By: Tim Mc (timcaffrey.delete@this.aol.com), May 1, 2021 1:42 pm
Room: Moderated Discussions
Yuhong Bao (yuhongbao_386.delete@this.hotmail.com) on May 1, 2021 2:21 pm wrote:
> Mark Roulo (nothanks.delete@this.xxx.com) on May 1, 2021 2:12 pm wrote:
> > Yuhong Bao (yuhongbao_386.delete@this.hotmail.com) on May 1, 2021 1:01 pm wrote:
> > > The fun thing is that 4K pages probably used to be too large. On a 80386, just 8 tasks would consume
> > > at least 64k and probably 128k just for the page tables alone. (80386 page tables were two levels)
> >
> > What assumptions are you making about hard drive seek times? And the amount
> > of DRAM to the number or processes more-or-less running at one time?
> >
> > 3,000 RPM seems like a reasonable guess for late-1980s hard drives, so 50 Hz.
> >
> > Halving the page size would (roughly) double the number of seeks to page in (and
> > out) the same amount of memory. Have you measured (or modeled) this tradeoff?
> >
> >
>
> I am particularly thinking of PCs with 4MB RAM which used to be common until 1995 or so.
MFM hard drives were 3600 RPM, 3-5ms seek time (but faster than that for head switching), and
typically 2:1 interleaving (by 1987, original XT hard drive were formatted at 5:1 interleaving IIRC),
17 sectors per track. So best case was a transfer rate of ~260K bytes/sec. Including seek time,
rotational latency, and 4K pages that span heads/cylinders would effectively reduce the transfer rate significantly. By 1995 I think the transfer rate had increased to about 13 MB/s, but the seek times were about the same. Tracks held more sectors though, and disk spun faster (5400 or 7200 RPM) so that helped.
> Mark Roulo (nothanks.delete@this.xxx.com) on May 1, 2021 2:12 pm wrote:
> > Yuhong Bao (yuhongbao_386.delete@this.hotmail.com) on May 1, 2021 1:01 pm wrote:
> > > The fun thing is that 4K pages probably used to be too large. On a 80386, just 8 tasks would consume
> > > at least 64k and probably 128k just for the page tables alone. (80386 page tables were two levels)
> >
> > What assumptions are you making about hard drive seek times? And the amount
> > of DRAM to the number or processes more-or-less running at one time?
> >
> > 3,000 RPM seems like a reasonable guess for late-1980s hard drives, so 50 Hz.
> >
> > Halving the page size would (roughly) double the number of seeks to page in (and
> > out) the same amount of memory. Have you measured (or modeled) this tradeoff?
> >
> >
>
> I am particularly thinking of PCs with 4MB RAM which used to be common until 1995 or so.
MFM hard drives were 3600 RPM, 3-5ms seek time (but faster than that for head switching), and
typically 2:1 interleaving (by 1987, original XT hard drive were formatted at 5:1 interleaving IIRC),
17 sectors per track. So best case was a transfer rate of ~260K bytes/sec. Including seek time,
rotational latency, and 4K pages that span heads/cylinders would effectively reduce the transfer rate significantly. By 1995 I think the transfer rate had increased to about 13 MB/s, but the seek times were about the same. Tracks held more sectors though, and disk spun faster (5400 or 7200 RPM) so that helped.