F E A T U R E

The Survivor's Guide to 2001: Network & Systems Infrastructure

December 11, 2000
By Jim Hutchinson

If you consider yourself a network and systems infrastructure guru of any sort, chances are good you've subscribed to Popular Mechanics in the past, and at some point you've invested in at least one "how-to" type of book. Most infrastructure people enjoy knowing the inner workings of any type of technology. But the image of the infrastructure professional is starting to change.



Have you noticed that the network and systems infrastructure arena is getting something of a bad rap these days? Some people aren't paying as much attention to infrastructure technologies because they're looked down upon as boring and noninnovative. Sexy technologies like wireless and digital convergence attract the hype, but some folks forget that infrastructure is what makes it all work. The underlying technologies have reached a level of maturity such that management and, in some cases, IT professionals are beginning to take them for granted. Despite the bad rap, though, there's plenty of cool stuff happening in the infrastructure space.
Corporations continue to spend huge chunks of their IT budgets on the technologies that make up their network and systems infrastructure. Because this infrastructure is now perceived as crucial, expenditures on it are justified as, simply, the cost of doing business. This is a far cry from the early days, when getting a purchase order signed for a file server or router involved steps approaching a market analysis and committee vote. All those battles have been fought, so the challenge today is how to spend your infrastructure budget wisely, on what's needed the most.

What exactly is huddled under the infrastructure umbrella? The list is long and detailed: cabling systems and power management, which are the lifeblood of any network infrastructure and which, when implemented properly, let us dream about 99.999 percent uptime; directory and naming services, which range from the standard DNS (Domain Name System) to Microsoft Active Directory and Novell NDS; LAN/WAN technologies, which include everything from modems, switches and routers to 10-Gigabit Ethernet, OC-192 and DWDM (Dense Wavelength Divisional Multiplexing); and, finally, operating systems and servers, comprising an endless number of combinations--from single-processor systems to mainframe-class clustered configurations with 32 processors under the hood. Each category is worthy of its own "how-to" book, and the network and systems infrastructure professional must be well versed in all of them.

Cabling Systems and Power Management

Cabling systems represent the "passive" portion of the infrastructure, while power management drives the "active" parts -- servers, routers and other networking components.

When it comes to the cabling infrastructure, too many people think you can simply pay someone to do the dirty work -- the installation and support -- and then forget about it. But does that really make sense when you're spending anywhere from $100 to $400 per desktop? Those dollars add up quickly.

You need to have a solid understanding of the types of copper and fiber that are available and how they affect your ability to run your infrastructure. Seemingly minor issues such as pair-twist rules for Category 5 patch-panel installations and unaccounted fiber attenuation can stop a network project dead in its tracks unless the cable plant is carefully planned and certified after installation.

And the choices in copper cable are dizzying: More than 150 versions of CAT 5 cable are on the market, and some are considerably better than others. Do you know which flavor is snaking its way through the walls of your enterprises? Then, CAT 6 and CAT 7 cabling is added to the mix just to make things more interesting. Time to brush up on what crosstalk and return loss really mean.

But there's a bright spot in the confusing world of cabling: Many vendors are now offering 15- to 20-year warranties on their installations to help IT professionals "future-proof" their cable plants. You might pay a bit more up front for the installation than you would if you hired your buddy who does these jobs on the side, but you'll have a much better chance of getting a certified installer to come out to fix a problem with the CEO's drop.

The same holds true for your fiber plant--attention to detail during a fiber installation is critical. With SONET and gigabit-interface costs on network gear dropping lower every day, it's worth spending a few days to understand what it takes to certify your fiber so these services can run across it.

Once you have your "passive" infrastructure in place, you'll start plugging in all those "active" gizmos. And unless you're buying solar-powered routers, good old AC and sometimes DC will play a significant role. But the network and systems infrastructure person must plan ahead for the inevitable--a power failure. This is where UPS vendors save the day.

IBM once ran a study that showed that a typical computer is subject to more than 120 power problems each month. If the thought of a $2,000 desktop system frying doesn't raise an eyebrow, how about a $100,000 backbone switch? Power concerns should be part of any IT buying decision, and certain equipment options must be considered; all the major UPS vendors have added features that make their products much more than just big batteries. These features include real-time UPS monitoring and alarming via the Web; environmental monitoring of temperature, humidity and other variables; power-outlet management, allowing remote devices to be power-cycled from anywhere; and logical server shutdown, which kicks in when all else fails.

Directory and Naming Services

There's no lack of passion when it comes to talking about directory services. The Microsoft Active Directory-Novell NDS holy wars have been in full swing for more than a year now, with new domains being added to the Internet's DNS at the same time. How it all shakes out will be interesting to see.

