Further to the training questions
Cor Bosman cor at xs4all.net
Thu Apr 13 18:59:47 CEST 2000
Hi, > Do other LIRs experience this?? The waitqueue is indeed very long. My last assignment request (admittedly large) got its first response 9 days after it was sent to a hostmaster from the wait queue.. Next response was 5 days later. Eventually it was approved 16 or 17 days after the initial request. Now, I realise the request wasnt standard, but it's a bit long :) (i dont blame the hostmaster in question btw!). What I wonder is if it takes this long because way too much effort is put into reviewing and replying to very small assignments. Say 75% of the hostmaster capacity is spent on assignments smaller than a /28. This means 8 or so hostmasters are on average working on these small assignments. Total costs for those 8 hostmasters are thus very large (probably more than 500000 Euro). Then I would not find it strange if we would discuss initial assignment windows, and growth in assignment windows in the LIR wg. If we could drop 75%, or hell, even 25% of the hostmaster load by being a little bit less strict in assignment windows (which to me is _not_ equal to wasting IP space), then this should be a subject of discussion. It would be in any commercial organisation :) The number are most likely not correct, but what Id like to see sometime are the actual figures. These numbers are necessary for the LIR wg to decide on policy, because they are, and if not, should be, linked. Numbers of interest are for instance: - amount of hostmasters - amount of tickets done per day - amount of initial tickets - amount of followup tickets - amount of closed tickets - amount of time spent per ticket - average response time per initial ticket - average response time per followup ticket - distribution of size of assignments - time spent on each size of assignment. - average assignment window - average time after change from no assignment window to first assignment window These numbers are not very different from what we ask of our helpdesk employees. I know some of these are already made, but I just mention them for completeness. Things you can then find out are for instance: - if the amount of closed tickets continues to be lower than the amount of new tickets, then we should worry shouldnt we? - if time spent per ticket is X, how much time/effort/money should be put into making sure it is less than X. And this is where Stephen's remarks probably come into the discussion. X could probably be lowered a lot with training. I at least have the impression that not very much effort is put into lowering X, and the effort that is being put in is simply hiring more hostmasters. Which unfortunately increases our bill :) - Will changes in assignment window policy dramatically lower the wait queue and costs? * should initial assignment window be set? * should we increase the assignment window more quickly? * should we base assignment windows on people too? Like UUnet, we have 2 LIRs, with the same hostmasters doing the tickets, one has a /23 assignment window, the other /28 or so. This seems like a quick win situation in lessening the load on the hostmasters. I dont have the answer to all these questions ofcourse, but at first glance I see a lot of room for improvement. - what are the main reasons for followup replies on assignment requests. if some main reasons can be found, we could look at improving this. Say every new request generates 4 followups, and we can lower this to 2 follow ups, we save 40% in email load. Assuming most effort is put into 1 of the 5 replies, you can probably still save 20 to 30% in resources. Saving 20% of hostmaster resources translates directly into a lower bill. Not all of the above issues are things the LIR should discuss (like someone said, it'd be micromanagement). But Id like to see reports on it. And id like to see improvements being carried out. And some of the numbers are imho essential to forming policy. I would gladly bring about a policy to increase assignment windows to LIRs with a 'proven' track record, if that would dramatically lower the wait queue and costs and increases efficiency. And 'proven' could be something to discuss in a LIR wg meeting. Regards, Cor Bosman
[ lir-wg Archives ]