LiveCDPerformance

Differences between revisions 15 and 16
⇤ ← Revision 15 as of 2005-04-27 08:00:11 →
Size: 3620
Editor: intern146
Comment:
← Revision 16 as of 2005-04-27 08:07:35 → ⇥
Size: 2541
Editor: intern146
Comment: Reformat and tidy up
Deletions are marked like this. Additions are marked like this.
Line 32: Line 32:
Performance issues in two areas:  1. Instrumentation is required. d-i logs stage 1, we can wrap init scripts to provide timing information on their startup.
 2. Optimise d-i modules that consume a large amount of time. Locale generation seems excessively slow, and potentially unnecessary for many language setups.
 3. Readahead may be a performance win with 256MB or more, but slower otherwise. Verify with different memory configurations, and ensure that readahead is only enabled when it's a performance benefit. http://unit.aist.go.jp/itri/knoppix/readahead/index-en.html has figures for Knoppix on cloop.
 4. Enable DMA on livecd - it may not work on some systems, but it's effectively required
 5. Filesystem comparison - squashfs and unionfs may give a performance benefit. http://listas.hispalinux.es/pipermail/metadistros-dev/2005-April/000580.html suggests a 50 second improvement. Unionfs is currently unavailable on PPC, but can be used for testing on i386.
 6. Background network setup - this can be left for when the user has hit the desktop (NetworkMagic)
Line 34: Line 39:
1) Installer

 * Can generation of locales be made faster?

Generic CD required - we need to be able to ship them to the rest of the world.

 * Background things like network setup?

We don't actually need setup the network at the beginning

(trivial: get Colin to do it)

2) Post-di boot

 * Readahead configuration may be suboptimal
Seems not very helpful. See this experiment on Knoppix (the same compress filesystem: cloop):
http://unit.aist.go.jp/itri/knoppix/readahead/index-en.html

If we have more than 256Mb we enable readhead otherwise no. Needs implementation.
(high priority).

 * The dma activations help, but no so much.
(requires more investigation)

 * We need check out the filesystem compression (whichever we want to use: cloop, squashfs, unionfs...) performance. But seems this is not the main problem.
(unionfs is in the kernel but not for ppc - i386 is go for basic testing)

 * Somebody did a comparative between the official Kubuntu live CD and the same content of the image but compress with ''Squashfs'' instead of ''cloop'' and ''unionfs'' instead of ''device-mapper''. Also he used the [http://metadistros.hispalinux.es Metadistros] live system. The result was the second one takes 50 sec less than Kubuntu to boot.
http://listas.hispalinux.es/pipermail/metadistros-dev/2005-April/000580.html
(High priority: work in progress - unionfs is in the kernel - needs testing)

NEEDS INSTRUMENTATION - do we have detailed enough logfiles?

Use invoke-rc.d to timestamp runlevels, we have timestamped printk in the kernel
and d-i produces logs with enough details.

IMPLEMENTATION:

 1. Full timings for initial install and second-stage boot.
 2. For packages that spend an extended amount of time in configure, see if that can be improved
 3. Test different filesystems
 4. Investigate the interaction of readahead with the current compressed filesystem - are we spending more time seeking than we'd spend just loading the programs anyway?
 5. Speed up any d-i modules (locale generation?) that consume large amounts of time

Live CD Performance

Status

Introduction

The boot time for the Ubuntu live CD should be comparable (or superior) to other popular live CDs.

Rationale

The better the load time, the better the live CD experience, the greater chance an end user will want to use it and move on to installing Ubuntu or another distro.

Scope and Use Cases

Implementation Plan

  1. Instrumentation is required. d-i logs stage 1, we can wrap init scripts to provide timing information on their startup.
  2. Optimise d-i modules that consume a large amount of time. Locale generation seems excessively slow, and potentially unnecessary for many language setups.

  3. Readahead may be a performance win with 256MB or more, but slower otherwise. Verify with different memory configurations, and ensure that readahead is only enabled when it's a performance benefit. http://unit.aist.go.jp/itri/knoppix/readahead/index-en.html has figures for Knoppix on cloop.

  4. Enable DMA on livecd - it may not work on some systems, but it's effectively required

  5. Filesystem comparison - squashfs and unionfs may give a performance benefit. http://listas.hispalinux.es/pipermail/metadistros-dev/2005-April/000580.html suggests a 50 second improvement. Unionfs is currently unavailable on PPC, but can be used for testing on i386.

  6. Background network setup - this can be left for when the user has hit the desktop (NetworkMagic)

Data Preservation and Migration

none.

Packages Affected

could be several.

User Interface Requirements

none.

Outstanding Issues

unionfs on ppc.

UDU BOF Agenda

  • d-i performance
  • casper performance

  • Ubuntu boot performance (FasterBoot)

UDU Pre-Work

  • Profile the live CD boot sequence to measure the time taken for each step
    • d-i startup (time-to-first-question)
    • Pre-casper d-i activity (measured per menu entry)
    • Casper d-i activity (measured per casper.d script)
    • Standard boot sequence

LiveCDPerformance (last edited 2008-08-06 16:14:57 by localhost)