Linux/Unix
The two vendors to watch in the Unix space in 2002 are IBM and Sun. When it comes to scalability, Sun rules. You can get a Sun box for less than $2,000 and scale it all the way up to the new Sun Fire Server with 106 processors, for only about $10 million. The critical point is that the application doesn't have to change anywhere along the way. Sun is obviously attempting to eat into IBM's mainframe market with the Sun Fire. At the same time, IBM is attempting to change its upgrade requirements with its big push--to the tune of $1 billion--into Linux. IBM also released a new 64 CPU platform. What's the significance? It's based on Intel processors yet isn't designed to run a Microsoft OS, only Linux.
Given the economic climate, you may be considering the cost advantages of Linux. The fact that IBM is pushing Linux so hard means it won't go away any time soon. Although you probably won't find application support comparable to what's available for Unix, Linux is well-equipped for Internet applications and its business-application availability continues to grow. Before deploying Linux, though, consider the skill sets in your organization. If you have Unix skills, they can be generalized to Linux without too much pain; if you have staff with no Unix experience, think twice. The low cost for the OS and the low cost of Intel hardware are attractive, but consider the cost of training and ramping up new skills.
Apple Computing Products
Watch for Apple to make inroads -- or at least try to -- into the corporate network. With the release of Mac OS X and more native applications, Apple is attempting to regain market share. The new system is a Unix derivative, similar to Linux. Unix and NeXT admins will like it, as this new OS supports a command line and remote-administrative capabilities. OS X even comes with an SMB client preinstalled for easier integration with Windows. There are a bunch of pretty interface differences (including transparent windows) that don't add to the system but may be easier to work with than X Window counterparts. Organizations migrating users from Microsoft Windows to a Unix platform may find Mac OS X easier to administer and train users on than Linux systems. The whole thing depends on how far Linux progresses user-interface design during the next few months. Apple's hardware prices have dropped from unreasonably high to somewhat affordable, and some of the components are now standard off-the-shelf PC-based hardware. But in spite of Apple's efforts, we can't advise any changes from a Windows environment. No matter how compelling the new features of OS X, there is no business case for adding support for an OS so lacking in business-application support.
Traffic Management and Shaping Bandwidth Claims
In recent years, traffic management has become synonymous with traffic shaping. Packeteer and Sitara Networks own the market, with reseller partnerships providing support on both the LAN and the WAN. What will be exciting in the coming year is a wave of products doing traffic management in a new way. Network Physics, netVMG, RouteScience, SockEye and other vendors are examining the issue of traffic management at the routing level and have promised to show some hot products in 2002. We've spent the past few months talking with the vendors in this space and almost all the technologies revolve around the ability to update BGP (Border Gateway Protocol) tables dynamically based on current network conditions, cost, link preferences and other factors.
Traffic shaping isn't dead, but if routing-based solutions can provide a better option for guaranteeing delivery, it'll certainly make the market rough. Radware provides products that make routing choices across links based on network conditions, and F5 Networks and others are close behind. These products provide a better ROI on provisioned bandwidth, but this is only a partial solution. While it makes service providers happy to provision all this additional bandwidth, the rallying cry of routing-based solutions vendors is that the additional bandwidth is unnecessary.
These products are still being tested in the real world and marketing departments haven't swung into full gear just yet, but watch for them in 2002. They aim to start a revolution in the traffic-management market and if their technology proves to treat the disease rather than the symptoms, enterprises will be clamoring for this newest cure to network woes.
Server Architectures
The server market is in flux. The recent announcement of a merger between Compaq and Hewlett-Packard has left many server admins confused, while Compaq and HP competitors are dancing with glee. In the short term, companies will continue buying what they've been buying. The lack of a clear and concise road and technology map for Compaq and HP customers is a challenge to IT workers: Do you stay with your tried and true HP or Compaq systems, or is this an opportunity to check out other vendors and other platforms? Recent challenges to Carly Fiorina's leadership have cast further doubt on the entire HP-Compaq deal. It all comes down to protecting your current investment while moving forward to fulfill your business needs.
A few things are happening in the server field. PCI-X will be a factor to consider in both the servers and the cards you purchase. PCI-X cards function in standard PCI slots, and standard PCI cards function in PCI-X capable machines. Just be aware that if you put a standard PCI card and a PCI-X card into a system, the bus will run at the older, slower PCI rate. Strategically buying PCI-X cards for current systems is a good idea, as they will be useful in future systems. On the other hand, if you're trying to get the performance benefits of PCI-X in your current systems, make sure all cards in the system are PCI-X-compliant or you won't get PCI-X performance benefits.
At the end of 2002, we'll begin seeing some mainstream implementations of Infiniband, so keep your eye on the Infiniband initiative, which will be a force in the next several years. However, we don't believe it will significantly affect your server purchases in 2002.
In the processor arena look for new offerings from Intel in the form of the code named "McKinley processor." This latest version of the Itanium processor is what Intel hopes to take mainstream. Previous versions of the Itanium processor were mostly for the development community. If you're looking to jump into the 64-bit computing arena, this is the place to be. New RISC offerings from IBM and Sun look to further squeeze the dwindling mainframe market. AMD has introduced its Athlon MP processor, bringing the power of its Athlon processor to the workstation and server market. AMD will release 64-bit chips in 2002, code-named "Hammer," which will provide 64-bit computing but without the need to recompile applications that Itanium requires. The upside is that 32-bit applications won't have to run in the handicapped emulation mode they do with Itanium; the downside is that there are no architecture changes that can provide potential performance benefits. For AMD to really be a power, it needs a Tier 1 company, such as IBM or Dell, to make machines for businesses with one of its processors or multiple processors. AMD has some business-class wins, but all are with Tier 2 and Tier 3 vendors.
The key words for servers are small and manageable. New boxes reflect the fact that companies are looking to pack more into less. Data-center space is expensive and needs to be conserved as much as possible. Smaller, faster, cheaper servers are the order of the day. The only time these slim servers don't fill the bill is when you need a lot of PCI slots or an internal RAID array. In all other situations, these can be the silver bullets for your data-center space needs.
Peter Morrissey is a full-time faculty member of Syracuse University's School of Information Studies, and a contributing editor and columnist for Network Computing. Send your comments on this article to him at ppmorris@syr.edu.
REPORTS
ANALYTICS
Follow 
