Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
10/12/2010 CSE 124 Network Services Fall 2010
CSE 124Networked Services Fall 2010
Lecture-6
Instructor: B. S. Manoj, Ph.D
http://cseweb.ucsd.edu/classes/fa10/cse124
1
Updates
Project-2
10/12/2010 CSE 124 Network Services Fall 2010 2
10/12/2010 CSE 124 Network Services Fall 2010
Content distribution networks (CDNs)
Content replication
challenging to stream large files (e.g., video) from single origin server in real time
solution: replicate content at hundreds of servers throughout Internet
content downloaded to CDN servers ahead of time
placing content “close” to user avoids impairments (loss, delay) of sending content over long paths
CDN server typically in edge/access network
origin server
in North America
CDN distribution node
CDN server
in S. America CDN server
in Europe
CDN
server
in Asia
3
10/12/2010 CSE 124 Network Services Fall 2010
Content distribution networks (CDNs)
Content replication
CDN (e.g., Akamai) customer is the content provider (e.g., CNN)
CDN replicates customers’ content in CDN servers.
when provider updates content, CDN updates servers
origin server
in North America
CDN distribution node
CDN server
in S. America CDN server
in Europe
CDN server
in Asia
4
10/12/2010 CSE 124 Network Services Fall 2010
CDN example
origin server (www.foo.com)
distributes HTML
replaces:http://www.foo.com/sports.ruth.gif
withhttp://www.cdn.com/www.foo.com/sports/ruth.gif
HTTP request for
www.foo.com/sports/sports.html
DNS query for www.cdn.com
HTTP request for
www.cdn.com/www.foo.com/sports/ruth.gif
1
2
3
origin server
CDN’s authoritative DNS server
CDN server near client
CDN company (cdn.com)
distributes gif files
uses its authoritative DNS server to route redirect requests
client
5
10/12/2010 CSE 124 Network Services Fall 2010
More about CDNs
routing requests
CDN creates a “map”, indicating distances from leaf ISPs and CDN nodes
distance could include geographical, topological, latency, or energy considerations
when query arrives at authoritative DNS server:
server determines ISP from which query originates
uses “map” to determine best CDN server
CDN nodes create application-layer overlay network6
10/12/2010 CSE 124 Network Services Fall 2010
CDN as an overlay cache system
7
Popular CDN techniques
Three popular CDN techniques DNS redirection (DR) URL rewriting (UR) Hybrid scheme (URDR)
DNS redirection (DR) The authoritative DNS name server is controlled by the CDN Two types
Full-site delivery (DR-F) Partial-site delivery (DR-P)
Full-site delivery (DR-F) The entire origin server is hidden from the client The zone (subtree) of the DNS is modified to reflect the authoritative
nature of CDN DNS All requests to content are forwarded to the CDN CDNs may either respond or forward the requests to origin server E.g., Adero, NetCaching, and Unitech Networks’ IntelliDNS
10/12/2010 CSE 124 Network Services Fall 2010 8
Popular CDN techniques
Partial-site delivery (DR-P) Origin server modified the URL of embedded objects host names in the URLs are resolved by the CDN’s DNS E.g. www.cnn.com/bar.gif to cnn.akamai.net/www.cnn.com/bar.gif CDN companies: Akamai, Digital Island, Mirrorimage, Solidspeed, and
Speedera
URL rewriting (UR) Origin server rewrites URL links as part of dynamic pages to redirect
clients to different CDNs Selected objects on origin sites are identified and pushed them to CDN
mirror servers The page is dynamically resolved with the IP address of the mirror
servers No further need for DNS (download the content from mirror) E.g., Clearway CDN
Hybrid scheme (URDR) DNS redirection +URL rewriting E.g., Fasttide
10/12/2010 CSE 124 Network Services Fall 2010 9
Other ways for Content Redirection
Manual hyperlink selection Users can be given a set of mirror choices Among the set, user can make a manual selection
HTTP redirection HTTP request option for redirect can be exploited
.htaccess file can contain redirection information line Redirect /path-to-origin-file /URL-to-CDN-mirror
HTTP refresh <meta http-equiv="refresh" content=“TImeout; URL=www.redirected-
address.com/index.html">
Layer 4 switching Transport layer information such as port number can be used to redirect the
queries to a different server Still the origin-server’s network resources may be consumed Not so suitable for effective CDN
Layer 7 switching Application layer switches can redirect the traffic to different mirror servers Mostly done through a collection of application layer switches at the entry point
to the origin server’s network Network resources for origin server may still be affected
10/12/2010 CSE 124 Network Services Fall 2010 10
10/12/2010 CSE 124 Network Services Fall 2010
Some popular CDN services
11
How measurements are conducted
National Internet Measurement Infrastructure (NIMI) A software system for building network measurement
infrastructures A command and control system for controlling measurement
tools Dedicated NIMI servers (probes) can be run at many locations
and controlled from a single point
PlanetLab Testbed More than 1000 machines world-wide Running virtual machines
Keynote’s streaming performance measurement services Multimedia performance
10/12/2010 CSE 124 Network Services Fall 2010 12
Distributed measurement of streaming services
• Agents attempt connection to a streaming server at 10 times/hour
• Attempt playing media files (e.g for 60s)
• Agents measure the following– Network parameters
• DNS time, end-to-end delay using traceroute, packet statistics, and player statistics
– Streaming statistics• connection success rate, bit rate, connection setup
/buffer/rebuffer time
– Server statistics• server type, serving platform, streaming protocol
– Presentation statistics• frame rate, metafiles, urls, player errors etc.
