Messages in this thread | | | Subject | Re: OSDI paper - IO-Lite: A Unified I/O Buffering and Caching System | From | Jan Vroonhof <> | Date | 03 Mar 1999 15:36:37 +0100 |
| |
Oliver Xymoron <oxymoron@waste.org> writes:
> Very interesting.
indeed
> effect on CPU caches, SMP scaling, etc. So parts of it are going to be a > hard sell
In fact does anybody see why this needs to be in the kernel? I maybe dense but if I understand the paper correctly from the (casual I admit) glance I had:
1. Most of the gains come from less copying along pipes. 2. [this is not clear from the paper to me] For the gains the occur both sides of the pipe need use IOL. 3. The gains come from copying by reference. 3b. There is an additional gain from to having to recompute TCP checksums.
Excluding 3b the only reason that I can think of at the moment for this to be in the kernel seems to be that this way use of IO-Light by the peer is transparent. I.e. the kernel converts it back to a normal pipe on the fly. If that is the case couldn't this be also achieved by some hands hake between the two processes?
Jan
P.S. I also do not see where the improvement in the pure 'wc file' case is comming from. Doesn't GNU wc already use mmap() and thus avoid any copying?
- 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/
| |