Transcript Analyzing Cross-layer Interaction in Overlay Networks

Analyzing Cross-layer Interaction
in Overlay Networks
Srinivasan Seetharaman
September 2007
Overlay Networks
Overlay networking helps
overcome functionality
limitations of the Internet
by forming a virtual network
over the native IP network
that is:


Independent
Customizable
2
Service Overlay Networks
Offer enhanced or new services by deploying
intelligent routing schemes.
Overlay link
Relaying
3
Service Overlay Networks (contd.)
Characteristics:



Nodes and links are persistent
Perform overlay routing independent of native layer routing
Each Overlay path comprises one or more Overlay links,
based on a certain selfish objective
Many types of services can be offered
Multicast (e.g. ESM, Overcast)
 QoS (e.g. OverQoS, SON)
 Security (e.g. DynaBone, SOS)
 Better routes (e.g. RON, Detour, X-Bone)
… and much more

4
Cross-Layer Interaction
Performing dynamic routing at both overlay and
native IP layers leads to:



Conflict due to mismatch or misalignment of
routing objectives
Contention for limited physical resources
Functionality overlap (Both overlay layer and IP
layer perform similar set of functions)
5
Cross-Layer Interaction (contd.)
These issues are amplified in the presence of

Selfish motives and aggressive behavior

Lack of information about other layer

Increasing impact ( #overlays  |Traffic| )
6
Context
Native network topology


Intra-domain
Inter-domain
Attitude of native network



Restrictive
Oblivious
Cooperative
7
Thesis Organization
.
Oblivious
Cooperative
OR vs Load
balancing
Intradomain
Interdomain
Restrictive
[Chapter V]
Interaction
with failure
recovery
[Chapter III]
OR vs Policy
constraints
[Chapter IV]
Overlayfriendly native
network
[Chapter VII]
BT vs Load
balancing
[Chapter VI]
8
INTERACTION BETWEEN FAILURE
RECOVERY IN THE NATIVE
AND OVERLAY LAYERS
Chapter III
Dual Rerouting
Each layer performs rerouting, with no knowledge of
which layer leads to optimal restoration
Overlay
rerouting
C
A
C
OVERLAY1
F
G
D
A
B
A
E
E
F
H
LAYER
H
X Failure
B
NATIVE IP
D
G
Native rerouting
LAYER
12
Downside to Dual Rerouting
1. Overlap of functionality between layers causing


Unnecessary route changes (esp when connectivity in
native network is very dynamic)
Increased probing overhead
2. Unawareness of other layer’s decisions leading to

Multiple simultaneous failures
3. Lack of flexibility and control
15
Tuning Dual Rerouting
Intra-domain
(keepAlive-time = 1 sec, hold-time = 3 secs)
Dual
Rerouting
Average route
changes
Suppress
overlay
rerouting at
0.5 prob.
Defer overlay
rerouting by
0.375 secs
Native-only
rerouting
125.08%
101.59%
109.85%
1.567
Stabilized
inflation
100%
108.32%
100%
1.202
Time when
stable
113.7%
100.48%
107.33%
2.481
114.22%
109.98%
110.73%
1.202
Peak inflation
17
Further Improving Recovery
Adjust the functioning of native layer:
Tuning
keepAlive-time
keepAlive-time
Tuning the native layer keepAlive-time:
This produces the best
tradeoff between # of
route changes,
stabilization time and
recovery time
19
INTERACTION BETWEEN OVERLAY
ROUTING AND TRAFFIC ENGINEERING
Chapter V
Repeated Non-Cooperative Game
Player1: Overlay Routing - Latency-optimized paths between nodes
Player2: Traffic Engineering - Optimal load-balanced routes
Overlay Link
Latencies
Overlay
Routing
Overlay
routes

Native link
delays

Overlay layer
traffic
Traffic on each
overlay link
Native
routes
Traffic
Engineering
TM

