The Feasibility of Supporting Large-Scale Live Streaming Applications with Dynamic Application End-Points

Kay Sripanidkulchai, Aditya Ganjam, Bruce Maggs*, and Hui Zhang Carnegie Mellon University

* and Akamai Technologies

Motivation • Ubiquitous Internet broadcast – Anyone can broadcast – Anyone can tune in

9/1/2004

[email protected]

2

Overlay multicast architectures

Router Source Application end-point 9/1/2004

[email protected]

3

Infrastructure-based architecture [Akamai]

+ Well-provisioned

9/1/2004

[email protected]

Router Source Application end-point Infrastructure server 4

Application end-point architecture [End System Multicast (ESM)]

+ Instantly deployable + Enables ubiquitous broadcast 9/1/2004

[email protected]

Router Source Application end-point 5

Waypoint architecture [ESM] W

+ Waypoints as insurance

9/1/2004

[email protected]

Router Source Application end-point W Waypoint 6

Sample ESM Broadcasts http://esm.cs.cmu.edu Event

9/1/2004

Duration (hours)

Unique Hosts

Peak Size

SIGCOMM ’02

25

338

83

SIGCOMM ’03

72

705

101

SOSP’03

24

401

56

DISC’03

16

30

20

Distinguished Lectures

11

400

80

AID Meeting

14

43

14

Buggy Race

24

85

44

Slashdot

24

1609

160

Grand Challenge

6

2005

280

[email protected]

7

Feasibility of supporting large-scale groups with an application end-point architecture? • Is the overlay stable enough despite dynamic participation? • Is there enough upstream bandwidth? • Are overlay structures efficient?

9/1/2004

[email protected]

8

Large-scale groups • Challenging to address these fundamental feasibility questions – Little knowledge of what large-scale live streaming is like

9/1/2004

[email protected]

9

Chicken and egg problem

Publishers with compelling content need proof that the system works.

9/1/2004

System has not attracted large-scale groups due to lack of compelling content.

[email protected]

10

The focus of this paper • Generate new insight on the feasibility of application end-point architectures for large scale broadcast • Our methodology to break the cycle – Analysis and simulation – Leverage an extensive set of real-world workloads from Akamai (infrastructure-based architecture)

9/1/2004

[email protected]

11

Talk outline • Akamai live streaming workload • With an application end-point architecture – Is the overlay stable enough despite dynamic participation? – Is there enough upstream bandwidth? – Are overlay structures efficient?

• Summary

9/1/2004

[email protected]

12

Measurements used in this study • Akamai live streaming traces – Trace format for a request [IP, Stream URL, Session start time, Session duration]

• Additional measurements collected – Hosts’ upstream bandwidth – Hosts’ GNP coordinates

9/1/2004

[email protected]

13

Akamai live streaming infrastructure

Source

A

A

A

A

A

A

Reflectors

9/1/2004

Edge servers

[email protected]

14

Extensive traces ~ 1,000,000 daily requests ~ 200,000 daily client IP addresses from over 200 countries ~ 1,000 daily streams ~ 1,000 edge servers ~ Everyday, over a 3-month period

9/1/2004

[email protected]

15

Request volume (daily)

Number of requests

Weekdays

Weekends

Missing logs

9/1/2004

[email protected]

16

Large-scale streams • Definitions – Large-scale: peak group size of over 1,000

• Characteristics of our traces [IMC ’04] – 660 large-scale streams – 71% audio vs. 7% video vs. 22 % unknown – 79% non-stop vs. 21% short duration

9/1/2004

[email protected]

17

Largest stream 75,000 x 250 kbps = 18 Gbps!

5pm 9/1/2004

6pm

7pm

8pm

9pm

[email protected]

10pm

11pm 12am 18

Talk outline • Akamai live streaming workload • With an application end-point architecture – Is the overlay stable enough despite dynamic participation? – Is there enough upstream bandwidth? – Are overlay structures efficient?

• Summary

9/1/2004

[email protected]

19

When is a tree stable? Not stable

More stable Stable nodes

X X

Interruptions

Time Ancestor leaves 9/1/2004

Less stable X nodes • Departing hosts have no descendants • Stable nodes at the top of the tree

[email protected]

20

Extreme group dynamics

15% stay longer than 30 minutes (heavy-tailed) 45% stay less than 2 minutes!

