On 10/3/07, Orlando Andico <[EMAIL PROTECTED]> wrote: > This is pretty interesting! > > I'm familiar with (at least on Perl) tie-ing a hash map to a disk file > for persistence. However, how is the redo log implemented? > > Or is that one of those Google secrets? :-) >
Not really. I found some really interesting thesis on the net describing in memory databases. Just a plain vanilla database preimage. i actually use structures. Just cast the structure pointer to an unsigned char then write it. on startup just copy them over and over until no more entries are left. every now and then i generate a snapshot from the log and replace it. i was surprised how useful append only files are specially on a replicated global filesystem. i have tried many different implementations for the redo log actually. xml, strings, and structures and binary xml like. the most generic is the preimage. if you dont have access to the internal structures then the only thing possibility is application specific logging. i like to use binary xml or tag-length-value formats for this one. > > On 10/3/07, Rogelio Serrano <[EMAIL PROTECTED]> wrote: > .. > > its easy to add a redo log. i use large memory bound arrays with an > > append only file backup on a coda fs as a database for very frequently > > accessed data. i need only one lock for the entire table. > > > > freaks out my boss though. > _________________________________________________ > Philippine Linux Users' Group (PLUG) Mailing List > [email protected] (#PLUG @ irc.free.net.ph) > Read the Guidelines: http://linux.org.ph/lists > Searchable Archives: http://archives.free.net.ph > -- Lay low and nourish in obscurity _________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