Background
traffic
21
Simulation Results
TE
objective
Round
Overlay
objective
Overall
stability
26
Our goal
.. is to propose strategies that
obtain the best possible performance for a
particular layer
while steering the system towards a stable
state.
28
Resolving Conflict – Our Approach
Assume: Each layer has a general notion of
the other layer’s selfish objective
Designate leader / follower
Operate leader such that
a. Follower has no desire to change
b. Follower has no alternative to pick
 Friendly
 Hostile
Use history to learn desired action gradually.
31
Performance of Preemptive Strategies
We proposed four strategies that improve performance
for one layer and achieve a stable operating point
Inflation factor
=
Steady state obj value with strategy
Best obj value achieved
Inflation
Leader
Strategy
Overlay
TE
Overlay
Friendly: Load-constrained LP
Hostile: Dummy traffic injection
1.082
1.023
1.122
1.992
Native
Friendly: Hopcount-constrained LP
Hostile: Load-based Latency tuning
1.027
1.938
1.184
1.072
36
CROSS-LAYER INTERACTION OF
PERFORMANCE-AWARE OVERLAY
APPLICATIONS
Chapter VI
BitTorrent File-Sharing
Popular file-sharing application that
generates a large volume of Internet traffic
Characteristics:



Service capacity increases with demand
Centralized tracker regulating neighborhood
Dynamically change active peers by
choke/unchoke protocol
39
Comparison to Overlay Routing
Data2
Data1
A2
A3
B1
B2
A1
AX
BY
40
BitTorrent Protocol
Tit-for-tat based incentive for uploading decisions

Leecher: Unchoke the fastest uploaders

Seed: Unchoke the fastest downloaders
Popular strategy to improve performance

Optimistic unchoke: periodically look for faster peers
41
BitTorrent Dynamics
B
A
C
D
E
Choke
Choke
Unchoke
Request
Unchoke
When
bottlenecked
on link L1
L2
L1
X
Peer
TFT
Upload statsacross links is balanced
Download stats
Load distribution
Status
Opt?
Interested
Pieces?
BitTorrent apps use all available b/w
Status
My interest
43
BitTorrent Dynamics
A
B
D
E
Choke
Choke
C
Unchoke
L2
L1
When NOT
bottlenecked
on link L1
X
Load distribution across links is unbalanced
44
Cross-Layer Interaction
Operating BitTorrent disrupts load balance
and can result in high max util:

This can be a problem for background traffic
Objective of native layer:

Minimize ( Max Util.)
Objective of BitTorrent:

Minimize (Overall finish time)
45
Simulation Setup
Pick 100 ASes with 60% of them being non-stub ASes
49
Simulation Setup
L1
D
F F
B
A
L2
C
E
Generate 1-50 peers. Each associating with 1-3 torrents
50
Simulation Performance Metrics
Max util across access links
= MaxaE ( Xa/Ca ),
E is set of all links
X is the load, C is the capacity
Average finish time inflation of leechers
= 1/Nl
Nl
 ( ’ /  )
i=1
i
i
Nl is # of leechers
’ is finish time after strategy
53
Reducing Impact – Traffic Engg
TE can be performed across inter-domain
access links, in order to minimize (Max util)
Two flavors: Ingress / Egress
Determines which access link to pick for a
certain destination or source IP address
54
Reducing Impact – Traffic Engg (contd.)
Performance of a random AS (Focus AS)
Applying TE does not
make much difference
55
Reducing Impact – Tuning BitTorrent
Alter certain BitTorrent protocol components or
tune the associated parameters


Minimal reduction of the max util
Significant inflation of finish time
Specifically, we tried each of the following:





Make peer selection random
Make piece selection random
Reduce duration of optimistic unchoking
Freeze list of unchoked peers after 10 mins
Tune the unchoking timers
57
Reducing Impact – Locality-awareness
Locality-based traffic management



Give priority to peers within AS
No change to BitTorrent clients
Also try caching of requests sent outside AS
59
Reducing Impact – Bandwidth Throttling
Limit bandwidth consumed by BT traffic


Popular strategy among most ASes
Involves lesser infrastructure cost
60
Cross-layer Conflict
Native layer and BitTorrent layer constantly
retaliate to other layer’s disruptive behavior
Peers deploy BitTorrent Protocol Encryption
to avoid detection by native layer
We develop two “friendly” BitTorrent
strategies that achieve a mutually agreeable
point by reducing peak load
61
A. Limit # of parallel downloads
Average
The unchoking protocol and their timeline is
uncoordinated across neighbors
62
A. Limit # of parallel downloads (contd.)
Reduces peak load from 0.94 to 0.852
Finish time inflation is 1.1501
63
B. Avoiding common neighbors
Problem is that two peers in same AS often
contact same peer outside AS
Algorithm