9/1/2004

[email protected]

21

Stability evaluation: simulation • Hosts construct an overlay amongst themselves using a single-tree protocol – Skeleton protocol of the one presented in the ESM Usenix ’04 paper • Findings are applicable to many protocols

– Goal: construct a stable tree • Parent selection is key

• Group dynamics from Akamai traces (join/leave) • Honor upstream bandwidth constraints – Assign degree based on bandwidth estimation

9/1/2004

[email protected]

22

Join

Join IP1 IP2 ...

9/1/2004

[email protected]

23

Probe and select parent

IP2 IP1

9/1/2004

[email protected]

IP1 IP2 ...

24

Probe and select parent

Parent selection algorithms • Oracle: pick a parent who will leave after me • Random • Minimum depth (select one out of 100 random) • Longest-first (select one out of 100 random) 9/1/2004

[email protected]

25

Parent leave

X Host leaves

9/1/2004

[email protected]

26

Parent leave

? Host leaves All descendants are disconnected

9/1/2004

[email protected]

27

Find new parent

Host leaves All descendants are disconnected All descendants probe to find new parents

9/1/2004

[email protected]

28

Stability metrics • Mean interval between ancestor change

• Number of descendants of a departing host

Interruptions

X Time Ancestor leaves

9/1/2004

[email protected]

29

Stability of largest stream

Longest-first Random Min depth Oracle: there is stability!

9/1/2004

[email protected]

30

Is longest-first giving poor predictions? Oracle, ~100% no descendants

Longest-first, 91%

Min depth, 82%

Random, 72%

9/1/2004

[email protected]

31

Percentage of sessions with interval between ancestor change < 5 minutes

Stability of 50 large-scale streams

9/1/2004

Longest-first

Random Min depth Oracle There is stability! Of the practical algorithms, min depth performs the best. [email protected]

32

There is inherent stability • Given future knowledge, stable trees can be constructed • In many scenarios, practical algorithms can construct stable trees – Minimum depth is robust – Predicting stability (longest-first) is not always robust; when wrong, the penalty is severe

• Mechanisms to cope with interrupts are useful – Multiple trees (see paper for details)

9/1/2004

[email protected]

33

Talk outline • Akamai live streaming workload • With an application end-point architecture – Is the overlay stable enough despite dynamic participation? – Is there enough upstream bandwidth? – Are overlay structures efficient?

• Summary

9/1/2004

[email protected]

34

Is there enough upstream bandwidth to support all hosts? Video 300 kbps DSL Upstream bandwidth only 128 kbps

DSL

DSL Saturated tree

What if application end-points are all DSL?

9/1/2004

[email protected]

35

Metric: Resource index • Ratio of the supply to the demand of upstream bandwidth Resource index == 1 means the system is saturated • Resource index == 2 means the system can support two times the current members in the system

9/1/2004

Resource Index:

[email protected]

(3+5)/3 = 2.7

36

Large-scale video streams

A few streams are in trouble or close. 1/3 of the streams are in trouble.

Most streams have sufficient upstream bandwidth. 9/1/2004

[email protected]

37

Talk outline • Akamai live streaming workload • With an application end-point architecture – Is the overlay stable enough despite dynamic participation? – Is there enough upstream bandwidth? – Are overlay structures efficient?

• Summary

9/1/2004

[email protected]

38

Relative Delay Penalty (RDP) • How well does the overlay structure match the underlying network topology? RDP = Overlay distance Direct unicast distance US

US 20ms

50ms

US 50ms Europe

US50ms Europe

Results are more promising than previous studies using synthetic workloads and topologies. 9/1/2004

[email protected]

39

Summary • Indications of the feasibility of application end-point architectures – The overlay can be stable despite dynamic participation – There often is enough upstream bandwidth – Overlay structures can be efficient

• Our findings can be generalized to other protocols • Future work: Validate through real deployment – On-demand use of waypoints in End System Multicast – Attract large groups

Thank you! 9/1/2004

[email protected]

40

The Feasibility of Supporting Large-Scale Live ...

Sep 1, 2004 - Ubiquitous Internet broadcast. – Anyone can broadcast .... Video 300 kbps. Upstream bandwidth only 128 kbps. DSL. DSL. Saturated tree. DSL ...

860KB Sizes 0 Downloads 176 Views

Recommend Documents

The Feasibility of Supporting Large-Scale Live Streaming ... - CiteSeerX
Aug 30, 2004 - Real, and Windows Media. In Figure 1(a), there were typically 800-. 1000 distinct ...... Receiver-driven layered multicast. In Proceedings of ACM ...

