> Richard Gooch wrote:
[ ... ]
> Suppose you want to make a distributed shared memory machine. What
> "far-memory-latency" would be acceptable? Inside the machine we
> currently achieve:
>
> primary cache: 0.005 us (2 cycles at 400MHz ?)
> secondary cache: 0.02 us
> main memory: 0.1 us
> disk: 12000 us (6ms avg rotational latency +
> 6ms avg seek time)
>
> This clearly demonstrates that you DON"T want to start paging...
Your post was mainly interesting but this section does not look like
relevant considerations to me. You seem to miss that:
- Playing with memory trashes hardware caches.
- Playing with virtual mapping trashes TLB entries.
- High-end disks have large caches.
- High-end disk firmwares are able to take decisions in order to
overlap seek-time + rotationnal latency.
- If you want memory for caching, they you donnot want to have
unused stuff in you caches.
- The main memory is shared by CPUs and DMA-able IO devices.
- Resources are limited, even if the current price of memory allows
to have much of this one. Just wait for next releases of software
and this comfort will disappear, IMO.
The reality an OS has to face is far more complex that the above 4 numbers
you seem to have gotten from your hat, in my opinion. What we want is to
use _efficiently_ the resources and paging helps to achieve it. We just
want to page intelligently and here is the problem, as you know.
Regards,
Gérard.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/