X.500 DSA requirements- uptime, liaison and long term.

[Russ Wright <[e-mail]>]

At 1:36 AM 9/20/94, Colin Robbins wrote:
...
>Looking at some of the other requirements:
>
>   * DSAs must be available at least 75% of the time.
>
>Are you serious?   For a production system at the very least 95% is a
>requirement.  The pilot in the UK has shown the above 95% percent is
>certainly obtainable.  When we started measuring, the figure were
>about 70%, but after a few weeks of publishing a league table of
>availability, a majority of DSAs are running at 99%.
>
>Assume you have a DIT with 100 organisations, if each DSA is only
>available 75% of the time, the implication is, on average, 25 of the
>100 organisations are unavailable - that is a pilot service, not
>production.
>
>I strongly recommend you set the target higher.  FYI, the service
>level target for the GB DSA is 98%, and has so far been acheived
>without major problem.

The basic question is whether you allow non-production quality DSAs into
your service.  We talked about putting the requirement up to 90% for all
non-country level DSAs, but backed down.  Maybe we shouldn't.

I'd be interested to hear from others-  Is 90% reasonable (for all DSAs
other than the master DSA, which is >95%)?  Do we want/need other folks
that don't have as good uptime?


>Don't get me wrong, I support the initiative, but as Peter Whittaker
>says, you need to look longer term, and consider what a production
>service really means.

I agree, someone has to look at the longer term.  I am hoping that our
requirements document will start the process of looking at more long term
goals and requirements.

I would to focus on more reasonable short term requirements that are easier
to justify.  We have immediate problems in the US with the level of quality
of the master DSA and many of the organizational DSAs.


>On a slightly different note, PARADISE has been trying to transition
>from pilot to production service for about two years now, it is not an
>easy road.  One of the next steps for PARADISE is to produce a set of
>requirements for a the future service.  I think it would be useful
>liaison is one or more of the authors of this ID, could attend the
>DANTE PARADISE meeting (date to be confirmed, but sometime in
>November) in Europe to participate in the discussion, then there could
>be some alignment of requirements.  (I seem to recall somebody in
>Australia has asked if they can also attend, maybe they could also
>bring some input).

We did try to bring in one of the DANTE people but after the initial
meeting, I haven't heard anything from her.  We would like to align the two
documents.  I don't have the funds to travel to Europe, but maybe Tim or
Peter was planning to go?


>Once the service level has been defined, (assuming it is X500 '93 based) a
>transition strategy from the QUIPU based pilot to a '93 service has to
>be drawn up.  This is an area I have been looking at, and believe it
>can be done cleanly - but more on that at a later stage.

Sounds good.

Russ

----------------------------------------------------------------------------
Russ Wright
Lawrence Berkeley Laboratory
Mail Stop 50B-2258
Berkeley, CA  94720
510-486-6965