Print

Print


Agree completely with Mike's "over engineering" comments. I would suggest two numbers, and possibly a date.

Total number of tries to connect
Total number which worked
When this was first started

With the total numbers and service reading them regularly, the services can build up its own picture of the current state of the Source. The total numbers allow a percent calculation (I think most of us can do that), as well as some idea of precision. Just a percent hides lots of this raw data.

When started (or number of days it's been running) is just a nicety to allow some sort of historical perspective and, assuming reasonable regular testing, a guess as to how often the tests are run.

As for what the tests do, I am with Mike on this and really don't care. It could be just a network level 'ping', or it could be complete query with validation of the results as accurate. Any service that wants to veer towards the latter can explain what they are doing elsewhere, and user services can process it as they like.

Simple and distributed, so everyone who wants to can take part and no one is forced to.

Peter

> -----Original Message-----
> From: SRU (Search and Retrieve Via URL) Implementors
> [mailto:[log in to unmask]] On Behalf Of Mike Taylor
> Sent: Thursday, April 15, 2010 5:35 AM
> To: [log in to unmask]
> Subject: Re: Add "reliability" index to CQL's "zeerex" context set
> 
> On 15 April 2010 13:22, Edward C. Zimmermann <[log in to unmask]> wrote:
> > The question is less that one might want to relate something but the
> semantics
> > of what one is relating unless its purely private--- and then its not
> trying
> > to relate to anyone but only itself.
> >
> > What does a number mean and what do the differences in the numbers
> mean? And
> > unless my numbers and your numbers have nothing to do with one
> another nor the
> > numbers of anyone else they need to have some shared models and
> references.
> >
> > Defining these would, I think, be more difficult (and long winded)
> than
> > creating a common service monitoring network and metrics (which would
> evolve
> > over time in response to shared experiences).
> 
> No-one is going to do that.  Let's stop fooling ourselves.
> 
> A simple number from 0-100 -- that we can probably handle.  Ask for
> more, and who has the time to do that work unfunded (or the time to
> apply for funding?)  If there is one lesson we should have learned
> from the NUMEROUS attempts at Explain over the years, it's that it is
> very, very vulnerable to over-engineering.  Aim too high -- even a
> little too high -- and we hit nothing at all.
> 
> 
> >
> >
> > On Thu, 15 Apr 2010 09:51:48 +0100, John Harrison wrote
> >> On Wed, 2010-04-14 at 23:06 +0100, Peter Noerr wrote:
> >>
> >> > What I'm really questioning is whether a zeerex response from a
> Source is
> > the right place for this information - whatever it is.
> >>
> >> Mike's original request/proposal was to add an index for some
> >> measure of reliability/responsiveness to the ZeeRex context set -
> >>  NOT the ZeeRex record schema.
> >>
> >> I've not seen any convincing arguments against this - in fact the
> >> conversations about how such a measure of reliability/responsiveness
> >> might be calculated seem to indicate that this would be useful.
> >>
> >> John
> >>
> >> > In one very real sense it is the right place, because then each
> Source can
> > decide for itself if it wants to obtain and make available this
> information.
> > But then each Source has to set up some external measuring point to
> get the
> > information, and the method of getting it back, and storing it. Not
> difficult,
> > but another thing to set up, maintain and manage.
> >> >
> >> > To make this easier it might be interesting to consider a "query
> mirror"
> > server, which the Source could send a query message to, and which
> would bounce
> > the message back to the Source. The dumber it is the better, so the
> Source can
> > handle its own authentication and API/protocol/language issues with
> the mirror
> > needing to know nothing about them. And a dumb query mirror could be
> easily
> > deployed at many locations to handle traffic for many Sources.
> Possibly as a
> > SRU server add-on? (The old psychiatrists joke "You're fine, how am
> I?" comes
> > to mind.)
> >> >
> >> > Peter
> >> >
> >> >
> >> > > -----Original Message-----
> >> > > From: SRU (Search and Retrieve Via URL) Implementors
> >> > > [mailto:[log in to unmask]] On Behalf Of Edward C. Zimmermann
> >> > > Sent: Wednesday, April 14, 2010 6:46 AM
> >> > > To: [log in to unmask]
> >> > > Subject: Re: Add "reliability" index to CQL's "zeerex" context
> set
> >> > >
> >> > > That's the point of "global" service monitoring...
> >> > >
> >> > > Running search to ones own targets in ones own networks does not
> need
> >> > > it.. but
> >> > > also it a club house.. ZeeRex/Explain is about going beyond the
> private
> >> > > club.
> >> > >
> >> > > On Wed, 14 Apr 2010 09:01:15 -0400, Ross Singer wrote
> >> > > > On Wed, Apr 14, 2010 at 6:51 AM, Mike Taylor
> <[log in to unmask]>
> >> > > wrote:
> >> > > > > Peter, we didn't envisage this as something that servers
> self-
> >> > > report,
> >> > > > > but as something measured from the outside.  To a client,
> it's
> >> > > > > irrelevant whether a server is running off a UPS, for
> example: all
> >> > > we
> >> > > > > care about is whether we can contact and search in and
> retrieve
> >> > > from
> >> > > > > the server.
> >> > > >
> >> > > > On the flip-side, though, if something's wrong between you and
> that
> >> > > > particular service (let's say a DNS server in your network
> goes
> >> > > > haywire and intermittently forgets certain domains, or
> something),
> >> > > > this service's reliability is tarnished through no fault of
> its own.
> >> > > > Or, let's say, for example, your ISP has decided that service
> X
> >> > > > hasn't paid enough of a toll to get priority to its
> customers...
> >> > > >
> >> > > > -Ross.
> >> > >
> >> > >
> >> > > --
> >> > >
> >> > > Edward C. Zimmermann, NONMONOTONIC LAB
> >> > > Basis Systeme netzwerk, Munich Ges. des buergerl. Rechts
> >> > > http://www.nonmonotonic.net
> >> > > Umsatz-St-ID: DE130492967
> >> --
> >>               '.    ,'.         John Harrison
> >>              '  `  '  '         University of Liverpool
> >>  c h e s h i r e  |  3          e: [log in to unmask]
> >>                   v             w: www.cheshire3.org
> >>               `-..;.'           t: 0151 7954271
> >>                 ..,     (c)
> >
> >
> > --
> >
> > Edward C. Zimmermann, NONMONOTONIC LAB
> > Basis Systeme netzwerk, Munich Ges. des buergerl. Rechts
> > http://www.nonmonotonic.net
> > Umsatz-St-ID: DE130492967
> >
> >