bassa presentation

19
Bassa: A Time Shifted Web Caching System for Developing Regions Wathsala W. Vithanage | Ajanth S. Atukorale June 29, 2011

Upload: wathsala-vithanage

Post on 19-Jun-2015

1.583 views

Category:

Technology


7 download

TRANSCRIPT

Page 1: Bassa presentation

Bassa: A Time Shifted Web Caching System for

Developing Regions

Wathsala W. Vithanage | Ajanth S. Atukorale

June 29, 2011

Page 2: Bassa presentation

Introduction

Web caches in developing regions have varying hit rates. (10% -25% in India according to J. Chen et al. [1] and 50% in Cambodiaaccording to Bowei et al. [2])

Even with web caching page load times are quite high indeveloping regions due to congestion on low bandwidth networks

Large file downloads can be time shifted by a web proxy. Whennetworks are highly congested web proxies can perform time shiftedcaching by queuing the request in order to fetch the file later.

Page 3: Bassa presentation

Observations

Networks in most developing countries are highly congestedduring the day time and underutilized during the night.

This has been observed in an analysis of WWW traffic inGuana and Cambodia [2].

This has been observed in Sri Lankan universities as well [3].

Page 4: Bassa presentation

Bandwidth Utilization

Figure: Bandwidth utilization before time shifted caching.

Page 5: Bassa presentation

Time Shifting the Caching Process

Based on content1 It might be possible to time shift caching of certain content

types. (Ex: Video, Application, etc)

Based on size1 It might be possible to time shift caching of content based on

an object size threshold. (Ex: Objects larger than 20MB)

Page 6: Bassa presentation

Time Shifting the Caching Process

1

10

100

1000

10000

1 10 100 1000 10000 100000 1e+06

Fre

quen

cy

Object Size

Video Content

(a) Video Content

1

10

100

1000

10000

100000

1e+06

1 10 100 1000 10000 100000 1e+06 1e+07

Fre

quen

cy

Object Size

Application Content

(b) Application Content

1

10

100

1000

10000

100000

1e+06

1e+07

1 10 100 1000 10000 100000 1e+06 1e+07

Fre

quen

cy

Object Size

Peak HoursOff Peak Hours

(c) Split by Time

1

10

100

1000

10000

100000

1e+06

1e+07

1 10 100 1000 10000 100000 1e+06 1e+07

Fre

quen

cy

Object Size

All Object for 24 Hours

(d) All Objects

Figure: Logarithmic plot of object size (in 8KB bins) distribution.

Page 7: Bassa presentation

Time Shifting based on Object Size

Deciding the appropriate object size threshold for time shifting.

Figure: Object size threshold vs bandwidth usage with the fitted curve ofthe form

a/x + b where a = 2.075 and b = 0.059.

Page 8: Bassa presentation

Abstract Time Shifted Caching System

Figure: Proxy server with time shifted caching capability.

Page 9: Bassa presentation

Time Shifting Results

0

0.2

0.4

0.6

0.8

1

0 5 10 15 20

Am

ount

of D

ata

Time of the Day

TotalLarger than 70MB

Figure: Object size threshold vs bandwidth usage

Page 10: Bassa presentation

Time Shifting Results

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 50 100 150 200 250 300

Use

d ba

ndw

idth

/Tot

al B

andw

idth

Object Size Threshold

Data for 24 Hours

Data for Peak Hours

(a) Bandwidth Utilization After (b) Bandwidth Utilization Before

Figure: Bandwidth Utilization vs Size Threshold

Page 11: Bassa presentation

Time Shifting Results

The 23% of the bandwidth saved by time shifting of caching hasincreased the number of data volume and number of requests.

Table: Details on Deployment Environment.

Number of Requests Data Volume in GB

Before 7,909,912 760After 9,967,372 1030

Page 12: Bassa presentation

Time Shifting Results

Table: Summary percentages.

Content Type Before As a

Percentage

of 760GB

After As a

Percentage

of 1030GB

Change

in Data

Volume

Video 40.38% 47.42% +59.15%Application 33.06% 33.03% +35.40%

Audio 0.29% 0.59% +175.72%Image 13.74% 8.95% -11.52%Text 12.04% 5.99% -32.57%Other 0.49% 4.02% +1011.86%

Page 13: Bassa presentation

Time Shifting Results

7% of the total bandwidth was used during the night fordownloading video content.

4% of the total bandwidth was used for downloading binary,octet-stream types (ISO disk images and Executables).

Bandwidth consumed for text and images were reduced by-32.5% and -11.5% respectively.

35% increase in the data volume after time shifted cachingwas introduced.

Even though 23% of the bandwidth was freed during thedaytime only 21.7% of this amount was utilized.

1 This implies that our network was not saturated afterdeploying Bassa.

2 This also tells us that it is possible to increase the object sizethreshold to a larger value.

100% increase in large file downloads.

Page 14: Bassa presentation

Conclusion

Bassa has reduced the network congestion during the day.Increased video content is an evidence.

Bassa has utilized the network that idles throughout night.

Bassa has encouraged users to download large files such asISO disk images and Videos. 100% increase in large filedownloads.

Some users got annoyed as 8% of the downloads failed due tosession timeouts.

Page 15: Bassa presentation

Bassa Architecture

Figure: Bassa Components and Messages.

Page 16: Bassa presentation

User Interface

Figure: Bassa Main User Interface.

Page 17: Bassa presentation

User Interface

Figure: Bassa Personal User Interface.

Page 18: Bassa presentation

Questions and Comments

Page 19: Bassa presentation

References

Jay Chen, Saleema Amershi, Aditya Dhananjay, and LakshmiSubramanian.Comparing web interaction models in developing regions.In Proceedings of the First ACM Symposium on Computing

for Development, ACM DEV ’10, pages 6:1–6:9, New York,NY, USA, 2010. ACM.

Bowei Du, Michael Demmer, and Eric Brewer.Analysis of www traffic in cambodia and ghana.In Proceedings of the 15th international conference on World

Wide Web, WWW ’06, pages 771–780, New York, NY, USA,2006. ACM.

W V Wathsala, B Siddhisena, and A S Atukorale.Next generation proxy servers.In Proceedings of 10th International Conference on Advanced

Communication Technology, volume 3 of ICACT ’08, pages2183–2187, 2008.