Let's start off with the familiar DNS. The Internet Software Consortium recently released version 9.0.1 of BIND (Berkeley Internet Name Domain), which is the basis for DNS as we know it. Enhancements are continually being made to keep pace with the changes in the Internet community; for example, the latest release adds key security features and support for IPv6. But applications for managing DNS namespace abound, and are usually offered as packages with features such as IP address management and DHCP/DDNS integration.

Most of the recent DNS controversy has centered on the addition of new domain identifiers. The widely used .com extension will have some competition from newcomers such as .biz, but only time will tell if these new extensions will gain acceptance. In any case, the DNS will continue to be the bread and butter of Internet naming services for years to come.

Now on to the holy wars. There's no shortage of opinions as to whether Microsoft Active Directory or Novell NDS is a better solution for the enterprise and which one ultimately will be the market-share victor. A directory service is a fundamental component of the network and systems infrastructure; it's completely integrated into how a corporation exchanges information, and careful planning is essential to ensure a successful implementation.

Software vendors and ASPs (application service providers) are lining up to help enterprises migrate their existing infrastructures to an Active Directory or NDS solution. IT professionals will need copious information in order to make a platform decision, and to determine the cost benefit of implementing a directory conversion that will support all aspects of the business.

LAN/WAN Technologies

Once upon a time, a small company named Cisco Systems gained a competitive advantage over just about everyone else in the LAN/ WAN arena by selling large, complex boxes called routers, and the company leveraged that advantage into the king-of-the-hill status it enjoys today.

However, Cisco has been kept on its toes by other vendors that can do things faster, cheaper and sometimes better. To combat these competitors, Cisco has taken a novel approach: It's bought them, to the tune of almost 70 acquisitions since 1993. And you can expect this pattern to continue.

But the LAN/WAN space is much more than Cisco. This is where technology is constantly being pushed to its limits--where the percentage of revenue spent on research and development is considered a competitive advantage. Here, even marketing people converse in terms like "aggregate throughput" and "scalable architecture." Face it -- this is where the geeks hang out, and where networking happens in its rawest form.

If you follow this market, you know that vendors have been trying to differentiate themselves by moving up the network stack. For example, Layer 2 switches now operate with routerlike features, such as support for multicast data streams and QoS. Also, Layer 3 switches and routers are entering the Layer 7-ish, or content, realm, where they're making traffic decisions based on information within a specific application protocol, such as URL or HTML metatags. One side benefit of all this is that prices per port continue to go decrease.

The push for higher network throughput is keeping vendors busy. For instance, just as the Gigabit Ethernet standard has started to gain market share, the industry is already pushing for the new 10-Gigabit Ethernet standard (www.10gea. org). In ATM, too, demand for higher raw interface speeds is taking precedence over features and functions. Having an OC-48 link installed in your backbone was all the rage a short time ago, but that's been replaced with demand for OC-192 products. And once we start burning through all this extra bandwidth, new interface speeds will be defined and in demand.

Operating Systems, Servers

The cross-platform support between hardware and operating system vendors has never been better than it is today. You can go out and buy just about any flavor of an Intel-based server and install your operating system of choice. Although the same can't be said for all RISC-based platforms (Microsoft makes sure of that), Linux has come along as the great equalizer. Even IBM seems to agree, based on its new support for Linux on its computing platforms.

On the server front, the Intel-versus-RISC debate has intensified with the push to 64-bit processors. Intel plans to ship its version of the 64-bit CPU, known as the Itanium, early next year. Sun Microsystems already ships the superscalar SPARC version 9. And Compaq has the venerable Alpha Tru64 Unix product line.

These next-generation processors already support large parallel-processing configurations or they are poised to add that support in the future. Databases, ERP and other compute-intensive applications will scream on these platforms. Also, new internal I/O architectures, such as infiniband and PCIX, will move data through these servers at rates that could flood almost any network infrastructure. IT professionals will need to have their existing network and systems infrastructure ready when these "supercomputers" start rolling through the door.

The OS space has been just as active. With the Linux hype fading, real-world testimonials are starting to emerge. Linux's performance and cross-platform support for nearly any hardware architecture make it an attractive choice in business applications--an area in which only Microsoft and Sun are accustomed to competing. Microsoft's Windows 2000 hasn't taken off as expected, but as with Linux, the hype is disappearing, and the OS is proving itself to be a stable choice for desktops and servers. Time will tell how well it will do competing in the data center against vendors such as HP and Sun.



Research and Reports

Storage Virtualization Guide
May 2012

Network Computing: May 2012

TechWeb Careers