DICE Meeting 2002-11-28
Conference Suite, BP, 1100-1300
Agenda
DICE Reorganisation
- Definitions
- Why reorganise
- Feedback
- Groups/Teams
- Issues
- What next
Minutes: Jeremy
The discussion ranged widely and between different areas but comments
have been minuted under the relevant agenda item.
Definitions
Due to ambiguity and different interpretation of what various phrases
meant there was an attempt to clarify the following definitions:
- DICE
Many people currently think of DICE as only being the commodity
desktop machines, rather than being some overall system.
The Distributed Informatics Computing Environment should perhaps more
correctly refer to all aspects of supported computing within
Informatics. This would include any LCFG-managed machines (commodity
boxes and servers), networks (connectivity), DNS, LDAP, backups, etc.
The principle is that even if there are self-managed machines on the
network they are still making use of services such as networking,
firewalling, DNS, etc. Whatever we finally decide, we need to come up
with terms that allows us to distinguish between the different aspects
and in particular will allow us a convenient and non-confusing way to
refer to the standard DICE commodity box.[ACTION: CEG]
- Line management
Under the new structure for DICE COs a line manager will have the
following responsibilities
- management of an individual's majority work, typically associated with
their 'home' group.
- their personal welfare and personnel related matters (appraisals,
etc)
- to resolve prioritisation matters with respect to other demands on a
CO's time
- High Quality Service
One of the overall aims of DICE is that once things have settled down
and migration from legacy is complete, there should be more resource
available for developing individual services. With running larger
scale services there will probably be more effort needed to make them
robust and reliable. However, it is also expected that services will
be developed to provide more features/facilities than was perhaps
possible under the legacy services due to lack of resource.
Why reorganise
With the initial rollout of DICE, which included basic infrastructure,
some services and the first wave of commodity machines, there are now
lots of Informatics-wide aspects to daily operations. The scale of
operation is much larger than before. We need effective ways of
cross-site working to provide:
- clear indication of who is running and/or responsible for what
- a set of escalation procedures for dealing with problems
- clear indication of how planning and development of DICE will be
done.
In particular we need a structure for the longer term future, given
that legacy operations will gradually dwindle.
Feedback
There has been a variety of feedback on the original proposals, via
email and informally in conversation. A number of issues have come up:
- uncertainty:
- there has been a degree of uncertainty about the impact of DICE
on each of the sites. This has been exacerbated by the need to deploy
DICE to such tight timescales which has left no time to clarify issues
and any misunderstandings. Some of the items discussed at the
meeting, particularly about support, helped clarify certain issues.
up.
- support:
-
There was a long discussion about support for users in general, and
in particular for non-DICE platforms.
- support model
we'll help anyone and we'll try to fix anything, within reason.
What we need to avoid is giving any single user a
disproportionate amount of time as it isn't scalable.
- supported hardware
The types of hardware supported by default is in transition as we
migrate from support policies at the three sites to a more common
one. Sites have existing supported architectures and will
continue to support them for the life of hardware that is already
purchased.
However, Computing Committee needs to make a formal policy
statement on what hardware is supported under DICE. This will
not preclude any particular hardware but will make
recommendations and indicate levels of support.
- transition from Solaris:
-
there is a significant support load related to moving users from
Solaris to DICE for the non-KB sites. This is both for getting
users migrated from their environment/setups that work under
legacy and also for getting required legacy functionality
available under DICE (eg, BP's
/corpora). This will
be a noticeable proportion of work in order to get remaining
users migrated by September 2003. CEG acknowledged that this was
important and would require effort to be allocated to it.
The recent meeting between COs and users at FH about migrating to
DICE had indicated that this aspect of the transition would be
greatly helped by access to local CO/CSO help in person rather
than just via the normal channel of the web form.
Issues
- line management/groups
-
The current proposal ties line management to groups. Concerns were
noted that this might present a difficulty for an individual if their
natural interests were in an area where they would get a different
line manager. At a time of reorganisation there is a natural
tendency to want to change as few things as possible. It
might even influence a CO's expression of interest as a way to not
changing line manager. It was also suggested that with COs still
split between sites it was perhaps better to have an individual's line
manager located at the same site. In practice there are several COs
not managed locally and it does not appear to have been a problem.
The concern was noted.
- site management
- Under the current proposals there is no officially designated role
that corresponds to the local site managers that have existed at each
site. The main reason for this is the concern that it would act as a
magnet for a wide variety of tasks that would be better being dealt
with under the relevant group or team. However, it was acknowledged
that this is a difficult area and will need further thought and consultation.
- database
- management and organisation of the
database is part of the Services group. It was inadvertently missed
out of the most recent draft proposals (posted to
cos@inf
22/11/2002). A full list of teams in fact has still to be
created---see below.
- role of Technology/Development group
-
The first draft proposals (posted to
cos@inf 29/10/2002)
mentioned that the Technology/Development group would tend not to have
a specific set of members but rather have people seconded as required.
The intention is that all COs get an equal opportunity to do
development work if they are interested rather restricting it to a
limited set of individuals. This idea was not mentioned in the
revised proposals so it was confirmed that this view still applies. A
comment noted was that consequently its operation is different from
other groups with respect to line management, since by definition it
will have very few, if any, permanent members.
- teams
- the current full list of teams has not
been finalised. It was agreed that this will be done as soon as
possible by CEG in order to facilitate the process of COs expressing
their interests.
[ACTION: CEG]
What next
- Timetable
- CEG are keen that a new structure is in place sooner rather than
later to reduce the uncertainty and provide the mechanisms for
managing DICE. The current goal is to the reorganisation decided
before the Xmas break so that the new structure takes from the start
of the New Year.
- Interests
- COs need to let CEG know what their interests are. Current line
managers will be discussing these with individuals over the next few
days. There is a meeting planned for Thursday 5 December where CEG
hope to collate the information coming back from COs and make a first
pass at a list of team members.
|
|
Please contact us with any
comments or corrections.
Unless explicitly stated otherwise, all material is
copyright The University of Edinburgh
|
|