* Sven Rudolph <[EMAIL PROTECTED]> (Mon, Jul 21, 2003 at 03:48:12PM +0200)
>>> Amanda's estimate phase takes here between 3 and 8 hours (with GNU >>> tar). So I thought about ways of skipping the extra estimate phase. >> Topic comes up periodically. There have been mentions of ways to >> substitute your own estimate technique. I think they generally >> involve the "calcsize" command. Probably making it a wrapper >> on the client. > I haven't found many useful info on this. > <http://marc.theaimsgroup.com/?l=amanda-users&m=98888253924842&w=2> > hardcodes the calcsize calls in sendsize. A bit unflexible. yes, but it works for me. I once started work on doing this proper (use exclusion lists, make it a dumptype option &c &c ...) but it was too much work > <http://groups.yahoo.com/group/amanda-hackers/message/2541> is some previous info on > this patch. > Both messages mention speedup numbers. calcsize is three times faster, > and it does all levels in one run. It's 3 times faster *AS* it does all 3 levels (level 0, level incremental, level incremental+1) in one go. > <http://groups.yahoo.com/group/amanda-hackers/message/2542> tries to > find reasons why calcsize isn't used in default amanda. I guess it > must have a drawback; otherwise it would be amanda's default method. > It sounds like calcsize is basically unmaintained code. > What are the current advantages and disadvantages of calcsize? Since I tried to do it properly (2 years ago ?) I don;t think anyone has ever worked on it. The biggest issue is that the estimates from calcsize do not match the real estimates for incremental dumps . People who tested this mentioned upto 5% difference in size. In my case 5% is about the same differnece I get anyway, given the time difference between the estimate and the actual dump, and the fact that lots of data is generated overnight. Currently listening to: the radio Gerhard, ([EMAIL PROTECTED]) == The Acoustic Motorbiker == -- __O But as it did, the heavens opened =`\<, And rain began to fall (=)/(=) Final judgement now, spread before you all
