Re: B*-trees
From: <cbbrowne_at_hex.net>
Date: Fri, 12 Jan 2001 04:17:45 GMT
Message-ID: <wklmsh3cwv.fsf_at_mail.hex.net>
>>> Depending on the structure of your data, reading from one file
>>> could also be faster, if one read operation gets more than one
>>> index table. Sure, the head of the harddrive doesn't have to move
>>> that far... now it's clear to me.
Date: Fri, 12 Jan 2001 04:17:45 GMT
Message-ID: <wklmsh3cwv.fsf_at_mail.hex.net>
>>>>> "Christoph" == Christoph Rupp <chrrupp_at_rz.fh-augsburg.de> writes:
>>> Depending on the structure of your data, reading from one file
>>> could also be faster, if one read operation gets more than one
>>> index table. Sure, the head of the harddrive doesn't have to move
>>> that far... now it's clear to me.
>> You can never be sure, what the head does. Typically repositioning >> the file cursor takes more time than reading a multiple amount of >> data. Christoph> Yesyes, that's what i meant. Therefore one file should beChristoph> faster then several files. Thinking about many fragmented Christoph> files on my hard drive :)
Further complications are available...
-> If the data is cached in memory [competently designed OSes have
commonly, since the 1970s, offered this feature], the disk may not get hit for a read operation if the data is already in RAM.
-> Disk drives sometimes lie about what they're doing internally.
They may claim to have some number of heads/sectors/clusters, but underneath have other "geometry" so that you don't really know what's going on.
-- (reverse (concatenate 'string "ac.notelrac.teneerf_at_" "454aa")) <http://www.ntlug.org/~cbbrowne/> ?OM ERRORReceived on Fri Jan 12 2001 - 05:17:45 CET