WiFi Direct with Delay-Optimized DTN is the Base Recipe for Applications in Location-Shared Wireless Networking Virtualization

.

.

The Big Picture

M.Zhanikeev -- maratishe@gmail.com --

WiFi Direct with DTN in Location-Shared Wireless Virtualization -- http://tinyurl.com/marat140123 --

2 /23
2/23

.

.

The Big Statements
1.

MIMO and other throughput-boosting technologies are far from

2.

MIMO is not a solution

widespread/personal use
◦ higher last mile throughput

3. current
4.

̸= higher e2e throughput

LTE/3G trend is bad -- less throughput -- more restriction 03

multi-stream aggregation is a proven throughput booster 01 02

5. DTN has matured to the level of

delay-intolerant DTN

03 myself+0 "Virtual Wireless User: A Practical Design for Parallel MultiConnect Using WiFi Direct..." MobiQuitous (2013)
01 myself+0 "Multi-Source Stream Aggregation in the Cloud" ADCN Wiley Book (2013)
02 myself+0 "Experiments with application throughput in a browser with full HTML5 support" IEICE ComEx (2013)


3 /23
3/23

.

.

The Big Idea of WiFi Direct
LAN
WiFi
Direct
3G

Wireless
WINNER In presence of two connection
possibilities, defaults to WINNER.
The two connections can be
JOIN
used together ( joined ).
14

WLAN

"Wi-Fi Peer-to-Peer: Best Practical Guide" Wi-Fi Alliance (2010)



4 /23
4/23

.

.

Proposal vs Multipath
•

traditional multipath:

basically

split TCP -- realtime, server support, etc.

05 06
• group communication is impossible under multipath
• the proposal is pure client-side virtualization -- like OpenvSwitch 07

Single Connection
Singular Traditional
Connectivity Applications

Multipath
Traditional
Multipath
Group Communication

Multiple No known cases
3G/LTE/* + WiFi Direct
Connectivity (wasted potential) THIS PROPOSAL
05 Y.Chen+5 "A Measurement-Based Study of MultiPath TCP Performance over Wireless Networks" ACM SIGCOMM (2013)
06

"MultiPath TCP: Linux Kernel Implementation" http://multipath-tcp.org (2013)

07 K.Yap+6 "Making use of all the networks around us: a case study in android" ACM SIGCOMM (2012)


5 /23
5/23

.

.

Practical LTE/3G and WiFi Direct



6 /23
6/23

.

.

Practice: LTE/3G Throughput
ISP3

Color-filled area: +/- 1 sigma

ISP2

Holiday 1
1400
Throughput (kbps)

Throughput (kbps)

1400
1050
700
350
0

1050

• first presented at 03

700

•

350
0

0

5

10
Hour

15

20

0

5

10
Hour

15

20

Workday

Saturday

1050
700
350
0

actual 3G providers -b-mobile (300kbps),
OCN LTE, Biglobe/
NEC LTE

• only Biglobe -- a recent

1400
Throughput (kbps)

1400
Throughput (kbps)

ISP1

Holiday 3+

entry into the market is
above 1Mbps ... sometimes

1050
700
350
0

0

5

10
Hour

15

20

0

5

10
Hour

15

20



7 /23
7/23

.

.

Practice: WiFi Direct
Other info

1m

10m
18

12

6

1m
10m

27

21.6

10m

WiFi Direct: Deterioration effect of distance (%)

24

Throughput ratio ( WiFi Direct / Bluetooth)

Avg. throughput (Mbps)

1m

32.4

30

Bluetooth

WiFi Direct

0

-10

-20

-30

-40

16.2

0
0

2

4
6
8
No. of files


10

0

2

4
6
8
No. of files

10

0

2

4
6
8
No. of files


10

8 /23
8/23

.

.

Practice: LTE/3G/WiFi + WiFi Direct
Gx: WiFi Direct and 3G for X files

25

G10
Intranet throughput (Mbps)


25

20

G10
G5
15

G1

10

W5
5

G5

W1
W10
W10W5
W1

G1

0

25

15

10

W5

W1
W10
G1

0
0.09 0.12 0.15 0.18 0.21
Internet throughput (Mbps)


Distance: 10m

G10

20

5

Wx: WiFi Direct and Tranditional WiFi for X files

Distance:1m


All data

G5

20

G10
G5
15

G1

10

5

W10W5
W1

0
0.09 0.12 0.15 0.18 0.21

0.09 0.12 0.15 0.18 0.21


9 /23
9/23

.

.

The Proposal: WiFi Direct + DTN



10 /23
10/23

.

.

The Proposal: Assumptions
granular service -- many files, cloud multisource 01, substream 02
2. a smartphone app -- WiFi Direct, delegation over DTN, etc.
1.

◦ plus a registry of users -- for discovery

3.

shared location -- university campus, company, etc.

01 myself+0 "Multi-Source Stream Aggregation in the Cloud" ADCN Wiley Book (2013)
02 myself+0 "Experiments with application throughput in a browser with full HTML5 support" IEICE ComEx (2013)


11 /23
11/23

.

.

The Proposal: The Unit Triangle
End Service
Remote connectivity
Local Connectivity
Unit Triangle

Main
Client


Delegated
Client


12 /23
12/23

.