10/12/2010 CSE 124 Network Services Fall 2010 13
Source: Keynote, Inc.
Keynote’s agents’ location
Network studies
Two dozen NIMI client locations
East and west cost
Send queries to CDN and origin servers
Note the IP addresses, TTL values, transfer times
Attempt to overcome the institutional caches
There may be a chance of institutional cache influencing the performance measurement
10/12/2010 CSE 124 Network Services Fall 2010 14
10/12/2010 CSE 124 Network Services Fall 2010
CDN statistics (late 2000)
HOTMM127 (Hot sites from Media Metrix)
127 sites obtained Media Metrix top 50 list and the lOOhot.com list
31% used CDN for images
An increase from 2% in 1999
URL588-MM500
1030 sites obtained Media Metrix top 500 list and another web site list
17% used CDN for images
15
Impact of DNS TTL and load balancing
10/12/2010 CSE 124 Network Services Fall 2010
TTL:10s
TTL:20s
TTL:20sTTL:120s
TTL:230s
TTL:N/A
16
CDN Response Time performance
Response time (with and without DNS lookup time)
10/12/2010 CSE 124 Network Services Fall 2010 17
10/12/2010 CSE 124 Network Services Fall 2010
CDN performance (averaged over 18 objects)
Download time in seconds (mean,median, and 90%ile)
18
10/12/2010 CSE 124 Network Services Fall 2010
CDN performance (serial Vs pipeline http)
19
10/12/2010 CSE 124 Network Services Fall 2010
Number of unique IPs/client returned
20
Key Traffic categories observed in CDNs
There exists four main traffic categories for CDNs
Web traffic (html objects over http)
Large software downloads
Media downloads (streaming video clips, full movie delivery, or movie download)
Large sized, long-tail content
10/12/2010 CSE 124 Network Services Fall 2010 21
Combined traffic pattern
10/12/2010 CSE 124 Network Services Fall 2010
Web traffic: html/images/scripts/formsPeak-to-Average: 1.8Peak traffic: 10-16pm EST
Large footprint, long tail traffic: OS downloadsPeak-to-Average: 2Peak traffic: 12-18pm EST
Software downloads Peak-to-Average: 1.4Peak traffic: Peaks early Morning; Otherwise Fairly flat
Multimedia: Streaming performance requirementPeak-to-Average: 7Peak traffic: 20-1am EST
22
Some numbers from Akamai
In 2010 Akamai reports
Located in more than 1000 rented data center spaces
66K+ servers
400K+ disks
120K+ CPUs
Petabytes of content traffic
Very high rate per sec: 3.5Tbps
Power usage effectiveness (PUE) is a metric used to determine the energy efficiency of a data center. PUE is determined by dividing the amount of power entering a data center by the power used to run the computer infrastructure within it
10/12/2010 CSE 124 Network Services Fall 2010 23
Some numbers from Akamai
In 2010 Akamai reports
Located in more than 1000 rented data center spaces
66K+ servers
400K+ disks
120K+ CPUs
Petabytes of content traffic
Very high rate per sec: 3.5Tbps
How to save energy in such a large distributed set of servers
10/12/2010 CSE 124 Network Services Fall 2010 24
Saving energy in CDN systems
Power consumption Metric for data centers Power usage effectiveness (PUE) PUE is a ratio: the amount of power entering a data center/the
power used to run the computer infrastructure within it Current rates are low: 1.5-3.0
Most CDNs rent data centers With rented data centers, uncontrollables are
Data center PUE Source of Energy Efficiency of hardwares and other equipments
And the web traffic grows at a high rate
Solution: CDN traffic management server software efficiency are critical
10/12/2010 CSE 124 Network Services Fall 2010 25
10/12/2010 CSE 124 Network Services Fall 2010
Reading
CDN from Kurose and Ross
CDN research paper uploaded at the website
CDN paper for Writeup-2
26