Characterizing%20User%20Access%20To%20Videos%20On%20The%20World%20Wide%20Web

1
Characterizing User Access To Videos On The World
Wide Web
Soam Acharya Inktomi Corporation Foster City, CA
Peter Parnes Center For Distance Spanning
Technology Luleå University of Technology Sweden
Brian Smith Department of Computer
Science Cornell University Ithaca, NY
MMCN 2000
2
Overview

• Analysis of traces from an ongoing VoW trial (VoD
  over the Web)
• 2 year period
• 13100 requests
• 246 titles

3
Why?

• Audio/Video content
• coming online rapidly
• constitute a large percentage (17) of bytes
  transferred online
• Useful to
• Cache Designers
• Codec Engineers
• Network Engineers
• Other Multimedia Researchers
• MM Storage Systems

4
Questions We Asked

• Do accesses to videos exhibit temporal locality?
• How frequently are videos accessed?
• Do users exhibit specific browsing patterns when
  viewing videos?
• What are the file size trends?

5
Roadmap

• VoW Setup
• Analysis Methodology
• Results
• Conclusion
• Future Work

6
VoW Setup

• Lulea University, Sweden
• Center for Distance Spanning Technology
• High speed network (34 Mbps)
• mMOD software system

7
VoW Setup II

• Two years (end of Aug 97 - mid Oct 99)
• 246 video titles
• encoded using H.261 (CIF - 320x240)
• 500 campus machines involved in access, 1400
  outside
• title categories
• general
• movies
• educational
• courses
• tutorials, seminars

8
Analysis

• Video file characteristics
• size
• duration
• bitrate distribution
• Trace access analysis
• Trace refinement
• Actual analysis on refined data

9
Median Movie Size 96 MBytes
10
Median Duration 70 minutes
11

• Quality of video streams deliberately kept low
  (for external users)
• Compression scheme designed to produce lower
  bitrates

12
Trace Access Analysis - Log Filtering

• Initially eliminate from the trace
• HTML documents
• Java applet requests
• images
• Joining a session already in progress

020133 salt.cdt.luth.se GET Movie1 020323
spock.cdt.luth.se GET TVSerial_970206 030412
aniara.cdt.luth.se GET Movie2 031011
aniara.cdt.luth.se STOP Movie2
13
Log Filtering II

• Eliminate from trace
• requests from demo machines
• resolve IP addresses for machine names
• reduce user errors
• hitting STOP button too many times
• hitting GET requests too many times
• Removed 1160 requests, 11965 remaining

14
Trace Analysis Methodology

• General
• How do video requests vary by day?
• Mathematical distributions?
• Do some machines request more than others?
• Pattern Detection
• Inter-access times
• Do users access videos all the way?
• Type of file
• Temporal locality

15
11965 accesses over twenty five months
16
Movie Popularity
Movie popularity did not follow Zipfs law -- P
1/(p1-t ) P freq. of access to a document, p
its rank in popularity
17
Distribution of Requests By Machine

• About 73 of all requests from campus and
  surrounding community
• For requests from within campus
• 2 of all machines (11) gt 21 of requests
• 10 of machines (53) gt 50 of requests
• Lab machines

18
(No Transcript)
19
Partial Access

• 61 of accesses went to completion
• 39 stopped early
• Suggests browsing pattern

20
File Category Variations

• Access patterns vary by file category
• Lectures have temporal locality of access
• Many accesses shortly after going online
• Entertainment videos do not

21
Temporal Locality

• LRU stack analysis

Previous Stack Position Counter
Stack
Trace
Position
Counter
GET Movie1 GET Movie2 GET Movie2 GET Movie2 GET
Movie3 GET Movie1
1 2 3
0 0 0
(increment previous location of currently
referenced document)
22
Temporal Locality

• LRU stack analysis

Previous Stack Position Counter
Stack
Trace
Position
Counter
Movie1
GET Movie1 GET Movie2 GET Movie2 GET Movie2 GET
Movie3 GET Movie1
1 2 3
0 0 0
(increment previous location of currently
referenced document)
23
Temporal Locality

• LRU stack analysis

Previous Stack Position Counter
Stack
Trace
Position
Counter
Movie2
GET Movie1 GET Movie2 GET Movie2 GET Movie2 GET
Movie3 GET Movie1
1 2 3
0 0 0
Movie1
(increment previous location of currently
referenced document)
24
Temporal Locality

• LRU stack analysis

Previous Stack Position Counter
Stack
Trace
Position
Counter
Movie2
GET Movie1 GET Movie2 GET Movie2 GET Movie2 GET
Movie3 GET Movie1
1 2 3
1 0 0
Movie1
(increment previous location of currently
referenced document)
25
Temporal Locality

• LRU stack analysis

Previous Stack Position Counter
Stack
Trace
Position
Counter
Movie2
GET Movie1 GET Movie2 GET Movie2 GET Movie2 GET
Movie3 GET Movie1
1 2 3
2 0 0
Movie1
(increment previous location of currently
referenced document)
26
Temporal Locality

• LRU stack analysis

Previous Stack Position Counter
Stack
Trace
Position
Counter
Movie3
GET Movie1 GET Movie2 GET Movie2 GET Movie2 GET
Movie3 GET Movie1
1 2 3
2 0 0
Movie2
Movie1
(increment previous location of currently
referenced document)
27
Temporal Locality

• LRU stack analysis

Previous Stack Position Counter
Stack
Trace
Position
Counter
Movie1
GET Movie1 GET Movie2 GET Movie2 GET Movie2 GET
Movie3 GET Movie1
1 2 3
2 0 1
Movie3
Movie2
Plot this after running through the entire trace
28
Temporal Locality Result
29
Conclusion

• Videos are relatively large (to capture entire
  lectures, movies)
• Users browse portions of video
• A small number of machines accounted for a large
  number of accesses
• High temporal locality of trace accesses

30
Future Work

• Further analysis on inter-access patterns
• Repeat analysis on traces from other VoW type
  experiments, cache traces ...