Last November some of the largest OpenView users got together and drew up a shopping list of stuff they wanted Hewlett-Packard Co to implement in its network management environment. Hewlett-Packard’s OpenView programme manager, Gordon MacKinney, says the company will begin delivering the first of those requirements next June or July with the release of OpenView 4.0. On top of the user demands list is the ability for OpenView to support and manage a much larger number of sites, or nodes, than it is capable of at the moment. The problem, as described by MacKinney, is that even with polling rates set to minimum time frequencies, say once every half an hour, a central management site is generally able to cope with network information generated by only a maximum of 4,000 or 5,000 nodes. Some users have networks with many more nodes that need managing. Next summer, Hewlett-Packard will begin offering a distributed version of its network node manager so that the collection of network information can be off-loaded to satellite sites, freeing resources at the central site; it will then be possible to distribute status polling, discovery and event forwarding to the satellite OpenView nodes, each of which will manage a network region and interact with a central station.
Information feed
The distributed managers will feed information back into a common database of network information, Hewlett-Packard’s Meta-Schema data repository, which is also due next summer, at a central site, which will synchronise collection operation and enable different network applications to share a common pool of data. This would mean that applications can be more closely integrated. MacKinney believes central site resources saved by farming out much of the hard work of network information collection, should enable tens of thousands of network nodes to be supported in a single management topology; he claims it will also filter out as much as 99% of the ‘junk traffic’. Hewlett-Packard also plans to offer a mechanism for distributing OpenView’s user interface processes so that operators in different sites can access the same network displays to monitor and review status information. Other planned enhancements include the ability to off-load graphical user interface processing from the system, making it easier to support multiple system administrators and cutting the amount of network traffic. Other features will include support for SNMP version 2 and Oracle. Users, meanwhile, are calling for other improvements, such as better automatic detection, recognition and mapping of networks that are added to a system, report generation and Japanese language support. It is now up under HP-UX, Solaris, AT&T Corp boxes and a range of other systems, and Hewlett-Packard is promising support for other Unixes down the road: RS/6000 and Digital Equipment Corp versions would make sense, plus the Windows NT OpenView implementation expected by end of next year.
