UK Post Office Report on using Access Gen

Management Summary

Response Systems' Access Gen product is a tool designed to provide browser capabilities to Computer Associates' CA Gen Central Encyclopedia (CE). It claims to provide substantial savings in terms of the time and effort often taken in encyclopedia reporting and model file transferring by providing information regarding models whilst they are still in the encyclopedia. In addition, there are facilities for impact analysis and problem diagnosis and resolution. A lot of the feedback received also praised the additional functionality that Access Gen provides over the CE toolset. These additional functions helped make the development teams' jobs a lot easier. This is not quantifiable in terms of possible savings but will nonetheless be of great benefit to project teams. It will also prove a useful product to aid in Component Based Development, as a number of the functions it provides are especially beneficial for these methods.

Individual Projects

Each project provided feedback and comments on the use of the Access Gen tool. Below is a summary of each projects' feedback.

Project 1 (Block mode)

The project team use CA Gen to maintain and enhance a very large application. This system is implemented on the OS/390 mainframe and uses 3270 block mode clients. The feedback from the team was very positive. They are using Access Gen, on average, 4 to 5 times per day. They have been principally using Access Gen for the following purposes:

Navigating through processes in order to see where processes are being used for impact analysis purposes. They feel this functionality is superior to the existing CE cross reference printing capability and they find that Access Gen even provides some facilities that are not available even when a CA Gen model has been file transferred to a workstation and is accessed using the toolset. They estimate that this saves at least 10-20 minutes per investigation.

Finding entity types from database table numbers and vice versa, and displaying entity type and attribute information. They estimate that this saves 5-10 minutes against accessing the same information on the CE directly or a lot more in file transferring the model to the workstation and accessing it there.

Printing action blocks. They find it quick, easy and straightforward to print action blocks, which actually fit the paper and come with views already expanded. They estimate that this saves at least 3-4 minutes per print especially if needing to print several action blocks. Impact analysis to find out which subset an object is in. They estimate that the time saved could be anything from 30 minutes to an hour depending on the subset size.

Seeing versions of objects across models and thus being able to prove that an object has not changed, or if it has, to be able to see exactly what has changed, when and who the change has been made by. This is particularly useful for a model manager. This can be achieved within a few seconds rather than hours.

They view the facility to detail view sizes particularly helpful to see if views exceed the 32K limit.

The team feels that in model management terms Access Gen is already saving them at least two hours each week, plus an hour or so file transferring. It is also enabling them to perform checks that they would not previously have been able to do.

Project 2 (Block mode)

This system is also implemented on the OS/390 mainframe and uses 3270 block mode clients. The feedback from the Project team was also very positive. They did not use it as much as the Project 1 team because there is not a great deal of active development required on the system at present. However, they found Access Gen essential as a tool for every day looking up of bits of information and described it as the only tool available for using CA Gen to its fullest. They estimate that it probably saves each team member at least an hour each week but that this figure would rise if they were to do more active development work on this system.

Project 3 (Client / Server)

This is a new system currently being developed. It will be implemented on NT/Windows 95 over a DCE middleware architecture. The team have found Access Gen to be a useful tool during the development of this new client-server application. They have been principally using Access Gen for the following purposes:

• They found the ability to find out in which subsets and in which modes an object is checked out of particular use.
• Quickly generating lists of worksets and action blocks, and being able to view their descriptions, has been beneficial in preventing work being duplicated for example by being able to find out if a workset or common action block has already been created that covers functionality that another developer needs to implement.
• Browsing of action diagrams also has been of some use but they expect this feature to be really useful in established application systems, especially block-mode ones.
• The team estimates that their time saving is 20% to 30% in terms of file transfers (thus creating a saving in mainframe resources as well as person time), plus the quick-to-look-up advantage, which they found difficult to quantify because "the temptation is to not bother when the easy way of doing it isn't there". They estimate that overall they would save an hour per person per week when at the peak of development.

Project 4 (Block mode)

This system is also currently implemented on the OS/390 mainframe and uses 3270 block mode clients. The feedback from the team was again very positive. They have been principally using Access Gen for the following purposes:

• Where Used Information, saving approximately 30 minutes each report. Browsing/Printing action diagrams. This now takes 90 seconds. They estimate this saves around 20 minutes a time as well as providing a better report.
• Impact analysis, they feel that this has led to more successful migrations, and has cut down quite dramatically on the number of 'extra' migrations where previously they had to perform another migration because an object had been missed..
• Reporting - they also found that reports are far easier to find compared with using the CA Gen toolset on the developer's workstation.
• They estimate that each team member saves somewhere between 45 and 180 minutes a week using this tool as well as an approximate saving of 25 calls to the CE each week.
• They also feel that the possible savings made by a team at the height of development would be considerably greater as Access Gen would be used for additional functions.

Project 5 (Client / Server)

This is a large application incorporating new processes and rewrite or interface of all existing Business Systems. It will be implemented on NT/Windows 95/HP-UNIX over a DCE middleware architecture. The team has found Access Gen to be a useful tool during the development of this new client-server application. They have been principally using Access Gen for:

• Subset Checkout: Check who has subsets out and thus avoid contention and make sure they will get the right access. They estimate that with Access Gen the task is now taking 20 minutes instead of the previous average of 40 minutes.
• Navigation around the model, which was a lot quicker than using the encyclopedia directly.
• Debugging: Referring to action blocks written by other team members.

Project 6 (Block mode)

This system is also implemented on the OS/390 mainframe and uses 3270 block mode clients. The project team used Access Gen and found it to be very useful for:

• Impact analysis.
• Reports, where they found it very user friendly.

Mainframe And CPU Resources

During the evaluation of Access Gen at the Post Office it was important to ascertain that there were no detrimental effects to the mainframe environment in which it was installed. It was felt that the best way to assess impact was to compare the charges for process, CPU and i/o usage of the user IDs of some of those people partaking in the cost benefit analysis exercise before, during and after Access Gen was available. The statistics showed that in general there was either little change or a reduction in CPU and DB2 i/o usage for April and May (the two months when Access Gen was available). Three key user IDs involved did show a considerable reduction. It has been confirmed that these people's work profiles changed very little over the period January to June 1998 whilst they were in the development phases of their project lifecycles. This seems to indicate that the use of Access Gen during April and may contributed significantly to the reduction in mainframe resources used.