Perform bilateral info exchange where each peer
A finds out if its neighbor B has a neighbor C
inside its own AS
If yes, toss a coin to determine if we can
download from this peer B (Randomization acts as
a load balancing strategy)
64
B. Avoiding common neighbors (contd.)
Reduces max util from 0.94 to 0.85
Finish time inflation is 1.187
65
ANALYZING INTER-DOMAIN POLICY
VIOLATIONS IN OVERLAY
ROUTES
Chapter IV
Inter-Domain Policy Violations
Two types of violations exist
Provider
1
Client
1
A
Peer
$$
Overlay
route
B
Client 2
Transit
violation
Provider
2
Legitimate
native route
$
Client
3
C
Exit
violation
70
Measurement Results
Each transit violation has a corresponding exit violation
upstream
Extent of exit policy violations in multihop paths
Violation Type
% paths
Next hop AS violated
72.05
Exit point violated
15.63
Total
87.68
75
Policy Enforcement by Native Layer
As ISPs become aware of the negative impact of
overlays and commence filtering, this leads to

drastic deterioration in overlay route performance

commensurate with the number of ASes enforcing policy
76
Resolving Conflict
Overlay Service Provider (OSP) adopts a
combination of the following strategies for
achieving good legitimate paths:
1. Obtain transit permit from certain AS for $T
2. Add new node to certain provider AS for $N
3. Obtain exit permit from certain AS for $E
77
Illustration of Mitigation Strategy
With no filtering,
11
Tier-1 provider
13
AS hosting overlay node
Cust-Prov relation
Tier-2 provider
Stub customer
21
31
Peering relation
23
22
32
33
Transit
violation
78
Illustration of Mitigation Strategy (contd.)
With filtering, we have no multi-hop paths
11
Tier-1 provider
13
AS hosting overlay node
Cust-Prov relation
Tier-2 provider
Stub customer
21
31
Peering relation
23
22
32
33
79
Illustration of Mitigation Strategy (contd.)
Option 1: Add new overlay node to provider AS 22
11
Tier-1 provider
13
AS hosting overlay node
Cust-Prov relation
Tier-2 provider
Stub customer
21
31
Peering relation
23
22
32
33
Option 2: Obtain transit permit from stub AS 32
80
Objective of Mitigation Strategy
For a certain budget, determine which ASes



to obtain transit permit from
to add new node to
to obtain exit permit from
… so as to achieve the best possible gain
Gain = Native route latency – Overlay path latency
Native route latency
81
Mitigation Results
When all permit fee = P, new node fee = N
Permit
Add new node
83
Summary of Cross-Layer Interaction
Overlays offer valuable services needed by endsystems. But, lead to complex cross-layer interaction
with potentially detrimental effects
Layer awareness is essential to reduce negative
effects and to improve performance of both layers.
We propose simple strategies that achieve this goal
in an effective manner.
90
Contributions of Thesis
Knobs for better control over the cross-layer
interaction
Analysis and mitigation of the conflict in objective
between native and overlay layers:


inter-domain: OR vs Policy enforcement
intra-domain: OR vs Traffic engg
Ways to improve coexistence between BitTorrent
file-sharing and native layer
Framework for network layer support of overlay
services
91
Future of Overlays
Overlays are essential as…



Means for end-systems to collaborate
Environment for testing future innovations (GENI)
Architecture for Future Internet in the form of Network
Virtualization
 Cross-layer interaction will affect performance. How
best to design protocols and services in the future?
92
Future Work
Need to address the network impasse. How to tune
the network for


.. the new breed of Internet applications? (e.g., file sharing)
…and new paradigms of communication? (e.g., wireless)
Which layer to implement a service at? For example,
a service like multicast can be performed at both
native layer and overlay layer!


Which layer to use for a particular scenario?
How can the other layer support this service?
93