Buying a new server?
Apr 1, 2008 12:00 PM, BY BRAD GILMER
Servers have been used in the broadcast industry for a long time. From the earliest frame stores and audio workstations to the latest 3-D effects rendering farms, they play an important role. In most cases, the configuration of the server is set by the manufacturer. But increasingly, end users are purchasing and configuring their own servers as the capabilities of off-the-shelf equipment continue to improve. Here are some things you should consider when purchasing and installing your own server.
What is the purpose of the server?
This is the first question you should ask. Generally, I find two kinds of servers in broadcast facilities.
The first type is a server dedicated to a task. This server may be deployed as video/audio storage system or graphics rendering farm, or it may be dedicated to more general networking tasks such as e-mail gateway or a firewall.
The second type of server is a catch-all server, handling several tasks simultaneously. For example, it may be an e-mail server, a shared storage system and run a couple of databases. Whether a server is dedicated to a single task or it is performing several different tasks, it is important that you understand what the server is going to do so you can configure it properly.
Who will maintain the server?
Maintaining a server is sometimes difficult. Bear in mind that not everyone in a broadcast facility has the same depth of knowledge as the resident computer technology expert. What happens if that person is on vacation, sick or otherwise unavailable? What about the hardware? If there is a hardware failure, can the system be repaired using off-the-shelf parts available from the local computer store?
If the system can be down for a while, then there won't be a problem. If, on the other hand, the system needs to remain operational, you may want to consider a server with a warranty that includes software support and on-site hardware support. (Be sure that you understand how long it will take to get a technician on-site.)
What networking demands are put on the server?
Some servers in broadcast applications will need to handle heavy network traffic. Servers handling high-resolution video are clearly in this category. Servers that run centralized applications for several clients may also be in this category depending on how the application is written. Well-written centralized applications can reduce network traffic.
Figure 1. Before investing in a high-speed server to switch connectivity, be sure it
makes sense. Here, 400Mb/s is available between the server and switch, but only
100Mb/s is available between a different switch and workstations downstream.
Click image to enlarge
But before you order a high-performance network card or invest time and money in combining multiple Ethernet ports into a single, virtual network connection (called teaming), you should carefully study the overall network. Make sure that its infrastructure and your clients can make use of the network bandwidth provided from the server. As Figure 1 shows, it wouldn't make sense, for example, to team four 100Mb/s ports on your server and switch, when the connection between the switch and most of your clients flows through a single 100Mb/s port.
What are the storage requirements for the server?
I have been involved in many projects where I needed to specify the size of storage required for a centralized server. In the past, I would have invested a significant amount of time to understand the storage requirements of the server and to estimate future growth. Storage prices have dropped precipitously over the last few years — so precipitously in fact, that it is no longer worth the time to do detailed storage requirements studies. While I still encourage you to think about storage requirements, do not put too much thought into it. Storage is too inexpensive to justify the time it takes to accurately estimate your needs. Ultimately, many of these estimates end up being educated guesses anyway, and it is a relatively inexpensive proposition to add storage when needed or to replace the entire server with a new one having substantially more storage after a few years have passed. This does not mean that you should not think carefully about storage, as the next point illustrates.
How critical is the server to your operation?
If the server supports an on-air operation, then you should pay for options that will increase the robustness of the server. Relatively low-cost servers are now available with dual power supplies, RAID or other redundant storage mechanisms, and redundant cooling fans. All of these options will increase the chances of your new server remaining online after a component failure. Of course, you should have a solid backup plan for these servers as well. Depending on your operation, it may also be necessary to deploy remote monitoring and diagnostics in the servers. (What happens if one drive in the RAID fails, no one notices and then you lose a second drive?)
Figure 2. Commercial servers may include a diagnostics computer built into the
server itself. This computer monitors the health of the server and notifi es you of
any failures. Having a separate network and connection to the Internet can allow
you to access these diagnostics, even when your main network is down.
Click image to enlarge
In some implementations, an entirely separate diagnostics computer, with its own power supply, is mounted inside the server. This diagnostics computer monitors the health of the server and notifies you by e-mail or telephone if it detects a problem. Usually these cards also contain an on-board Web server through which you can monitor the server and perform basic operations such as rebooting the server remotely. Depending on the size of your operation, this may be an extremely beneficial tool. But configuring these systems can take some thinking, and designing the system so that the diagnostics computer is still available even in the event of a network failure may require that you install a completely separate network. As Figure 2 shows, this may be as simple as installing a small dedicated switch.
What operating system will you use on the server?
I know that for many of you, the topic of server operating systems (OS) borders on a religious discussion. For that reason, I am not going to get into any specific recommendations. That said, for most applications there are three choices: Microsoft, UNIX or Linux variants (*nixes), and Mac. The choice may be driven by server applications or manufacturers' requirements. But in many cases, the choice really comes down to what you are comfortable with.
If you know Microsoft products and this is your first server, then go with a Microsoft server. You can use a lot of what you know about desktop platforms. If you prefer *nix, I recommend Red Hat Enterprise Linux if you are new to servers and will need support. If you are comfortable in the *nix environment and want a full-featured, solid, free server, I recommend FreeBSD. If you are a Mac person, then a Mac server is the way to go. If you are used to working at the command prompt level, Apple's *nix OS makes a transition to Linux or FreeBSD relatively straightforward.
What is the amount of processing required for the server applications?
In today's computer market, manufacturers are always promoting their newest, fastest, most expensive server. But spending a few minutes thinking about your server and the applications it will support may save you hundreds (or thousands) of dollars by purchasing yesterday's technology. Only the most demanding applications require top-of-the-line processing power. Generally, these are 3-D rendering and compression platforms, and in almost all cases, the computer platform for these broadcast applications are carefully selected by the manufacturer. For almost all other applications, yesterday's processors — still blindingly fast — will work just fine.
Think about it. Will the server be used primarily as a storage platform? If so, it may have high I/O requirements. But unless all the video must run through the processor (which is not the case in most file servers), a high-speed processor isn't required.
How about streaming video on the Internet? Again, look at the whole picture, both from a processor and a network bandwidth standpoint. If you have a typical Internet connection, for example a T1 line, the connection is limited to 1.5Mb/s. Even with five T1 lines dedicated to streaming, only 7.5Mb/s will come out of the server to the Internet. At 750Kb/s per stream, you could source 10 simultaneous streams and still not exceed the capacity of even a single 10BASE-T network connection. Just about any commercial server manufactured in the last three years could support the processing requirements of this application. So think about the applications running on your server before spending a lot of money to get the fastest processor available.
Specifying, installing and maintaining a server in a broadcast facility is within the capabilities of most broadcast engineers. Just pay attention to details, and the installation will be successful.
Brad Gilmer is president of Gilmer & Associates, executive director of the Advanced Media Workflow Association and executive director of the Video Services Forum.
Send questions and comments to: brad.gilmer@penton.com
blog comments powered by Disqus
| Want to use this article? Click here for options! |  |