The Feasibility of Supporting Large-Scale Live Streaming ... - Conviva
formance of several design alternatives and show that simple algo- rithms have the potential to ..... over the past week on a publicly available web page. Approxi-.

The Feasibility of Supporting Large-Scale Live Streaming ... - Conviva
... is computed as the sum of all the degrees that the source and application end- .... bandwidth resources, such as from school or the workplace. To summarize ...

The-Feasibility-of-the-North-Eastern-USA-Supporting-the ...
... Oryx, 47(1), 96–104. Page 3 of 9. The-Feasibility-of-the-North-Eastern-USA-Supportin ... eturn-of-the-Cougar-Puma-concolor 2011 LAUNDRE.pdf.

Characterizing the Opportunity and Feasibility of Reconfigurable ...
best memory hierarchy configuration for each application, ..... Includes dynamic power only, no accounting ..... and Software (ISPASS), White Plains, NY, 2010.

Characterizing the Opportunity and Feasibility of Reconfigurable ...
tablet, laptop, and server environments. As Moore's law continues to deliver ... the memory wall [10], multi-level caches have been a key element of computer architecture for decades with research studies spanning organization [11], write and ...

Christian Morrisson - The Political Feasibility of Adjustment.pdf ...
of short-term measures — budget cuts and tight monetary policy — and through. devaluation. Adjustment was limited to a stabilization programme, the sole. criterion being the reduction of the external deficit as rapidly as possible. It was. soon r

The Feasibility of Streamlining Aid.pdf
This extends the analysis. of Dynarski and Scott-Clayton (2006) which was. limited to dependent students. Page 3 of 13. The Feasibility of Streamlining Aid.pdf.

Testing the feasibility of strategies to enhance flexible delivery ... - Avetra
Centre for Vocational Education Research (NCVER), and designed to test the feasibility .... Englewood Cliffs, New Jersey: Educational Technology Publications.

Is the human population a largescale indicator of the ...
QUANTUM GIS (Quantum GIS Development Team, 2009) interface. Italian provinces ..... Analysis Support System (GRASS) Software. Open Source. Geospatial ...

Is the human population a largescale indicator of the ...
local scale of analysis (Chiari et al., 2010; Magura, L ¨ovei &. T ´othm ´er ´esz, 2009; .... introduced Carabidae species in the data analysed. Esti- mated species ...

Testing the feasibility of strategies to enhance flexible delivery ... - Avetra
by their instructor, and by the course material. They did not typically .... and teachers ready for it? Australian and ... delivery including online learning. Brisbane: ...

The method of reflection-projection for convex feasibility ...
Feb 22, 2002 - positive semidefinite) solution to linear constraints in Rn (resp. in Sn, the space ...... These algorithms fall into two disjoint classes: the so-called.

Testing the feasibility of strategies to enhance flexible delivery ... - Avetra
Centre for Vocational Education Research (NCVER), and designed to test the feasibility of ... Figure 1: Two dimensional representation of factors describing VET learner preferences .... Boote J (1998) Learning to learn in vocational education and tra

Supporting Information
Jul 13, 2010 - macaque brain consists of 95 vertices and 2,402 edges [1], where the network models brain regions as vertices and the presence of long dis-.

Feasibility of Automated Monitoring of Lifting Equipment ...
May 1, 2005 - The range of existing technical solutions for monitoring con- struction ... time intervals (in the order of 5 s) makes it impractical to archive the data in .... for data exchange among a variety of applications (CIMSteel In- tegration 

Design of the Floating-Point Adder Supporting the Format Conversion ...
Aug 8, 2002 - converted into absolute number. Then, as the normal- ization step, the leading zero is calculated in the lead- ing zero counter[9] for the absolute ...

Supporting Information
Oct 26, 2010 - between 0.3 and 1.0 mL of an aqueous colloidal dispersion. (4 g∕L) of partially doped polyaniline nanofibers was mixed with. 3 mL of DI water.

Design of the Floating-Point Adder Supporting the Format Conversion ...
Aug 8, 2002 - Conference on ASICs, pp.223–226, Aug. 2000. [5] A.B. Smith, N. Burgess, S. Lefrere, and C.C. Lim, “Re- duced latency IEEE floating-point standard adder architec- tures,” Proc. IEEE 14th Symposium on Computer Arith- metic, pp.35–