SCIENTIFIC-LINUX-USERS Archives

July 2014

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
"Patrick J. LoPresti" <[log in to unmask]>
Reply To:
Patrick J. LoPresti
Date:
Tue, 15 Jul 2014 11:18:17 -0700
Content-Type:
text/plain
Parts/Attachments:
text/plain (22 lines)
On Tue, Jul 15, 2014 at 11:11 AM, ToddAndMargo <[log in to unmask]> wrote:
>
>> Who would this differ from the "sync" command?
>
>
> "who" should have been "how"

As someone else said, "sync" just synchronizes the copy on disk with
the copy in memory; i.e. it forces a flush of dirty pages to disk. The
kernel will still keep a copy of those pages cached in RAM, forever,
or until it needs that RAM for something else.

That copy in RAM can mess up your performance tests. Specifically, if
all file meta-data (specifically file sizes and timestamps) is cached,
rsync will not actually touch either the source or destination disk at
all.

"sysctl -w vm.drop_caches=3" forces the kernel to forget what it
cached, setting a clean slate for an honest benchmark.

 - Pat

ATOM RSS1 RSS2