Re: FIFO
Joerg Schilling <[email protected]>
| Newsgroups | gmane.comp.archivers.star.user |
|---|---|
| Message-ID | <4342C636.nailBNZ11G057@burner> |
David Brown <[email protected]> wrote: > > On Solaris you would be able to switch to a non-blocked mode by editing > > the st.conf file. > > Linux's driver appears to allow the variable blocksize mode on DAT, but not > AIT tape drives. It seems odd to me, since I can 'mt setblk 10240' and get > the same result. It just makes it inconvenient because I have to know what > the block size of the written tapes were when I wrote them. This is bad and a known problem from the 1980s when SGI IRIX by default did this instead of what was usual on UNIX before. > In the days when a 40 MHz CPU was considered powerful, increasing the > blocksize helped to keep the tape drive streaming. Tape speeds have > increased less than or about equal to an order of magnitude, while CPU and > system speeds have increased several. I guess an older machine might still > benefit from larger blocks, but I'm guessing that star's fifo helps more > than the larger blocks. It seems that the tape drive itself is the limiting factor for the archiving speed. Note that every "block" has a "gap" and some other things in addition and that even if the tape drive and the CPU do not limit the SCSI thoughput, you send/read the net. block size but the tape drive writes/reads the gross block size. A block size of 126k always has been a good compromise between exchangeability and read/write speed. In any case, I would not recommend to use more than 1MB. Jörg -- EMail:[email protected] (home) Jörg Schilling D-13353 Berlin [email protected] (uni) [email protected] (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily