Messages in this thread | | | Date | Mon, 27 Sep 2004 16:04:51 -0500 | From | Steven Pratt <> | Subject | Re: [PATCH/RFC] Simplified Readahead |
| |
Ray Bryant wrote:
> Hi Steve, > > On question I have (and I'm sorry, I haven't had time to look at your > patch to sort this out) is what happens if the user supplies a rather > serious I/O size, will you read ahead multiples of that, or what > happens? Or, for that matter, how well will it perform?
Same behavior as the old code. I/Os are broken up into at most max_readahead size pieces. In the case of the old code only 1 of these could be outstanding. In the new code there could be at most 2 outstanding at any point in time.
> > I've heard about HPC applications for IRIX that issue a 2GB read. :-)
This is why for these types of applications, especially on RAID arrays, you need to set max_readahead into the MBs. (But that is a different topic).
Steve
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |