1996 VM Workshop Requirements Voting

During the 1996 VM Workshop, the attendees engaged in several spirited discussions about the current "state of VM" and where it is heading. It became clear that there was a need for us to express our ideas to IBM using a formal requirements process as well as the usual casual discussion. Therefore, it was decided that the VM Workshop would write, vote on and submit requirements to IBM, like a normal user group, such as SHARE.

The requirements from the 1996 VM Workshop were drafted by an author, transcribed by the agenda organizer and discussed at a open, requirements session. Votes were then taken with the attendees who were still in attendence. Since many attendees had left by Saturday morning, the voting for these requirements will remain open for 2 weeks. Voting will close on 7/16/96. At that time, the results will be compiled and the requirements will be sent to IBM. The order in which these requirements are listed is the order in which they were received and does not imply any priority or order of importance. Note: the voting is only open to those who attended the 1996 VM Workshop.

Send your vote to Martha McConaghy URMM@VM.MARIST.EDU. If you feel the issue is critical to your installation, then specify that you want your name and company attached to the requirement. We feel that this gives added weight to the requirement when IBM is considering them.

Voting Ground Rules

  1. Each attendee can only vote for each requirement once. However, attendees from the same company can vote individually.
  2. The votes have the following meaning:

Requirements Voting is now CLOSED. These are the final results as of 7/17/96, which will be forwarded onto IBM.

Requirement 1 - Backup/Restore of VSAM Files for VM

Description of problem:

We are running FOCUS to generate on demand reports during our normal work day. To facilitate this without interference with our on-line CICS functions, a seperate VSAM catalog has been developed. The refresh of the data in the secondary catalog is a daily process that runs unattended at night and runs under VM in several service machines. While this process is stable and gives good performance, it was discovered that it would be avantageous to have the option of loading this catalog from a backup tape, which can be accomplished by running a IDCAMS RESTORE on a VSE system. This process is working and has been tested, but is cumbersome and confusing for operations staff.

Requirement:

If we could have the BACKUP/RESTORE feature in the VM implementation of VSAM, it would ease our confusion, provide single source of VSAM maintenance routines and, with REXX, provide better control of the refresh process.

Submitted by Sam Whiteman, Indiana-Purdue University, Fort Wayne

Voting:
YES - 2
NICE - 34
NO - 2

Requirement 2 - Nameserver Compatibility with non-VM Nameservers

Description of Problem:

During the past 2 years, the TCP/IP network at Marist College has increased in importance until now, it is central to its ability to function as an organization. Our VM/ESA system is the central focus of this network, handling millions of pieces of e-mail as well as suport our mission critical web pages. The quality of TCP/IP's nameserver support, however, has become a serious problem causing major network outages and inability to deliver e-mail. In particular, it has great difficulty in handling various types of records from UNIX based nameservers, including our Internet provider, SPRINT. It also has had many problems handling variations in MX record definitions commonly used by large providers such as America On-Line. It appears that the VM nameserver follows the RFC rules "to the letter", but does not properly handle variations which have become generally accepted practices on the Internet.

For example, we have experienced a problem for more than 1 year, where the VM nameserver reports that it is receiving "bad packets" from our authoritative nameservers, which are run by SPRINT. Apparently, the format of the packets do not match what VM's nameserver expects. However, SPRINT argues that they are using wellknown UNIX based nameserver software and that only VM nameservers have a problem with it. The problem impacts our nameserver performance as well causing the console to fill up with these messages.

Requirement:

We need a cache-only nameserver on VM which is compatible with the current and future UNIX servers on the Internet, and which handles variations on RFC definitions properly. We feel that a cache-only server on VM is very important to ensuring better performance on our VM based SMTP and web servers. This nameserver could remain CMS based, but could also be a port from a POSIX based system.

Submitted by Martha McConaghy, Marist College

Voting:
YES - 20
NICE - 35
NO - 0

This issue is also critical to the following attendees:

Paul Lewis, University of Louisville

Carl Forde, Beyond Software

Larry James, Southwestern Bell

Anne-Marie Marcoux, McGill University

Requirement 3 - VM Support for Synchronizing System Time with External Sources

Requirement:

Representation of dates and time of day presented to virtual machines via Query Time, Diag 0C, Diag 270 and REXX date and time functions should be consistent with one or more external reference sources.

CP's internal representation of time and its use of the TOD clock is not at issue here. The issue is the time presented to users, programs running virtual machines and stored by file systems.

It may be useful for a virtual machine to be able to determine the adjustment used for synchronization. This is similar to the kind of information supplied by Diag 00 for the GMT offset. A configuration file option for a system default and a class G command to combine the synchronization adjustment to the GMT offset returned by DIAG 00 would be useful. Additionally, there may be some instances when it would be appropriate for CP itself to use the date and time values it is presenting to a virtual machine, rather than it's own internal representation of the time. A few that come to mind are, the determination of when to issue the midnight message, the timestamp value associated with a spool file and the timestamp applied to messages written to the primary operator's console.

