Messages in this thread | | | From | David Holland <> | Subject | Re: transactions in VINO | Date | Tue, 22 Jun 1999 16:29:35 -0400 (EDT) |
| |
> Your number was that a Vino transaction begin/end pair is 1570 > cycles on a 100 Mhz Pentium. The L4 IPC operation, if memory > serves me, is something like 130 cycles in the case of interest, > which would give 12x. Withdrawing the 100x figure (with > apologies), I submit that this number amply makes my case.
First of all, I would say that comparing the cost of a transaction to the cost of entering the kernel and saying "12x overhead" is somewhat misleading - sure, it's true, in a way, but you really want to look at it in the context of a complete operation that actually does something. Not that this will probably help a great deal for operations that aren't I/O bound (like, say, setgid().)
However, wasn't your original case related to persistence? VINO transactions aren't persistent in that sense. That is, we make no attempt to commit the kernel state to disk or anything, because VINO's kernel isn't persistent; it's a "normal" kernel in that if it crashes, you reboot.
So given that you were claiming that transactions are too slow for use in a persistent kernel (right?) your case is actually a lot stronger: persistent transactions would require a commit to disk and be slower yet.
-- - David A. Holland | VINO project home page: dholland@eecs.harvard.edu | http://www.eecs.harvard.edu/vino
- 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/
| |