Heavy Metal

There has been a lot of recent speculation (some of it public) on the future of IBM's 'strategic' product, OfficeVision. I've been taking some behind the scenes looks at what is really going on.

This is a new column, and it's the only place where the facts behind the news are going to be discussed. I'm going to concentrate on IBM mainframes, but large Unix systems will be appearing, too. My background has dwelt on data base systems and performance; at the moment, a lot of what I am doing is focused on VTAM efficiency and productivity. I am a featured speaker at the GUIDE European Spring Conference, presenting papers on SAA and new advances in network security; I hope that some of you will be there. In my next column I'll pass on some highlights of the most interesting papers.

OfficeVision was announced in May 1989, as IBM's Office Automation product, combining earlier systems such as PROFS and DISOSS as one package for all office workers to learn, using the SAA (Systems Application Architecture) standard announced only a short while before by IBM.

OfficeVision is vitally important to IBM, being the only SAA/CA (Common Applications) package yet available from IBM. These are all of the actual applications that conform to IBM's SAA standard. The release of OfficeVision 2.0, which was expected in September 1990, has been substantially delayed, which is unheard of for any product that IBM labels 'strategic'.

The original version of OfficeVision was not much different from the earlier products that IBM pulled together to make this new application. This first release of OfficeVision was not making extensive use of SAA, and didn't present any new challenges to the system. It is this new release 2.0 that fully implements SAA, making use of the SAA/CUA (Common User Access) interface, and with SAA/CCS (Common Communications Support) co-operative processing between MVS hosts and OS/2 PCs.

Performance problems with OfficeVision 2 result from IBM's use of session managers under MVS. In fact, it may be impossible to implement as IBM intended, due to the high overhead of Netview/Access when managing the multiple sessions required by OfficeVision. The user community are saying (as with any IBM 'strategic' product) that they might not install OfficeVision now, but they will do nothing to prevent putting it in _when_ they need it. IBM says that you need to have NetView/Access (the NetView session manager) for OfficeVision. This isn't true, even according to IBM.

IBM have promised specifications for the OfficeVision 2 session manager interface to third parties; but so far, they haven't delivered. This has meant problems for suppliers who want to support OfficeVision; and who could solve IBM's problems with performance.

Netview/Access is a very basic session manager. It provides minimal facilities; just session switching, and little of the management of sessions that third party products provide. It also has possibly the highest overhead, in terms of CPU cycles and virtual storage required. For use with OfficeVision, the restricted facilities are not a problem; however, the overhead is.

OfficeVision 2 beta sites are saying that you need OS/2 Extended Edition for it, which means running on real IBM PS/2s with MCA (OS/2 EE is only available for genuine PS/2s). The problem arises from the use of multiple sessions to monitor mail arrival. The system overhead is just like an extra five users constantly typing away for every OfficeVision user that you will have. This means overhead; increased CPU, memory storage and channel capacity.

If you want to have all users in an enterprise running OfficeVision, you are talking of hundreds and thousands of users for most sites. However, it is widely recognised that the practical limit of users with multiple sessions products is in the dozens to hundreds range. Most organisations with a session manager don't have more than twenty users who are privileged to use it.

The delays on release are unusual from IBM; we hear it's for performance reasons. With the above story, this is not surprising.

IBM can't afford to abandon OfficeVision; it must work, because it is the only IBM SAA/CA (common application). This provides an SAA application that everyone must learn. Learning the next SAA application is cheap. The learning curve to transfer to applications based on DEC or on PC LANs will then be substantial.