Mobility Traces

• ​MobiLib-USC: Traces from the USC wireless LAN - 4500 users, 400 access points. 2005
  • ​IMPACT: Investigation of Mobile-user Patterns Across University Campuses using WLAN Trace Analysis - analyzes 4 campus traces
• ​MIT Reality trace - 2005
• ​Natl. Singapore U. trace - 2007
• ​Cambridge Haggle traces - multiple traces from 2006 and 2009
• ​San Francisco taxicab traces - 2009
• ​Shanghai Jiao Tong U. SUVNet traces - Traffic GPS data from ~2000 taxis and N buses during 1 month in 2007 - ​Paper on mobility model extracted from trace

Community Trace repositories: ​CRAWDAD (Dartmouth) ​MobiLib (UFL)

Mobility Models

• ​A Group Mobility Model for Ad Hoc Wireless Networks - Hong et al., UCLA - 1999
• ​A Survey of Mobility Models for Ad Hoc Network Research - Tracy Camp, Jeff Boleng, Vanessa Davies, 2002
• ​A Mobility Model Based on WLAN Traces and its Validation - Tuduce et al., 2005
• ​Building Realistic Mobility Models from Coarse-Grained Traces - Yoon et al., MobiSys 2006
• ​Modeling Time-Variant User Mobility in Wireless Mobile Networks - Hsu et al., 2007. ​Code
• ​META: a Mobility Model of MEtropolitan TAxis Extracted from GPS Traces - Huang et al., 2010

Trends

• ​Cisco Visual Network Index (VNI) Home ​Global Mobile Data Traffic Forecast Update, 2010–2015
  • 2010 mobile traffic starting to reflect the 1:20 skewed behavior long observed in fixed networks - top 1% clients generate ~20% of all traffic
  • Bottom 80% clients generated ~20% of overall 2010 mobile data traffic
• ​Morgan Stanley Report on 'Internet Trends' - 2010

DNS Queries

• MIT DNS client-side traces - 3 traces 2000-01, 2 from MIT, 1 from Korea
  • ​DNS Performance and Effectiveness of Caching - Jung et al. - SIGCOMM 2001.
  • Domain name popularity - Zipf-like distribution with param=0.91
  • Latency - median ~100ms, 90th%ile is ~400-1200ms
  • ​CoDoNS re-evaluated DNS perf. from PlanetLab nodes based on above trace: Latency mean/median/90th%ile = 382/39/337ms
• Gatech DNS client-side traces - 75 diff locations, 21 countries, 2002
  • ​Diversity in DNS Performance Measures - Liston et al., SIGCOMM'02
  • Latency - varies, mean is between .95-2.31s, breakdown shows only ~20% of time spent to query root/gTLD
• UCLA DNS local-resolver traces - 3 traces all during same 12 days, 2004
  • ​A Comparative Study of the DNS Design with DHT-Based Alternatives - Pappas et al - INFOCOM'06
  • Domain name popularity - combined Zipf param=0.93
• ​CAIDA DNS root server traces - periodic simultaneous collections across all root servers
  • ​A Day at the Root of the Internet - Castro et al. SIGCOMM CCR 2008
    • 98% of queries are categorized as root pollution - including ~80% that are identical, repeated or resulting from not caching prior referrals
  • ​Understanding and preparing for DNS evolution - comparative analysis across 2007-2009. Includes distribution of queries and query rates across client bins - useful to model client behavior. Very small % of clients generate most queries. Unfortunately, these heavy hitters are mostly sending invalid queries due to mis-configs, firewalls blocking responses, etc.

Web Site and Content Popularity

• ​Web Caching and Zipf-like Distributions: Evidence and Implications - Breslau et al. INFOCOM'99
  • Zipf-like distribution and no correlation between request frequency and response size or rate of change
• ​On the universality of rank distributions of website popularity - modified Zipf
• ​Client Behavior and Feed Characteristics of RSS, a Publish-Subscribe System for Web Micronews - Liu et al. IMC'05
  • Popularity of RSS feeds by hits is Zipf with param=1.37, Popularity by subscribers is Zipf with param=0.5
  • Con: popularity derived from passive logs from univ. dept - small user base

Peer-to-Peer Traffic

• ​Analyzing peer-to-peer traffic across large networks - Sen et al. ToN 2004
• ​Peer-to-peer architecture case study: Gnutella network - Ripeanu, P2P Computing 2001
• ​A Measurement Study of Peer-to-Peer File Sharing Systems - Saroiu et al. 2002