InformationWeek: The Business Value of Technology

InformationWeek: The Business Value of Technology
InformationWeek - Our New iPad App



InformationWeek Labs
Home
News
NewsFlash
News In Review
Financials
IW Community
AuthorITies
Shop Talk
Careers
Secret CIO
Columnists
Feedback
Business
Resource Center
Labs
Date Book

Services
Contact IW
IW Daily
Subscriptions
Media Kit/Ed Cal.
IW Marketplace
IW International
Site Map
July 7, 1997
TNG 2.0: Ahead By A Generation

Computer Associates takes the lead in the distributed systems management field


By Stephen V. Wright

T he career goal for most IT managers is a simple one: Keep the enterprise's operations running smoothly and continuously. They're happiest when the phones are quiet-because when their phones ring, the odds are that something is down, somewhere in the enterprise.

If they have responsibility for more than 1,000 desktops and servers, meeting that goal probably depends on one of these two distributed systems management (DSM) suites: Unicenter TNG from Computer Associates, or Tivoli TME 10 from IBM's Tivoli Systems unit.

With Unicenter TNG 2.0, CA has taken a commanding lead over rival Tivoli in almost all areas, including user interface, architecture, and functionality. Not only does Unicenter TNG provide all of the functions offered by TME 10 3.1, it also provides fully integrated applications for problem management, storage management, and workload management.

The outcome of our tests of Unicenter TNG in InformationWeek Labs is given here; a review of Tivoli TME 10 3.1-and the recently released Global Enterprise Manager (GEM)-will appear in a future issue. But based on our experiences with both products in daily use, there a re some obvious contrasts between them that bear noting.

As a former Tivoli employee who has spent five years as president of a consulting company that deploys Tivoli at large corporate sites, I was inclined to believe that TME 10 would come out ahead of Unicenter TNG on all fronts. But I must admit that Unicenter TNG is, in many respects, a full generation ahead of Tivoli's TME 10 in its architecture, functionality, and usability-three areas that are pivotal for producing results quickly.

It's important to stress that Unicenter TNG, while similar on the surface to TME 10, takes a completely different approach to solving systems-management problems. In this article, we'll focus on the core differences to that approach, in architecture, functionality, and ease of use.

Though often confused with the mainframe systems-management product called, simply, Unicenter, Unicenter TNG was originally written for Windows NT. CA has strategically moved TNG front and center during the last two years, continui ng to add applications through internal development and acquired technology.

The functionality set for Unicenter TNG encompasses not only Tivoli's list of functions, but also areas that TME doesn't yet cover completely, such as problem management (help desk), automated storage management (backup, restore, and archiving), a full-scale workload-management system (job scheduling), and Simple Network Management Protocol-compliant network management.

But more important than the quantity of applications is the way CA has completely integrated its applications, allowing modules to exchange messages in both directions. This integration should save clients time and money during deployment. CA provides integration kits for many third-party applications to help customers who have other applications installed. The company also offers a free development kit and training for integrating custom applications.

More Than Pretty Graphics
Unicenter TNG's designers must believe the old saying about the imp ortance of first impressions. With a knockout virtual-reality navigation interface that lets you "fly" though your network or peer into a faulty computer, CA's 3-D interface looks like an add-on level for your Duke Nukem PC game.

But it's not just a pretty picture; CA's Real World Interface lets you "look" inside a faulty PC (indicated by a red ball hovering over it) to verify the reported problem, such as a defective disk drive.

After zooming around the network until we'd had our fill of 3-D, we fired up the alternate 2-D user interface and constructed a graphical representation of our network using built-in geographic maps and vendor-specific icons. For the products of many hardware vendors, such as Cisco Systems, CA provides a class of device icons, complete with logos, that allowed us to build a realistic depiction of our network. Updated vendor icons and geographic information are provided on an update CD-ROM that CA provides quarterly.

In design mode, each device is automatically positioned anywhere in the United States simply by entering a telephone area code and prefix on its properties sheet. When we entered our phone exchange on the location property sheet, Unicenter TNG pinpointed our exact location.

Another nifty TNG feature is Nodeview, which lets you double-click on a node to closely examine a computer with a reported problem. You can display detailed SNMP performance information or create graphs of utilization statistics.

Unicenter TNG makes this possible because, unlike TME 10, it allows the flexibility of both polling and nonpolling operational modes at any time. This flexibility provides two benefits: In the nonpolling mode, network traffic is created only when a fault is detected, and in polling mode, administrators can examine a node at any time.

Unicenter TNG is so well integrated that we could easily instruct it to automatically open up a trouble ticket if a network hub fault was detected-without programming and without using third-party software. On the downside, we occasionally had to resort to the command prompt for actions the graphical user interface doesn't support, such as starting and stopping Unicenter. In addition, CA's user interfaces for the various applications could stand to have a more consistent look and feel.

In the event-management arena, Unicenter TNG lets administrators construct hierarchical Event Managers-that is, establish local or site event managers that receive events from local servers. Local or site managers apply rules to incoming events. The rules, which are constructed through the Event Manager GUI, can take actions such as restarting a stopped process or deleting unnecessary files from a disk.

If the local manager cannot resolve the condition, the events can be automatically passed on to the next event manager in a user-defined hierarchy. Round-the-clock operations staff typically attends this higher-level manager.

After receiving a forwarded event from a lower-level manager, the operations staff can page designated support st aff to investigate the problem or send corrective actions back to the local event manager to be carried out on the server in question. Such a two-way hierarchy of event managers can be a powerful tool, and Unicenter TNG makes this configuration easy to implement through the event manager interface-you simply click on dialog boxes and radio buttons.

Unicenter TNG really excels in software distribution from the server to the desktop. While TME uses a local client to push the software down to the desktop, Unicenter TNG uses a client pull model. An agent on a client queries a Unicenter TNG server when booting up, and at regularly scheduled intervals. The agent will automatically download software that is marked by software administrators as mandatory, and will also show the user on the client all available software packages and their distribution status.

Another nice feature of Unicenter TNG's software distributor is that administrators can mark a software package as mandatory for installation by a specif ic future date. Users can download the package earlier, but if they haven't downloaded it by the mandatory date, Unicenter TNG will force the distribution.

Room For Improvement
Unicenter TNG lags behind TME 10's Software Distribution application in server-to-server software distribution. Unlike TME, Unicenter TNG allows only a single level of distribution, from one server to a number of other servers. This single layer of fan-out means that Unicenter TNG can't have secondary servers distribute to subsets of servers; all distribution must take place from the main server.

Another weak point in CA's software distribution scheme is its lack of support for bandwidth throttling, which lets an administrator specify the maximum amount of a network's capacity that the software distribution should use. This prevents an administrator from accidentally taking all available network bandwidth. CA says the next release will use technology from NetCon Corp. in Crystal River, Fla., to provide both m ultiple-level fan-outs and bandwidth throttling.

Unicenter TNG comes with an excellent Concepts Guide and Getting Started manual. Most other documentation is distributed only on CD-ROM-not a wise policy for a product of this price and complexity. CA should take a tip from Tivoli and include a complete set of printed documentation.

Stephen V. Wright is president of Strategic Solutions Corp. Technical editing was provided by Sean Gallagher, senior technology editor at InformationWeek.


  Back to Labs

  Send Us Your Feedback

  Top of the Page

bottom navbar





Sign up for the InformationWeek Daily email newsletter

*Required field

Privacy Statement



This Week's Issue

Technology Whitepapers

Featured Reports







Video