Subject: Re: Scheduling solutions?
From: James Conway <jwc@snf.stanford.edu>
Date: Tue, 27 Jan 2004 13:49:10 -0800
Tue, 27 Jan 2004 13:49:10 -0800
Greetings Mark Topinka and Raith Users:

This is an interesting idea.  Similar to one proposed by arvind, 
hendrikb, and myself as we all seek fair solutions to this issue.

For my take I do not wish to administrate the Coral reservations and 
only wish to resolve disputes between people at the time of access.  I 
will not baby sit people or the schedule -- there is no time!

 >> This would be a self-policiing rule, but if one notices anybody 
violating it, ...
Once we agree on a straightforward policy my intention is to become very 
severe with violations after the first warning...
If it come to the point where I must intervene then the user(s) get put 
off the system with no access for one week on second occurrence, then 
one month on the third.  That I expect would keep people toeing to the 
line of the soon to be established new reservation policy.

Hopefully we can come to a consensus and a new  policy at our EBEAM Town 
Meeting February 2, 2004.
We will likely run this policy for a trial one or two month period; 
continue discussions on the raith@snf.stanford.edu list and then 
reconvene and re-evaluate.

Thank you and keep the ideas coming...

James Conway


Mark Topinka wrote:

> Hi all-
>       Thanks to John, Hendrik, James, and John for the ideas so far on 
> this issue.  It is indeed next to impossible to sanely schedule time 
> on the Raith these days.  I have a proposal which I believe is simple 
> to put into place, takes nobody any extra time, and could be started 
> immediately...   How about if the rule becomes no more than 10 hours 
> of sign-up per person for the next 7-day (rolling) period.  As soon as 
> you start one block of time, that time comes off your total and you 
> can sign up for more time.  This would be a self-policiing rule, but 
> if one notices anybody violating it, one could contact James and James 
> could remove the offending reservations (I would propose not All of 
> that persons reservations be removed, just the most distant offending 
> ones).  I took a quick look at the upcoming week, and this policy 
> would free up about 35 hours of time in the next week.  It might 
> slightly slow down super-hard-core users, but it would allow less 
> intensive users the occasional use they need.  Just a proposal...  
> There is no perfect solution (short of maybe buying five more Raiths), 
> but this might be a decent compromise for the time being...  What 
> think y'all? -Mark
>
> At 12:03 PM 1/22/2004 -0800, Hendrik Bluhm wrote:
>
>> Hi,
>>
>> I like the idea of having a request based time allocation system which
>> takes into account how much time people had on the machine recently,
>> but implementing this via a human agent who sits in front of coral seems
>> inefficient, unflexible and probably also unreliable. However, it
>> should be possible to automate this by hosting a web interface on the
>> snf homepage.
>>
>> The procedure could roughly look like this:
>>
>> people request time some 10 days in advance, specifying
>> preferred/impossible time of the day/week, minimum length etc.
>>
>> 8 days ahead, those requests are assigned a priority based on
>> recent machine use (to be obtained from the coral database, reflecting
>> actual use) and granted according to the priority score.
>> People could then have 2 days to transfer the allocated slots onto 
>> coral,
>> which would require that everyone agrees not to make reservations
>> during this time.
>> Less than 5 days in advance, any free spot on coral would be available.
>> Of course, the whole system can be refined and supplemented by 
>> additional
>> rules regarding cancellation etc, e.g. canceling less than 24 hrs
>> in advance counts as actual beam time when calculating priorities.
>>
>> Implementing this would of course require some cooperation with the
>> SNF admins, but should not be to hard. I would be willing to work on the
>> scheduling algorithm, but would greatly appreciate an help regarding how
>> to set up a web interface.
>>
>> The same system could in principle be used for other equipment
>> (innotec, stsetch?, etc.), and if it turns out to be very useful, it 
>> could
>> be incorporated directly in coral one day. (Not by us, though)
>>
>> Hendrik
>>
>>
>> ---------------------------------------------------
>> Hendrik Bluhm
>> Department of Physics
>> Stanford University
>>
>> Work adress:                    Moler Lab
>>                                 Lab. for Advanced Materials
>> Phone: (650) 723-4012           McCullough Bldg.
>> Fax:   (650) 725-2189           476  Lomita Mall
>>                                 Stanford, CA 94305
>


Greetings Mark Topinka and Raith Users:

This is an interesting idea.  Similar to one proposed by arvind, hendrikb, and myself as we all seek fair solutions to this issue.

