delorie.com/archives/browse.cgi   search  
Mail Archives: cygwin/2004/02/15/13:34:06

Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-subscribe AT cygwin DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin/>
List-Post: <mailto:cygwin AT cygwin DOT com>
List-Help: <mailto:cygwin-help AT cygwin DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-owner AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
Delivered-To: mailing list cygwin AT cygwin DOT com
Message-ID: <402FBB84.3020407@tlinx.org>
Date: Sun, 15 Feb 2004 10:33:40 -0800
From: linda w <cygwin AT tlinx DOT org>
User-Agent: Mozilla Thunderbird 0.5 (Windows/20040207)
MIME-Version: 1.0
To: "'Cygwin List'" <cygwin AT cygwin DOT com>
Subject: dd if=/dev/sda3 bs=1024 conv=sync|bzip2 >//server/backups/windows_backup.bz2?
X-IsSubscribed: yes

I thought the above would yield an image backup of my windows partition.

It does seem to, at first, but for some odd reason or another, it really 
starts slowing down toward the end of the partition.  (Took about 2-3 
days vs. ~2-3 hours under linux).  Which
would be a first RFE -- but as an ex-manager at SGI used to say, it's 
only performance -- as long as it works, that's the primary concern.

But it doesn't seem to know when to stop.

"df" claims:
C:                    30813648  15304720  15508928  50% /

That would be 30,813,648 1k-blocks or 31,553,175,552 bytes.

Examination of the output of a "dd" shows it is accessing the correct 
partition.

However, at some point, it reached a point where bzip2 stopped
pegging the cpu and dd started being the most guilty party -- at about 
60-70% pegging with csrss.exe taking up the slack at
20-30% cpu (was 0 when bzip was crunching along at high speed).
It will stay like that for a few minutes, at least, then bzip2
will arise from the dead

Now, looking at I/O read/write bytes, I see:
              input               output
dd       29,921,032,311   29,920,942,080
bzip2    29,920,931,594    9,655,162,880

peak working sets for both show bzip2 peaking at ~8.8M and dd at 2 meg.  
They started 69 hours ago with cpu times for both
(dd,bzip) showing 10.5 and 20.5 hours, respectively.


When they started, the output of bzip2 on the server was
increasing file size at the rate of, roughly, 800-900k/second.

Now the file size is increasing at the rate of about 600 bytes/second.

So what the bleep?  I know bzip2 can be slow at times, but output of 600 
bytes/sec?
Server shows zero load, roughly 6G free -- not likely to be a source of 
delay.

The read seem to show a logorithmic slowdown -- not sure if it will ever 
finish at this
rate.  Any idea why it would slow down so much?  I know 'dd' chokes 
somewhat on large files -- or used to as of a few months ago.  Was 
written a bit inefficiently for reading from
random access devices -- telling it to "skip" blocks at the beginning 1) 
didn't like >4 byte
integers, and 2) used the "read" call to skip the bytes.  Meaning you 
can't copy the
last 100 bytes of a 30G file by seeking to it as one might expect, but 
instead, 'dd' reads the data on disk.  Might be necessary if blocks are 
variable length defined by line feeds, but
certainly a drag for what is the more common case (for me) of using 
fixed size blocks.

But bzip seems to be hogging the cpu again now....cranking out ...well 
down to 500 or
so bytes/second now...(sigh).

Wouldl be cool if this worked "right" and would be open-software's 
answer to
the 60-70$ "disk image" program by "Overpriced SW Inc.".

-linda

-- 
    In the marketplace of "Real goods", capitalism is limited by safety
    regulations, consumer protection laws, and product liability.  In
    the computer industry, what protects the consumer?



--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/

- Raw text -


  webmaster     delorie software   privacy  
  Copyright © 2019   by DJ Delorie     Updated Jul 2019