On 20/02/2013 12:01, Tim wrote:
On Tue, 2013-02-19 at 19:56 +0000, Gordan Bobic wrote:
40 seconds vs 60 seconds to boot up really matters? Really? I find
my machines, laptops included, take longer to POST than they take to
boot up even with mechanical disks, let alone with SSDs.
I wouldn't have thought drive speed would be a major concern at boot
time, these days, unless the boot sequence of an OS was so inefficient
that it loaded up scads of big files. Sometimes I seriously wonder
about what is being loaded at boot time, about whether they're really
part of booting up. As opposed to just doing something.
It's not big files that are a problem, it's lots of files that aren't
sequentially laid out that are a problem on mechanical media. You may
think your disk is good for 100+MB/s, but you can only do 120 seeks per
second on a 7200rpm disk. Assuming your typical file required at boot
time is about 4KB, that's a whopping 480KB/s your disk is managing to
squeeze out at a push.
Take GRUB, for instance. Load a menu, load a graphic, try to play some
audio (yes, I was surprised to see that in the GRUB files). As opposed
to load a menu, start booting from the choice, do nothing else. All
these little extra steps adds another delay, especially when they're
sequential (the next thing happens, after the prior thing). The move to
a parallel boot process is supposed to speed things up on that premise.
As the OS has a whole chain of things it does while booting, and
probably not all them are really needed to be part of the boot process.
What bothers me is that the traditional sysvinit has been fine for the
past 20 years, with much slower hardware. If it wasn't a problem then,
why is it a problem now that we have 1000s of time faster hardware (even
the disks got 2-4x faster in that time)?
But I agree with your position. My computer takes quite some time
getting ready before it even reads from the hard drive.
And the computer just seems to take way too long doing some things,
anyway. You have a multi-gigahertz computer that play games that
through many megabytes of data around in real time for a lovely
impressive picture, yet parsing a single text configuration file, take
so long that you're surprised (e.g. Apache or Squid start-up can be
significantly sped up by stripping the comments out of the config file),
and it's compounded by there being lots of such files read as the
computer starts up. Then there's things like plugging a USB stick in,
or putting a disc in a drive, it takes an extraordinary age before those
tasks complete.
The only conclusion I can make is that the quality of the code and
skills to write it has deteriorated even faster than the rate of
performance improvement of hardware.
As a random example that annoyed me the other day - I used to like an
arcade game called Raiden in the early-mid '90s. IIRC the hardware spec
of the machine was a couple of 10MHz NEC V30 CPUs and a Z80 for sound
processing. It ran beautifully. Today most similar game on my Android
phone struggle horribly because I "only" have a 600MHz CPU and the 3D
OpenGL acceleration in hardware just isn't able to make up the
difference. That's a 30x difference on the MHz figure, and probably
another 10x on top of that for performance-per-clock.
There's inefficiencies everywhere. Considering them insignificant is a
failure. Particularly because they're not in isolation. They compound
together.
I am as disappointed by the state of affairs as you are. I blame it on
poor education, or more specifically miseducation. Things like operating
systems courses being taught based on Java, programming courses
including teaching that the compiler can do all the optimization for you
(last time I checked, most compilers sucked quite badly:
http://www.altechnative.net/2010/12/31/choice-of-compilers-part-1-x86/
and ICC made the performance of the rest look downright embarrasing),
and JavaScript being considered "fast"(
http://www.theregister.co.uk/2013/02/05/gnome_standardises_on_javascript/
).
The thing that concerns me is that nobody seems to be noticing, let
along thinking there's something wrong with it.
Gordan
--
users mailing list
users@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
Have a question? Ask away: http://ask.fedoraproject.org
