Fabien Benetou's PIM | Wiki / Numbers

Consumption of materials WithoutNotes

During the last 7 years, excluding books and research papers, as of early March 2016

last articles I have read
5079 items
- 4649 read
- 430 watched
- 6383 notes (generally 1 line long)

In their titles 6 items have quantified self, 8 items with quantified and 76 with self.

added

in 1 sentence: ~3 years now with ~1 edit of ~5 words per daylight hour (as of July 2011) to add and generate at the root

Number of errors

- done nightly via ToDo#Maintenance
- approximately ~3 errors per page

Number of pages

1464 pages
- with their pages viewed through the visit distributions since 18/02/2011
  - see stats file from 15/02/2011 to 7/11/2010 and until 18022011 which included Robots.Txt as if it was a wiki page
  - note that it includes my own viewing but now viewing before the 7th even though the content was available for years
  - few events that can explain peaks
    - tweeted about my link of the month (~8pm 10/11/2010)
    - tweeted about my bypassing page (09/10/2010) and getting RT by @PPInternational
    - the WithoutNotes page of the current month (e.g. WithoutNotesApril24 ?) will probably always remain high as I tend to edit it multiples times a day

Number of edits

26K as of 22/01/2012
- past edits ~~23K as of 29/06/2011~~ ~~21K as of 21/05/2011, 14K as of 11/09/2010~~
- with ~51 "words" per edit and peak at 5 words
  - based on space this "1" or "a" would be a word too
- bin/wiki_diff_distribution.sh
  - add via GnuPlot the associated distribution (e.g. Poisson with μ=5)
- exploring it in VR http://vatelier.net/Demos/RollerCoasterEdits
- made with Path:/pub/socialbehaviors/edits_in_unixtime.plt and Path:/pub/socialbehaviors/edits_in_unixtime.data via pmwiki_edits_in_unixtime
- see also an example of edits over time of the Wiki group

distribution of top 10 authors by contributions as of 11/11/2010

  14752 Fabien
    586 Utopiah
    362  Fabien
    315 paola
     90 fabien
     86 (anonymous)
     60 Paola
     26 Adam
     14 lea
     13 Admin

Average number of edits per hour

~1 as of 22/01/2012
- 26000 / (((1327257767 - 1214268628)/3600)-(1307*8))
~1 as of 12/09/2010
- 14000 / (((1284233397 - 1214268628)/3600)-(810*8))
  - unixepoch timestamps of start 1284233397 and current 1214268628 dates thus in seconds
  - 810 number of days * 8 hours of sleep
  - 14K is the number of edits

Size

99M as of 22/01/2012
- 72M ./pub
- 25M ./wiki.d
32M as of 26/08/2010
- 17M ./pub
- 14M ./wiki.d
  - top 5 largest pages as of 27/08/2010
  1. 784KB PersonalInformationStream.WithoutNotes
  2. 484KB ReadingNotes.ReadingNotes
  3. 396KB Content.PersonalInformationStream
  4. 204KB Cookbook.Cognition
  5. 176KB Languages.English
3MB for the entire wiki files with all revisions included as pure text without pics, skin, etc

Number of internal links

- extrema, BackEnd and pages with 0 links, have been removed

ping statistics (from client in Paris)

40 packets transmitted, 40 received, 0% packet loss, time 39057ms
rtt min/avg/max/mdev = 41.586/42.421/47.036/1.269 ms

Page generated time

Path:/pub/internal_measurements/pages_generated_times.txt
- Path:/pub/internal_measurements/pages_generated_times_until_1283271492.txt
- fomat = UNIX epoch when the request ended - time to process the code in seconds - URI
- wiki engine encapsulated by measure_open + measure_close
- we can see a direct correlation between page size (with inclusions) and time required
- note that fetching a small static page is done under 10ms
  - e.g. time wget http://fabien.benetou.fr/pub/page_generated.txt --delete-after returns ~0.010s
- Path:/pub/internal_measurements/sorted_page_generated.txt (as of 19:53:15 CEST Aug 27 2010)
  - cat ../page_generated_measurements.txt | awk '{print $2, $3}' | sort -n -r
  - consider yerrorbars or financebars/candlesticks gnuplot visualization
- specifically for min/avg/max (no mdev)
  - overall 0.1507/1.4903/20.241
  - action=edit 0.2625/0.3391/1.4947
  - action=print 0.1689/0.8612/4.7228
  - action=pagefeed 0.1507/0.3265/0.5045
  - action=search 1.3009/1.6325/2.4968 (as of 1283247915)

Search through social logs

Path:/pub/internal_measurements/socialmemorization_data.txt
improvment
- track the moment of the day when the command is most often used
- pre-load before the next most likely moment
- naive version : 15min before average first daily access time
  - more or less 15min depending on mdev and what is most likely to replace it in cache

Parsing time

~2m to render and simply parse across every page of every group

Resources

- made with Path:/pub/internal_measurements/load_avg.plt and Path:/pub/internal_measurements/load_avg.data
- using cat /path/to/uptime.log | sed "s/.*: //" | sed "s/,.*//" > /path/to/internal_measurements/load_avg.data + * * * * * uptime >> /path/to/uptime.log
vmstat

procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
0  0  43868  23000  78392 257892    0    0     1     0    0    0  3  0 95  2

hdparm tests (as of 1283250693)
- sda1 / device read timings
  - 14 MB in 3.28 seconds = 4.27 MB/sec
- sda1 / cached read timings
  - 792 MB in 2.00 seconds = 395.85 MB/sec
- sda2 /home device read timings
  - 18 MB in 3.14 seconds = 5.72 MB/sec
- sda2 /home cached read timings
  - 778 MB in 2.00 seconds = 388.32 MB/sec
see also
- RRDtool

Entire "request" time study

gather
- brain delay
- hand movement time
- ping
- DNS requests
- server-side generation time of page
- full page downloading time
  - including JavaScript libraries
- client-side page rendering
- use time wget URL --delete-after
  - see also --no-dns-cache and --no-cache
  - allows to also see the retrieval time thus deduce part of the rest
organize in a visually coherent manner
- probably a time axis

Numbers {Wiki}