Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!samsung!emory!att!pacbell.com!ucsd!ucbvax!hsdndev!husc6!nmr-m.mgh.harvard.edu!jon From: jon@nmr-m.mgh.harvard.edu (Jon Parmet) Newsgroups: comp.periphs Subject: Re: exabyte record size limit Message-ID: <5703@husc6.harvard.edu> Date: 12 Feb 91 12:27:33 GMT References: <1991Feb8.154740.11155@helios.physics.utoronto.ca> <85054@sgi.sgi.com> Sender: news@husc6.harvard.edu Organization: Mass General NMR Imaging Center Lines: 60 Newsgroups: comp.periphs Subject: Re: exabyte record size limit References: <1991Feb8.154740.11155@helios.physics.utoronto.ca> <85054@sgi.sgi.com> Distribution: usa Organization: Mass General NMR Imaging Center Keywords: gap_threshold mode_select In article <85054@sgi.sgi.com> jeremy@perf2.asd.sgi.com (Jeremy Higdon) writes: >In article , jeremy@cs.adelaide.edu.au (Jeremy Webber) writes: >> Very small transfers also seemed to use more *tape* than larger transfers. I'm >> not sure why. 1K blocks got less than half the storage on the tape that 64K >> blocks did. > >This is because the Exabyte writes bands of nulls while waiting for input. >The default is for it to write up to 7 bands, and each band is 8k. >Depending on your motion threshold, the speed of the Exabyte, and the >turnaround time of your application, I could see you getting far >less storage with a 1K transfer size. The 'mode select' command for the Exabyte allows you to change the number of gap tracks which are written when data is not sent to it fast enough. If throughput isn't a large concern, then changing the gap threshold will allow you to get back some of that lost storage at the expense of some extra re-positioning. Newsgroups: comp.periphs Subject: Re: exabyte record size limit Summary: Expires: References: <1991Feb8.154740.11155@helios.physics.utoronto.ca> <85054@sgi.sgi.com> Sender: Followup-To: Distribution: Organization: Mass General NMR Imaging Center Keywords: Newsgroups: comp.periphs Subject: Re: exabyte record size limit References: <1991Feb8.154740.11155@helios.physics.utoronto.ca> <85054@sgi.sgi.com> Distribution: usa Organization: Mass General NMR Imaging Center Keywords: gap_threshold mode_select In article <85054@sgi.sgi.com> jeremy@perf2.asd.sgi.com (Jeremy Higdon) writes: >In article , jeremy@cs.adelaide.edu.au (Jeremy Webber) writes: >> Very small transfers also seemed to use more *tape* than larger transfers. I'm >> not sure why. 1K blocks got less than half the storage on the tape that 64K >> blocks did. > >This is because the Exabyte writes bands of nulls while waiting for input. >The default is for it to write up to 7 bands, and each band is 8k. >Depending on your motion threshold, the speed of the Exabyte, and the >turnaround time of your application, I could see you getting far >less storage with a 1K transfer size. The 'mode select' command for the Exabyte allows you to change the number of gap tracks which are written when data is not sent to it fast enough. If throughput isn't a large concern, then changing the gap threshold will allow you to get back some of that lost storage at the expense of some extra re-positioning.