Internet streaming and network neutrality: comparing the

Italian Networking Workshop - 2016
Internet streaming and network neutrality:
comparing the performance of video hosting services
Alessio Botta, Aniello Avallone, Mauro Garofalo, Giorgio Ventre
University of Napoli Federico II, Napoli, Italy
NM2 SRL, Italy
Network neutrality
General idea: Data on the Internet should be treated in
the same way despite several its characteristics such as
technology, device, application, service, user, provider,
and country they come from or go to
A long debate
• Should operators prioritize traffic?
• Has neutrality ever existed?
• What do users get, in the end?
Alessio Botta - Italian Networking Workshop 2016
2
Our position
• We do not want to take a position pro or against
neutrality
• We rather aim at analyzing current situation with
respect to a particular service: Internet video streaming
• We analyze the performance perceived by users for
different providers from several locations around the
world
• And provide a snapshot on the current situation
Alessio Botta - Italian Networking Workshop 2016
3
Related work
Main topics
• Video popularity, user behavior, delivery policies, and
some performance of YouTube and Daily Motion (Cha
et al. 2007, Plissonneau et al. 2012)
• Infrastructures (e.g., Padmanabhan et al. 2001, Calder
et al. 2013)
We want to compare the performance of different
services from a user viewpoint and update the previous
snapshots
Alessio Botta - Italian Networking Workshop 2016
4
Our contribution
• We introduce a general methodology to acquire and
analyze performance statistics
• We measure, analyze, and compare the performance
of YouTube, Vimeo, and Dailymotion, from several
locations around the world and from a user viewpoint
• We also provide insights on geographical location of
the infrastructures
Alessio Botta - Italian Networking Workshop 2016
5
Providers and infrastructures
• Dailymotion
• French provider
• No indications of CDN even if some news report agreements
with Orange and Akamai
• Youtube
• Google, very much studied
• Complex infrastructure, cache servers inside ISP or at
exchange points
•
• Uses Akamai
• Different improvements for video delivery
Alessio Botta - Italian Networking Workshop 2016
6
Methodology and tools
• 200 PlanetLab nodes from 36 countries
• We selected 4 videos for each provider
• ~500 views , 10k<views <120k, 120k<views<1M, and
views>1M
• Tools
•
•
•
•
youtube-dl to download the videos and get the throughput
netstat to get the IP address of the servers
ping and traceroute to get the RTT and the path/hops
geoping and geotrack to have clues about the location of
the servers
• We performed measurements over 6 months, with
each cycle during 24h
Alessio Botta - Italian Networking Workshop 2016
7
Results
Average Throughput
Alessio Botta - Italian Networking Workshop 2016
9
Average RTT
Alessio Botta - Italian Networking Workshop 2016
10
Daily patterns?
Alessio Botta - Italian Networking Workshop 2016
11
Throughput by country: DM
Alessio Botta - Italian Networking Workshop 2016
12
Throughput by country: Vimeo
Alessio Botta - Italian Networking Workshop 2016
13
Throughput by country: YouTube
Alessio Botta - Italian Networking Workshop 2016
14
RTT by country: Daily Motion
Alessio Botta - Italian Networking Workshop 2016
15
RTT by country: Vimeo
Alessio Botta - Italian Networking Workshop 2016
16
RTT by country: YouTube
Alessio Botta - Italian Networking Workshop 2016
17
Observations (though further analyses)
• Dailymotion
• Entire infrastructure in France (i.e. Paris)
• Abnormal activities by some nodes (Korea and Singapore
maximum throughput is about 12000 KiB/s)
• Vimeo
• Akamai very spread (two-hops away from clients in some
cases) and very high performance
• Clients are always re-directed to the same server, besides day
time or network overload
• YouTube
• Cache-servers in almost all the countries tested
• Delivery strategy assessing both the “distance” and the load
Alessio Botta - Italian Networking Workshop 2016
18
Conclusions
• We performed a performance comparison of three
video hosting services, namely Dailymotion, Vimeo and
YouTube
• We estimated the throughput, RTT, number of hops,
and geography location of the infrastructures
• Dailymotion seems to have a centralized infrastructure
• Vimeo and YouTube use CDNs to deliver their contents,
and the former shows better performance than the
latter
• Highlighted performance differences may be regarded
as lack of neutrality, but they are not due to different
treatments of traffic, but rather to different
infrastructures
Alessio Botta - Italian Networking Workshop 2016
19
Thanks!!
Any question?
[email protected]