For my take I do not wish to administrate the Coral reservations and only wish to resolve disputes between people at the time of access.  I will not baby sit people or the schedule -- there is no time!

>> This would be a self-policiing rule, but if one notices anybody violating it, ...
Once we agree on a straightforward policy my intention is to become very severe with violations after the first warning...
If it come to the point where I must intervene then the user(s) get put off the system with no access for one week on second occurrence, then one month on the third.  That I expect would keep people toeing to the line of the soon to be established new reservation policy.

Hopefully we can come to a consensus and a new  policy at our EBEAM Town Meeting February 2, 2004.
We will likely run this policy for a trial one or two month period; continue discussions on the raith@snf.stanford.edu list and then reconvene and re-evaluate.

Thank you and keep the ideas coming...

James Conway


Mark Topinka wrote:
Hi all-
      Thanks to John, Hendrik, James, and John for the ideas so far on this issue.  It is indeed next to impossible to sanely schedule time on the Raith these days.  I have a proposal which I believe is simple to put into place, takes nobody any extra time, and could be started immediately...   How about if the rule becomes no more than 10 hours of sign-up per person for the next 7-day (rolling) period.  As soon as you start one block of time, that time comes off your total and you can sign up for more time.  This would be a self-policiing rule, but if one notices anybody violating it, one could contact James and James could remove the offending reservations (I would propose not All of that persons reservations be removed, just the most distant offending ones).  I took a quick look at the upcoming week, and this policy would free up about 35 hours of time in the next week.  It might slightly slow down super-hard-core users, but it would allow less intensive users the occasional use they need.  Just a proposal...  There is no perfect solution (short of maybe buying five more Raiths), but this might be a decent compromise for the time being...  What think y'all? -Mark

At 12:03 PM 1/22/2004 -0800, Hendrik Bluhm wrote:
Hi,

I like the idea of having a request based time allocation system which
takes into account how much time people had on the machine recently,
but implementing this via a human agent who sits in front of coral seems
inefficient, unflexible and probably also unreliable. However, it
should be possible to automate this by hosting a web interface on the
snf homepage.

The procedure could roughly look like this:

people request time some 10 days in advance, specifying
preferred/impossible time of the day/week, minimum length etc.

8 days ahead, those requests are assigned a priority based on
recent machine use (to be obtained from the coral database, reflecting
actual use) and granted according to the priority score.
People could then have 2 days to transfer the allocated slots onto coral,
which would require that everyone agrees not to make reservations
during this time.
Less than 5 days in advance, any free spot on coral would be available.
Of course, the whole system can be refined and supplemented by additional
rules regarding cancellation etc, e.g. canceling less than 24 hrs
in advance counts as actual beam time when calculating priorities.

Implementing this would of course require some cooperation with the
SNF admins, but should not be to hard. I would be willing to work on the
scheduling algorithm, but would greatly appreciate an help regarding how
to set up a web interface.

The same system could in principle be used for other equipment
(innotec, stsetch?, etc.), and if it turns out to be very useful, it could
be incorporated directly in coral one day. (Not by us, though)

Hendrik


---------------------------------------------------
Hendrik Bluhm
Department of Physics
Stanford University

Work adress:                    Moler Lab
                                Lab. for Advanced Materials
Phone: (650) 723-4012           McCullough Bldg.
Fax:   (650) 725-2189           476  Lomita Mall
                                Stanford, CA 94305