One acceptable implementation of this requirement is to provide a diagnose interface that allows a suitably privileged virtual machine to obtain and set CP's current synchronization adjustment. CP merely needs to apply this adjustment at the appropriate times.

Submitted by Perry Ruiter, Government of British Columbia

Voting:
YES - 35
NICE - 10
NO - 1

This issue is also critical to the following attendees:

Paul Lewis, University of Louisville

Carl Forde, Beyond Software

Larry James, Southwestern Bell

Dick Waite, Software AG

Requirement 4 - ISLINK Connection between 1st and 2nd Level Guests

Requirement:

Allow for connection of 1st level CP to a 2nd level VM guest via ISLINK, either via support for a VCTCA from 1st and 2nd level, or by somehow enhancing the ISLINK command to 'point' directly to a logged on VM userid (theoretically running VM/ESA). Currently, 2 ESCAN CHPIDS must be wasted in order to get up a real connection from 1st to 2nd level.

Submitted by Byron Graham, Kimbal International

Voting:
YES - 7
NICE - 29
NO - 4

Requirement 5 - Support Remote Messaging

Requirement:

Enhance CP commands MSG, MSGNOH and WARNING to communicate across an ISLINK or AVS connection to a userid on a remote system. Example syntax:

MSG userid AT rem-system text............

or

MSG node(userid) text...........

Second syntax is preferable to avoid problems when sending local messages beginning with 'AT'.

This would greatly enhance the inter-system communications and open the door to many new capabilities and automation via service machines.

Submitted by Byron Graham, Kimball International

Voting:
YES - 2
NICE - 13
NO - 23

Due to the majority NO vote on this requirement, I will not be submitting it to IBM unless the YES/NICE vote changes to out number the NO's.

Requirement 6 - ADSM Management of Deliquent Filespaces

Description of Problem:

One of the characteristics of ADSM is that it uses the label on a disk drive as the name of the node's filespace within the server. When the label is changed, it causes a new filespace to be created. This has two effects:

  1. It causes a new full backup to be done on the disk, even when it is not necessary.
  2. It leaves a filespace which is no longer used for backups, but takes up space in the filepool and in the database.

Upgrades to PC systems software, such as OS/2, frequently cause drives labels to change. Also, this is often done by LAN administrators, unaware of its effect on ADSM. This leaves the server with many unnecessary filespaces containing useless files, taking up much needed space. Currently, monitoring this is a manual task for the administrator and there is no ability to automatically remove 'delinquent' filespaces. This causes excessive use of storage rosources.

I would like the ability to monitor delinquent filespaces based on customizable criteria. This criteria probably should NOT be based on the "last successful backup" value since selective backups do not update this value. For example, I would like to monitor any filespace which has not been accessed in more than 8 months and determine if it is a 'delinquent' and a candidate for deletion.

Submitted by Martha McConaghy, Marist College

Voting:
YES - 11
NICE - 27
NO - 0

The number of voting attendees who are running ADSM is 9.

This is also a critical issue to the following attendees:

Wendy Alberg, Cornell University

Paul Lewis, University of Louisville

Larry James, Southwestern Bell

Postition Statement:

The purpose of this statement is to express the opinion of the attendees of the 1996 VM Workshop on an issue which is not strictly a technical one. Therefore, it was felt that this should be reflected in a position statement rather than a requirement.

For the past few years, IBM has promoted VM as a premier client-server operating system and this vision appears to be coming true. However, during this time, IBM's committment to a strong TCP/IP product on the VM platform has waned and the product has developed numerous weaknesses and overt problems. As of this statement, there are in excess of 50 formal and informal requirements against the product. This situation now threatens VM's ability to fulfill its role as a full participant in the client-server world. In many shops, it also threatens VM's and the IBM mainframe's viability.

The attendees at the 1996 VM Workshop wish to express their deep concern over this situation and their support for the continued development of the TCP/IP product on VM. It is the foundation upon which the future of VM will be built and it must be robust and responsive to the changing demands on the TCP/IP environment. We understand that development has begun on numerous improvements to the product and we fully support this. We encourage IBM to continue this effort and to implement as many of the requested improvements as possible. We also appreciate the effort of the product developers to communicate with the customers and listen to their needs.

We would like to see IBM issue a statement of direction for the TCP/IP product on VM so that its committment will be clear to all customers.

Voting:
YES - 44
NICE - 2
NO - 0

This issue is critical to the following attendees:

Martha McConaghy, Marist College

Carl Forde, Beyond Software

Val Christensen, SAS Institute

Larry James, Southwestern Bell

Anne-Marie Marcoux, McGill University

Betsie Spann, Oracle Corporation

Dave Jones, Velocity Software