One final Old-school question on MODs. (example mods aren't primed for simplicity)
If you have a data file of 460 average byte size, and you have 10,000 records on a 2K system, that would want a prime number near 4600. If you have a BP file with 37 records averaging 15,000 bytes, this math would indicate near 277 frames. Given that BP's are more static than data files, plus their records tend to be bigger, should the separation be considered or is that pretty much a lost art. If this file were 277 frames to begin with, you could never have 37 items hash into all of them. I've heard arguments that the MOD should be the number of BP items in this case. Or even 37,7 with a non-one separation. That doesn't guarantee that they will hash evenly as well. You could still have unoccupied base frames. Is all of this guilding the lily and except for non-prime or extreme under-sizing, the differences aren't worth policing? Thanks in advance. Mark Johnson ----- Original Message ----- From: "Dan Fitzgerald" <[EMAIL PROTECTED]> To: <u2-users@listserver.u2ug.org> Sent: Friday, September 02, 2005 4:39 PM Subject: Re: [U2] Prime number file modulos > Yeah... the problem with non-prime is that lumpy hashing thing. The wrong > prime or file type will do this, too; it's why we call tuning "iterative"; > sometimes it takes a few guesses to get lucky. > > File corruption doesn't occur just because you picked a bad modulus, type, > or sep, but it does increase your risk. Bad flinks and bad blinks don't > occur where there aren't any. The more data you can fit into primary space, > the less vulnerable you are when something else goes wrong. > > The Henry Eggers quote captures it. Re-stated, don't tempt Mr. Murphy. > > "Our greatest duty in this life is to help others. And please, if you can't > help them, could you at least not hurt them?" - H.H. the Dalai Lama > > "When buying & selling are controlled by legislation, the first thing to be > bought & sold are the legislators" - P.J. O'Rourke > > Dan Fitzgerald > > > > > > >From: "Mark Johnson" <[EMAIL PROTECTED]> > >Reply-To: u2-users@listserver.u2ug.org > >To: <u2-users@listserver.u2ug.org> > >Subject: Re: [U2] Prime number file modulos > >Date: Fri, 2 Sep 2005 08:53:01 -0400 > > > >I just got a copy of the stat report and found over 100 files who's mods > >are > >either obviously not prime, ie 1000 or someone's stupid assumption that > >1001 > >is prime. Plus many of these files are 200%-500% under mod'd (is that a > >word?). Plus it's the second slowest client of mine (my microdatas are > >currently the slowest, but sized properly). > > > >I'm checking to see if there's an erroneous resize program or an erroneous > >resize programmer. Not truly understood back in the day (circa 1978), but > >prime numbers were stringently insisted when learning MV101. > > > >Thanks. > >Mark Johnson > > > >P.S. Could "SLOW" stand for System Loves Overflow Workspace? > >----- Original Message ----- > >From: <[EMAIL PROTECTED]> > >To: <u2-users@listserver.u2ug.org> > >Sent: Tuesday, August 30, 2005 5:05 PM > >Subject: Re: [U2] Prime number file modulos > > > > > > > The funniest or most gruesome file tuning that I've seen was > > > a few years ago, at a site that I won't name, who had a master > > > file with a modulo of exactly 100 and where 99% of the ids > > > were numbers ending in 00. > > > > > > Definitely not tuned by "FAST" - - maybe they'd used "SLOW"? > > > > > > Answering Mark's question, "did it really matter to be prime?" > > > . . . No, but if your group distribution is extremely lumpy > > > or spiky, then, as Henry Eggers once said, your file is walking > > > around with a sign taped to the back of its shirt, saying, > > > "Hit Me!". > > > > > > -- > > > CONFIDENTIALITY: The information transmitted is intended > > > only for the person or entity to which it is addressed and > > > may contain material that is confidential, privileged and > > > exempt from disclosure under applicable law. > > > Any review, re-transmission, dissemination or other use of, > > > or taking of any action in reliance upon, this information > > > by persons or entities other than the intended recipient is > > > prohibited. > > > If you received this in error, please contact the sender > > > and delete the material in a secure receptacle or by > > > shredding the document (s). > > > -- > > > > > > > > > > > > -----Mark Johnson wrote: ----- > > > To: <u2-users@listserver.u2ug.org> > > > From: "Mark Johnson" <[EMAIL PROTECTED]> > > > Sent by: [EMAIL PROTECTED] > > > Date: 08/30/2005 09:16AM > > > Subject: [U2] Prime number file modulos > > > > > > A little non-U2 but I'm sure still answerable by many experienced MV > > > persons. > > > > > > One client (AP-Pro, Native) reported to me a bunch of GFE's. Upon > >further > > > investigation, all of the involved data files had non-prime modulos. > > > > > > This begs the question. Did it really matter to be prime. I understand > >the > > > concept of prime numbers and the many forms of hashing so let's not > >deviate > > > into a hashing thread debate. I'm just interested to learn first-hand > > > observations on native systems with non-prime file modulos. > > > > > > Thanks in advance. > > > Mark Johnson > > > ------- > > > u2-users mailing list > > > u2-users@listserver.u2ug.org > > > To unsubscribe please visit http://listserver.u2ug.org/ > > > ------- > > > u2-users mailing list > > > u2-users@listserver.u2ug.org > > > To unsubscribe please visit http://listserver.u2ug.org/ > >------- > >u2-users mailing list > >u2-users@listserver.u2ug.org > >To unsubscribe please visit http://listserver.u2ug.org/ > ------- > u2-users mailing list > u2-users@listserver.u2ug.org > To unsubscribe please visit http://listserver.u2ug.org/ ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
