Server Hardware Explained (Part 5)

by [Published on 22 Dec. 2011 / Last Updated on 22 Dec. 2011]

This article continues the discussion of server hardware by talking about server storage options.

If you would like to read the other parts in this article series please go to:

Introduction

Now that I have talked about some of the more basic server hardware components, I want to turn my attention to storage. I have to admit that I had to think long and hard about how to approach this particular topic because there are such a wide variety of storage options for servers. As I thought about the topic, I remembered the promise that I made at the beginning of this series to try to keep things simple. That being the case, I am not going to be discussing SAN storage. If there is enough interest in the topic, I might do a separate series on it.

Internal VS. External Storage

If you think back to the article where I discussed server form factors, you will recall that there are various form factors for rack mount servers. A 2U server takes up two rack units, and is therefore twice the size of a 1U server. Likewise, a 4U server is four rack units in size and takes up twice as much rack space as a 2U server.

My point is that servers come in a variety of shapes and sizes. A server’s physical size often determines the amount of physical storage that the server can accommodate. For example, most of the 1U servers that I have seen are equipped with two internal hard drive bays, whereas there are 4U servers on the market that can literally accommodate a couple dozen hard drives.

In some ways this fact can be a bit misleading because it implies that if you need a lot of storage space then you should purchase a large form factor server. However, small form factor servers are often connected to external storage. That way they can provide as much storage as the organization might require, but without taking up too much space in the server rack. Many organizations choose to use 1U servers connected to external storage because it is often less expensive to connect several 1U servers to an external storage pool than it is to purchase several, fully loaded 4U servers.

Before I move on, there are a few things that I want to clarify. First, 1U servers are by no means required to access external storage. As I said before, most of the 1U servers that I have seen have space for two internal hard disks. One of these disks can contain the operating system and any applications that might be required, while the other disk stores data.

A second point that I want to clarify is that storage capacity is not the only reason for connecting a server to external storage. External storage is often used as an alternative to internal storage for performance reasons. Suppose for instance that you have two identical servers. One of those servers has an internal 1 TB data drive. The other server has ten 100 MB drives configured as a stripe set. Assuming that everything else is equal the server with the ten hard drives will outperform the server with the large hard drive, even though both machines have the same total capacity.

The reason for this is that because the ten hard drives are arranged in a stripe set, data is evenly split among the drives. This means that read and write operations occur across ten disks simultaneously, with each of the ten disks containing a mere ten percent of each file. As such, a file could theoretically be read 90% faster on the server with the disk array. I use the word theoretically because there is some overhead in the process that slows things down a bit, but as a general rule a stripe set is faster than a single disk.

The third point that I wanted to clarify before moving on is that even though I have been using the example of a 1U server connecting to external storage, any properly equipped server can access external storage. This holds true for 1U, 2U, 4U, blade servers, and pretty much any other type of server. In fact in my lab I have high end PCs that are running server operating systems, but that are not true servers and they have no trouble connecting to an external storage array.

Server Hard Drive Types

Most of the servers that are being sold today can either accommodate SATA hard drives or SAS hard drives. As I’m sure you probably know, most desktop and laptop PCs use SATA hard drives. The SATA hard drives that are sold for server use are actually very similar to desktop hard drives. They use a similar controller, and both laptop hard drives and server hard drives are 3.5 inches in size.

Since server hard drives tend to cost a lot more than desktop hard drives, you may wonder if you can replace server grade SATA drives with SATA drives made for desktop computers. The answer is that it depends.

In many cases you can install a desktop SATA drive into a server and the server will recognize and use the drive. However, there are some server manufacturers that use proprietary connectors on their SATA drives as a way of ensuring that only server grade drives can be installed into the server.

With that in mind you might be wondering why any sane person would pay a premium price for a “server drive” when they could just purchase a cut rate desktop hard drive instead.

There are two main reasons why it might be better to go ahead and spend the extra money for a server grade drive. The first reason is that server grade drives have a higher duty cycle. The duty cycle for a desktop hard drive varies from manufacturer to manufacturer, but most desktop SATA drives are designed to be used around ten to twenty percent of the time. Server hard drives on the other hand are specifically designed for much heavier use. It is not uncommon for server drives to be in use a full 100% of the time.

What all of this boils down to is that if you use a desktop hard drive in a heavily used server then the drive probably won’t last as long in the server as it might have if it were installed in a moderately used desktop computer. The drive probably also won’t last as long as a server grade hard drive would have. This doesn’t necessarily mean that you shouldn’t use desktop hard drives. I have friends who use desktop grade hard drives in servers that handle non critical workloads. They have told me that their annual hard drive replacement cost is lower when they use desktop hard drives, even though they burn through more drives in a year’s time than they would have if they had purchased server hard drives. Generally speaking though, server hard drives are of a better quality and should be used in any server with a mission critical workload.

The other thing that you need to know about using desktop hard drives in servers is that you shouldn’t try to mix match desktop grade hard drives with server grade hard drives. Consumer grade hard drives tend to use different timeout values than server grade hard drives. These differences can lead to various types of IO problems if desktop hard drives are mixed with consumer grade hard drives, especially in a RAID array.

Conclusion

As you can see, there is a lot to talk about when it comes to server storage. In Part 6, I am going to turn my attention to SAS drives and solid state drives.

If you would like to read the other parts in this article series please go to:

Advertisement

Featured Links