Server Hardware Explained (Part 3)

by [Published on 13 Oct. 2011 / Last Updated on 13 Oct. 2011]

This article continues the discussion of server hardware by examining the management capabilities that are often built into servers.

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

Introduction

In my previous article I spent some time talking about server case designs. Now, I want to turn my attention to server management features.

Being able to effectively manage servers is a critical in any data center. As such, many server manufacturers build management features into the server’s hardware. These same vendors usually also supplement the hardware management capabilities with proprietary management software.

Because this article series is geared toward those who have limited experience with server hardware, I realize that the concept of hardware based management might seem a little strange. You might be wondering why organizations invest in products like System Center Operations Manager if the hardware already includes native management features.

The reason why organizations invest in management software even though management capabilities are built into the hardware is because servers are complex and management needs to be performed on many different levels. For example, a moment ago I mentioned System Center Operations Manager. In case you aren’t familiar with this product, it is a Microsoft utility that is designed for monitoring the server’s operating system for conditions that might indicate that there is a problem. When problems are detected, the software may attempt to automatically take corrective action, or it might alert the administrator, depending on the severity of the problem. Microsoft also offers add-ons that allow System Center Operations Manager to monitor the health of specific applications that are running on the server.

My point is that System Center Operations Manager can monitor the health of  a server’s operating system and possibly even the applications that are running on the server, but it does very little to monitor the health of the physical hardware. The monitoring software might report symptoms that result from underlying hardware issues, but the software will always take an operating system centric approach to server monitoring.

So with that in mind, I want to turn my attention to hardware level monitoring. As you have probably already figured out, monitoring capabilities that are integrated into the server hardware are used to monitor the hardware’s health.  However, there is more to it than that.

Software based server management products such as System Center Operations Manager and competing products usually have one very significant requirement. In order for these types of management products to work, the server that is being monitored has to be functional. Every management product is different, but normally this requirement means that the server must be bootable and that even if the operating system is having problems it must be running well enough to host a management agent so that the management software can communicate with the server.

When we talk about hardware based management, this requirement goes away. In fact, some forms of hardware management do not even require that the server be turned on. As long as the server is connected to a power source it can be managed at the hardware level.

Of course this statement raises a lot of questions. For instance you might be wondering how it is even possible to manage a server that isn’t turned on, and what types of things can be managed at the hardware level. Before I attempt to answer these questions, let me explain that when it comes to server management, there is no getting around the requirement for management software. As I said earlier however, there are several different types of management software. Hardware management requires management software that is able to interact with the server at the hardware level.

There are two main standards for hardware based management – IPMI and BMC. Most servers support one management standard or the other, but some servers support both standards. Since IPMI is arguably the more common of the two standards, I will focus my discussion around it.

With that said, servers that support IPMI include a dedicated management port. This port is simply a low bandwidth network port. In order to manage such a server, an administrator connects to the server through this port from another machine that is running IPMI compliant management software. Once the connection has been established, the management software is able to assess the server’s health by examining various status codes. The nice part is that because IPMI (and BMC) are industry standards, the status codes are universal. IPMI management software works with any IPMI compliant server regardless of the server’s manufacturer.

So if hardware level management is so great why isn’t it used as the sole method for managing servers within a datacenter? Well, although hardware level management is important, its capabilities are somewhat limited. IPMI is designed to report the status of sensors within the server. As such, IPMI is great for telling you things like the temperature of the CPU, the speed of the case fan, and whether or not the server’s case has been opened recently. In other words, IPMI is great for detecting low level problems, but it does not offer any advanced management capabilities.

Throughout this article, I have tried to stress the idea that there are multiple tiers on which servers need to be managed. I have talked about management software that focuses on the operating system and applications, and I have also talked about low level hardware monitoring. However, there is one more management tier that I want to mention.

Most of the better known server manufacturers offer their own hardware management tools. For example, HP offers a management utility called HP Systems Insight Manager and Dell has a similar utility called OpenManage. Unlike the IPMI management tools that I spoke of earlier, these tools are vendor specific. An IPMI based management tool can be used to manage any IPMI compliant server, but vendor specific management tools such as the ones mentioned above are proprietary and cannot be used for cross platform management. For instance you can’t use HP Systems Insight Manager to manage an IBM server.

So what do these proprietary management tools do? As I mentioned earlier, IPMI based tools are solely diagnostic. In contrast, the vendor specific tools are used to perform low level configuration tasks, such as setting up storage arrays. The actual capabilities of proprietary server management software vary from vendor to vendor, but generally speaking such utilities assist administrators in the initial deployment of network servers and alert them to any impending failures so that corrective action can be taken. Often times such tools also automate things like firmware and driver updates. Some of the server management utilities also assist with receiving technical support from the manufacturer. The utility might include remote assistance capabilities or even maintain a database of the warranty and service contract records for each server that the organization owns.

Conclusion

Being able to manage the servers within a data center is critical to an organization’s wellbeing. As such, most server manufacturers integrate hardware based management capabilities into the server’s system boards, and some manufacturers even offer supplemental management software.

In Part 4, I will turn my attention to server memory. As I do, I will be discussing memory fault tolerance and how server memory differs from PC memory.

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

Featured Links