I Double Dare You To Try A New Approach
B
y Robert Moskowitz
Did you have a good year in 1996? Did you make great progress implementing network technologies to support the corporate computing systems? Do you now plan to evolve your networking world step by step? Well, I double dare you to sit back and relax. You sweated buckets last year, getting
management to buy into technologies that you spent months researching and testing. You came up with a plan, and now it seems that all systems should be go for '97. But is that really the case?
The "State" of the Internets
As our networks get more far-flung, they get more fragile. Users must put up with poor performance at best and broken applications all too frequently. A common culprit in this behavior has been the statefulness, or lack
thereof, in the application. Application state can best be described as memory of the client/server relationship. Too much state is costly in bandwidth, application complexity and network error recovery. Too little state is costly in bandwidth and application complexity.
This apparent dilemma clearly is manifested in many of the current networking applications. Novell NetWare and NetBIOS are heavily weighted with state. The constant state maintenance traffic brings little value to the actual use of these protocols. HTTP 1.0 is very light on state. There is little possibility to optimize data flows between clients and servers. Thus both, in the hands of the typical user and support personnel, continue to produce severe network degradation.
This leaves you with two alternatives for the coming year. You can deploy additional bandwidth and server capacity to meet unpredictable traffic patterns or you can rethink your a
pplication strategy. On a strictly cost basis, the bandwidth and server upgrade is the way to go; redeploying applications is front-loaded with costs that are rarely recovered. However, you will never get a handle on your networking
business this way. Only a serious rethinking about your complete architecture and its implications will let you set a course that will give you the competitive edge your management is seeking.
Internets in Flux
The Internet protocols have been a fertile ground for application creativity. This very creativity has resulted in a lot of work to select those applications that are worth their cost. The give and take of effective network utilization can have a greater impact on your bottom line than any of the user interface competitions. Since a performing application is a more effective tool than an intuitive, unreliable application, attention to the network impact is the real measure of best-of-class applications.
The current attention to applications for internetwo
rking in a company, otherwise misnamed as intranetworking, is actually a disservice for all of us. Many of these applications are not deployable for the extended enterprise. They just demonstrate more of the same narrow thinking that our traditional LAN applications exemplify. It also has slowed needed changes to old war horses such as printing. A frequent assumption in these new applications is that latency and congestion are not present, and reasonable levels of bandwidth are cheap. This sort of design and application selection criteria is bad for your business in the long run. It is too limiting for the types of processes that your management will wish to deploy.
Stale New Year's Resolutions
We've been at this networking business way too long. We seem to have forgotten some of the basic premises we started with, such as effective sharing of resources and always striving to better the lot for our users while streamlining our support requirements. Many of us have achieved nearly total saturatio
n of networking technologies over these 10 long, hard years of the "year of the LAN." We have arrived. But a horse and buggy will also get you places, and those blinders on the horse are very effective in keeping the horse calm about its pac
e on the road. Is a 10 percent improvement in server availability what networking is all about? Is Internet e-mail integration really going to fulfill your messaging needs?
It is painful to contemplate fundamental changes. The implementation teams are swamped with the standard fair of tasks. They will be consumed by the Windows95 rollout this year. New client/server applications that were started last year will be competing for installation resources with needed server upgrades. It seems that there is no slack for any sweeping changes. This begs the question of when the pace might slack off so that sweeping changes will be comfortable to contemplate.
Realistic Goals for the New Year
The Windows95 rollout is an interesting leverage point for many business
es. Windows95 has ended up being a lot more than a simple software upgrade; it requires hardware changes as well. More memory and faster processors and disks are a part of the upgrade. The additional disk space on these faster drives allow for our primary resolution: Reduce our dependency on file servers and their expensive network protocols.
This can be tackled in part. The Windows95 rollout itself is the perfect place to "run all common applications from the users' drive." There are three immediate benefits for this: Notebooks and desktops are treated the same. Network utilization will drop by getting Dynamic Link Library (DLL) traffic reloads off the network. And third, server usage will drop significantly and many of the server upgrades planned for this year will not be needed.
We might actually be able to significantly decrease our server codependency by shifting shared files to a more effective architecture like Lotus Notes or the Web. Finally, if we can tackle those department databases on the
file servers and move them to departmental Structured Query Language (SQL) servers, we will have done ourselves a truly great service. A move to an SQL engine philosophy for even the little departmental databases will simplify the inevitable cross-
department sharing requirements that always come up and afford the users better network performance, effective remote access and superior backup functionality.
My final challenge to you is to revamp your printing strategy. The printing vendors are actively working on fixing the current printing maze, and this will bare close scrutiny in the coming months.
So I double dare you to really get ready to ring in the new millennium with an approach to networking that will serve you well over the next set of years and not just be a refinement of our initial ideas from the dawn of the LAN.
Robert Moskowitz is a software systems specialist at Chrysler Corp., Detroit, Mich., and a member of the Internet Architecture Board (IAB). He can be reached at rgm@htt-c
onsult.com.
|
REPORTS
ANALYTICS
Follow 
