue the series. Unfortunately, continuing the Networking for Beginners series isn’t an option for various reasons, but I wanted to do the next best thing and write a series that is designed to introduce beginners to the basics of server hardware.

 

The Difference Between PCs and Servers

When I first started out as a network administrator one of the concepts that I had trouble grasping was the difference between a server and a PC (from a hardware standpoint). Don’t get me wrong, learning the differences between a PC and a server isn’t rocket science. It’s just that I didn’t have anyone who could explain it to me.

I got my first job as a network administrator almost twenty years ago during my first year of college. At the time I had a really strong working knowledge of PC hardware and of DOS (which was the desktop operating system of choice at the time). However, I had never been exposed to network servers before. My only background in networking at that point involved IBM mainframes.

Prior to interviewing for the position I memorized a book on Novell NetWare and managed to pass myself off as a networking expert. When I started my new position I didn’t have any trouble working with the network operating system, but I was a bit baffled by the server hardware. I could only assume (incorrectly) that the servers that we were using were highly specialized computers that were only capable of running server operating systems.

A couple of years later Microsoft released Windows NT and I decided to attend a training class to see if Windows NT would benefit our organization. The training center where I took the class provided each student with a PC and then proceeded to teach us how to install Windows NT Server onto it.

This really confused me. I just couldn’t understand why the company that I worked for was spending thirty thousand dollars for each server when a PC that cost a thousand dollars would run Windows NT just as well. Since our servers used exactly the same CPU architecture as our desktop PCs, my only conclusion was that our hardware vendor was ripping us off. After all, it seemed that from a hardware prospective there was absolutely no difference between a PC and a server other than the price.

Obviously I was wrong. I won’t bore you with the details of what the differences between our PCs and our servers were because all of that hardware is completely obsolete today. Even so, I wanted to share this story with you as a way of illustrating the point that there are major differences between PC hardware and server hardware. Sometimes those differences just aren’t immediately obvious to beginners.

In many ways server hardware really isn’t all that different from desktop PC hardware. Both use the same basic components such as memory, CPUs, and power supplies. In spite of these similarities however, server hardware can seem completely foreign to those who have previously only dealt with desktop hardware. The individual components tend to be more advanced than the components that are found in desktop computers. Servers may also make use of redundant and / or how swappable components. Occasionally PCs might have redundant or even hot swappable components, but such features are much more common on servers.

 

Server Form Factor

So now that I have talked a bit about some of the differences between PCs and servers, I want to turn my attention to server form factors. When it comes to computers, the term form factor describes the physical dimensions and standards used by various system components and by the computer as a whole. To give you a more concrete example, many PCs use a form factor called standard ATX. A standard ATX case can physically accommodate any standard ATX system board and power supply, even if they are made by a different vendor.

Another common PC standard is Micro ATX. Many standard ATX computer cases are designed to accept either standard ATX or Micro ATX system boards, but a Micro ATX case would not be able to accommodate a standard ATX system board because Micro ATX cases use a smaller form factor.

Form factors apply to server hardware as well, but servers typically adhere to different form factors than PCs (but not always). There are several different form factors used for network servers, but the three most common form factors are tower servers, rack servers and blade servers.

 

Tower Servers

Tower servers look a lot like PCs. Each tower server is a standalone machine that is built into an upright case.

Tower servers are used mostly in smaller datacenters. Larger datacenters typically avoid the use of tower servers because of the amount of physical space that they consume and because they tend to be noisy.

Another disadvantage to using tower servers is that the cabling can get messy. Server racks and blade server chassis usually have features that are designed to manage cables, but tower servers have no such features.

 

Rack Servers

As the name implies, rack servers are servers that are mounted within a rack. The rack is of a uniform width and servers are mounted to the rack using screws. Each rack can accommodate multiple servers and the servers are typically stacked on top of each other.

Because racks are designed to accommodate standard sized components, many hardware vendors offer rack mountable networking components other than servers. For example there are rack mountable network appliances (such as hardware firewalls) and rack mountable switches.

Rack mount components follow a form factor that is referred to as a rack unit. A standard rack mount server is referred to as a 1U server meaning that it is 1 rack unit in size. A 2U server consumes two rack units of space within the rack. Some vendors also offer 4U and ½U servers. The larger form factors are usually used when the server needs to be able to accommodate a large amount of storage.

 

Blade Servers

Like rack servers, blade servers also adhere to a standard size and mount inside a special “rack”. In the case of a blade server however, the rack is known as a chassis.

Blade servers tend to be vendor proprietary. You can’t for example insert a Dell blade server into an HP chassis.

The reason why blade server design is proprietary is because unlike a rack server, which is fully self-contained, blade servers lack some of the components that they need to function. For example, blade servers do not have power supplies.

The blade server chassis is designed to accept various modular components, including the blade servers themselves. For example, a chassis might contain a power supply unit, a cooling unit, and a blade server. The actual chassis design varies from one vendor to the next, but most blade server chassis are designed to accommodate multiple power supplies, multiple blade servers, and a variety of other components (such as network adapters, storage modules, and cooling modules). With the exception of the cooling components, individual blade servers are mapped to the individual modules or components.