…but not always.
It should be no secret by now that MO:ULa’s CAVCON is something of a pet subject for me. This latest bit of musing on the subject was triggered by something Malfhok posted on the MO:UL forums.
In that post, Malfhok quoted RAWA’s definition of CAVCON 5:
“CAVCON 5 is the best state. It means that donations are exceeding MO:ULagain expenses, there are a couple of months of MO:ULagain expenses in reserve, and extra money can be put toward server upgrades, bug fixes, new content, incorporating fan created content, etc.”
RAWA was later to clarify (or de-clarify?) that “CAVCON 5 is a bit more nebulous” and required an executive decision to declare that conditions were right.
My moment of revelation here, and I’m kicking myself for not realizing this much sooner, is that the original definition was probably flawed in a fundamental way. Basically, the CAVCON indications give a relative measure of how well donations are going in comparison to expenses (and that was all explained in an earlier post here) – it doesn’t matter what the actual expenses are, whether they go up or down, because you simply ratio the incoming against the outgoing. We know what the expenses consist of even if we can’t assign real values to them:
- Standing charge for server lease – fixed annual fee (but assume it’s spread across each month of the year)
- Bandwith usage charges – self-evident that this is variable month on month
- Support and maintenance labor – variable, but probably fairly steady except when there is an update or major problem
In other words, expenses are linked to usage; it’s not directly proportional to usage but there’s a correlation. And that’s where the CAVCON 5 definition breaks down in my view. I’m guessing that when it was written up RAWA probably had a vague sort of idea like “We can probably do something kinda new if we had X amount of money in the fund and X looks to be roughly equivalent to two or three months of expenses.” But ‘X’ would be an absolute value and you can’t compare an absolute value with a relative one so while it may have equated to 2/3 months of expenses we also need to consider that at the time of writing the definitions in February 2010 we had something over 40,000 logins per week: We’re now down to around 7,000 per week. So if logins equate to bandwidth and bandwidth equates to cost then a months expenses back then was a lot more than it is now – I wouldn’t suggest it’d be six times more because not all of the expenses come from the bandwidth usage but three times more might be a fair guess.
Another factor to consider in the same vein is that over the past year or so, a couple of things have been done to reduce the fixed cost of the servers, most recently the move to reserved instances. In all, I reckon the fixed costs have maybe been halved from the early 2010 position. In the meantime, ‘X’ has remained the same or more likely increased due to a) inflationary effects and b) the fact that it’s harder to get a pipeline going again the longer it’s been switched off (even presuming the folks who could do it were still available). So now, ‘X’ is no longer roughly equivalent to two or three months expenses but more likely eight or nine months or even more, especially if you want to factor in trying to keep a minimum reserve to ward off any CAVCON 1 scares such as happened at the start of July.