.

The Proposal: Virtualization

Virtual
Client

• basic method presented in
03

Virtualized
resource

• virtualization is

Internet
Intranet
…

Real
User
…

•

seamless
reasonable range:
◦

note: triangles are
async! not necessarily
parallel!


up

to 5-10 intranet links


13 /23
13/23

.

.

The Proposal: Formal Optimization
minimize

Ndiff = NA,target −

•

∑
{

u∈ A,B,C,...

}

Ru Nu

subject to
∑
t=1,...,T

•

client
• network resource

user

Ndiff ≥ 0,
{
}
∀u ∈ A, B, C, ...

•

{
}
count( A, B, C, ... ) < k,

•

Nu,t < Nu,quota


T -- arbitrary time period -- a month?
{
}
A, B, C are users, A is the main
u

NA , Nu for generic

N has a limit Nquota -- like 1-3Gbytes
in ISPs today

•

Ntarget is the desired throughput of A
using DTN each user has a rating Ru


14 /23
14/23

.

.

Evaluation



15 /23
15/23

.

.

Evaluation: Approach
crawdad 15
• naive optimization -- simple heuristics/models
• use lag to evaluate performance -- definition later
• real mobility traces --

15

"CRAWDAD Mobility Traces" http://crawdad.cs.dartmouth.edu (2013)



16 /23
16/23

.

.

Evaluation: Setup
target: 60s of throughput between 300kbps and 50Mbps
• capacities: LTE 300kbps, WiFi Direct 30Mbps
•

◦ WiFi Direct gets -25% for each new delegation (0.75 multiplier)

•

range: 25m effective range for data exchange



17 /23
17/23

.

.

Evaluation: Models, Metrics

•

MODEL: PACED -- grab neighbors only to satisfy your NA but not
more

MODEL: GREEDY -- grab any neighbor but not more than 5
• METRIC: LAG -- simply actual minus target download time
•



18 /23
18/23

.

.

Proposed ( 1 + log( y) for y > 1, - 1 - log( y) for y < -1)

Performance: Lag
5
4
3
2
1
0
-1
-2
-3
-3


-2
-1
0
1
2
3
4
Traditional ( 1 + log( x) for x > 1, - 1 - log( x) for x < -1)

5


19 /23
19/23

.

.

Throughput for Greedy Model ( 1 + log( y) for y > 1)

Performance: Greedy vs Paced
4.5
4.35
4.2
4.05
3.9
3.75
3.6
3.45
3.3
3.3


3.45
3.6
3.75
3.9
4.05
4.2
4.35
4.5
Throughput for Paced model ( 1 + log( x) for x > 1)

20 /23
20/23

.

.

Wrapup



21 /23
21/23

.

.

Wrapup
>= 4.1
• playing with application components now, software
implementation soon
• immediate plan: a campus-wide application
• WiFi Direct is

natively supported in Android



22 /23
22/23

.

.

That’s all, thank you ...



23 /23
23/23

.

.

[01] myself+0 (2013)
Multi-Source Stream Aggregation in the Cloud
ADCN Wiley Book
[02] myself+0 (2013)
Experiments with application throughput in a browser with full HTML5 support
IEICE ComEx
[03] myself+0 (2013)
Virtual Wireless User: A Practical Design for Parallel MultiConnect Using WiFi
Direct...
MobiQuitous
[04] P.Schmidt+3 (2012)
A first look at multi-access connectivity for mobile networking
ACM CSWS
[05] Y.Chen+5 (2013)
A Measurement-Based Study of MultiPath TCP Performance over Wireless
Networks


23 /23
23/23

.

.

ACM SIGCOMM
[06]

(2013)
MultiPath TCP: Linux Kernel Implementation
http://multipath-tcp.org

[07] K.Yap+6 (2012)
Making use of all the networks around us: a case study in android
ACM SIGCOMM
[08] A.Makela+2 (2012)
Comparison of load-balancing approaches for multipath connectivity
Elsevier
[09] H.Pucha+3 (2008)
Adaptive File Transfers for Diverse Environments
USENIX
[10]

(2013)
OpenVSwitch project
http://openvswitch.org/



23 /23
23/23

.

.

[11] E.Kohler+4 (2000)
The Click Modular Router
ACM TOCS
[12] A.Vasilakos+2 (2011)
Delay Tolerant Networks: Protocols and Applications
CRC Press
[13] A.Balasubramanian+2 (2007)
DTN Routing as a Resource Allocation Problem
SIGCOMM
[14]

(2010)
Wi-Fi Peer-to-Peer: Best Practical Guide
Wi-Fi Alliance

[15]

(2013)
CRAWDAD Mobility Traces
http://crawdad.cs.dartmouth.edu



23 /23
23/23

WiFi Direct with Delay-Optimized DTN is the Base Recipe for Applications in Location-Shared Wireless Networking Virtualization

More Related Content

Viewers also liked (20)

Similar to WiFi Direct with Delay-Optimized DTN is the Base Recipe for Applications in Location-Shared Wireless Networking Virtualization (20)

More from Tokyo University of Science (20)

Recently uploaded (20)

WiFi Direct with Delay-Optimized DTN is the Base Recipe for Applications in Location-Shared Wireless Networking Virtualization