Francesc Altet wrote: > El dc 18 de 04 del 2007 a les 11:38 +0100, en/na Michael Hoffman va > escriure: >> What is the I/O block size that PyTables uses? I ask because on my >> Lustre system, reading blocks of less than 2 MB results in degraded >> performance. > > Yes, for chunked datasets (all except for Array), you can compute the > chunksize in the next way (using 2.0 here): > > # For arrays >>>> reduce(lambda x, y: x*y, array.chunkshape)*array.atom.size > 4096L # 4 KB > # For tables >>>> table.chunkshape[0]*table.description._v_itemsize > 4032L # almost 4 KB > >> Is there a way to change the I/O block size? > > Yup, you can in 2.0 by passing the desired value in new argument > 'chunkshape' in constructors (remember to divide by the size of your > atom first).
Thanks Francesc. I'll look forward to switching to 2.0 soon. ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ Pytables-users mailing list Pytables-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pytables-users
