CA-Unicenter: The Next Generation
By Bruce Boardman
Computer Associates International has released the latest version of CA-Unicenter TNG (The Next Generation), complete with a much-ballyhooed, 3-D GUI. Although the ability to "fly" around the network and even inside of machines makes for great trade show demos, the most important premise is CA's assertion that it offers the best
network management platform on the market--bar none. We decided to put that claim to the test. We brought TNG into the lab to see how well it would fare in the Syracuse University and our distributed labs. An Intergraph PC with a Glint video chipset displaye
d the show for us in 3-D.
CA is aiming TNG squarely at the likes of Hewlett-Packard Co.'s network management platform, OpenView, as well as at IBM's network and systems management framework, TME 10. According to CA, all of the necessary pieces--from calendaring to network monitoring, and from systems administration to an army of third-party developers--are part of TNG. Our quick look affirms this notion, and we can't wait to spend more time testing TNG in our labs.
Popcorn and 3-D Glasses
Although we knew that TNG's 3-D interface wasn't going to be all that useful, we still were a little disappointed with it. First, if you try this at home, be sure to get a track ball for navigation. Second, don't expect any magic in terms of layout.
TNG's autodiscovery lacks perfection--as do all network management platforms. Controls for limiting the scope and the required time for discovery let you specify address ranges, as well as either an Address Resolution Protocol (ARP) cache or a ping sweep. Su
bjectively, the 6,000+ node discovery seemed slightly faster than that of Cabletron Corp.'s SPECTRUM or HP's OpenView. However, we still spent one full day sweeping the entire network address range with pings. We ran a quick discovery test using just the ARP cach
e, and the basic layout wasn't too bad. The ping sweep, while filling in all the holes, overpopulated our network's top layer.
TNG has two basic levels of discovery. The first, like other network management platforms, discovers the existence of a network object. The second level is noteworthy because it adds the discovered object, filling the 3-D interior with actual device components. Fattening the discovered network object involves running a specially designed agent on the discovered network object. TNG ships with an agent for NT, and CA offers add-ons for various flavors of Unix. CA also includes a software developer's kit for building the 3-D interior of any network device.
We spent our testing time busily sorting out the resulting 3-D a
utodiscovered network map. The accuracy of the map, while average in our view, posed a new difficulty: placement of the 3-D objects on the desktop. We did the usual cut-and-paste map cleanup to accurately represent the network hierarchy and design, but TNG's 3-D interface was so heavily populated, that many objects were hidden behind a confusing mess of 3-D icons. Luckily for us, a good, old-fashioned 2-D interface let us make our changes and apply them to the 3-D interface.
Wherefore Art Thou, RMON?
We went out onto our discovered network, looking for modeled Remote Monitoring (RMON) objects (a favorite place to start picking at network management). Initially, we could not find any of the RMON-enabled hubs, switches or probes that we knew were on the network. We had discovered the network with different Simple Network Management Protocol (SNMP) community strings from those in these devices. Rather than go through the process of rediscovering the network, and to gain some experience in working w
ith TNG's database, we edited the description of two probes to gain RMON access.
Compared with most network management platforms we have worked with, changing the device with TNG was more difficult. This is because its object database structure is complex. As a resu
lt, we had to build an object that represented the probe correctly. Although not an incredibly difficult task, we required assistance from CA to familiarize ourselves with the object descriptions in the database. This is a fair indication that you will most likely need some training to use TNG.
The product's built-in correlation engine, which monitors systems and network events, is split into two levels of functionality. The first is the common approach of receiving a trap and performing an action based on that trap. The second approach offers a more complex correlation of multiple events across multiple network objects. CA refers to the rules that make up these multievent correlations as "policies," and TNG builds them using its Agent Fac
tory. The Agent Factory is available for NT and Unix, and it ships with predefined policies. The policies can correlate events across multiple agents similar to IBM's NetView. The difference in TNG's agents is their ability to monitor system and application events, as long as you use the Agent Factory application programming interface.
Bruce Boardman can be reached at bboardman@nwc.com.
|
REPORTS
ANALYTICS
Follow 
