Rohan Murty Harvard University Jitendra Padhye Ranveer Chandra Alec Wolman and Brian Zill Microsoft Research 1 Trends in Enterprise WiFi Networks Increased adoption and usage ID: 260862
Download Presentation The PPT/PDF document "Designing High Performance Enterprise Wi..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Designing High Performance Enterprise Wi-Fi Networks
Rohan Murty Harvard University Jitendra Padhye, Ranveer Chandra, Alec Wolman, and Brian Zill Microsoft Research
1Slide2
Trends in Enterprise Wi-Fi Networks
Increased adoption and usage [Forrester]Culture of mobility: Users tend to use Wi-Fi even when wired connections are available [Gartner, Forrester, Economist]Move towards an all wireless officeUsers want wire-like performance from wireless networks
2Slide3
Capacity of Conventional Corporate WLANs
Corporate WLAN Study:12 users< 1 Mbps each3Slide4
Characteristics of Conventional Corporate WLANs
Focus on coverageFewer APs than clientsClients talk to APs far away; worsens rate anomalyClients pick APs to associate withUse RSSI of beacon packetsAgnostic to channel load at APsLack adaptive behaviorNo load balancing; fixed channel assignmentsCongestion and hotspots worsen4Slide5
DenseAP
Focus on capacityLots of APs; densely deployedClients can talk to APs near by; mitigates rate anomalyInfrastructure picks client-AP associationsGlobal view of network conditions (channel load, interference, etc.)AdaptabilityLoad balance associations; Dynamic channel assignmentRedistributes load away from local hotspots5Slide6
DenseAP is Practical
No client modificationsWorks with legacy clientsChanges limited to the infrastructureEasy to deploySelf-managing6Slide7
DenseAP
Central Controller (DC)DenseAP System Architecture
Associations
Channel Assignments
Load Balancing
DenseAP
Nodes (DAPs)
Commands to DenseAP nodes
Summarized Data
from DenseAP nodes
Summarized
Data
Wired Network
Commands
Interface with clients
Send summaries to DC
7Slide8
Key Challenges
Controlling AssociationsMechanismsPolicyDynamic Channel AssignmentMechanismPolicyLoad BalancingMechanismPolicy8Slide9
Probe Request
Probe Request
Probe Request
ACL
ACL
ACL
00:09:5B:5A:1F:4F
Association Control in
DenseAP
9Slide10
ACL
ACL
ACL
00:09:5B:5A:1F:4F
Probe Request
MAC = 00:09:5B:5A:1F:4F
RSSI = 30
Probe Request
MAC = 00:09:5B:5A:1F:4F
RSSI = 42
Probe Request
MAC = 00:09:5B:5A:1F:4F
RSSI = 40
Association Control in
DenseAP
10Slide11
Accept Client
ACL
00:09:5B:5A:1F:4F
ACL
ACL
00:09:5B:5A:1F:4F
Association Control in
DenseAP
Client only sees
one
DAP at any given time
Probe Response
11Slide12
Association Policy
What is the quality of a connection between a client and a DAP? (rate)How busy is the medium around each DAP? Overall goal: Associate client with a DAP that will yield good
throughput
12Slide13
A Metric for DAP Selection
Expected Transmission-Rate (Mbps) Available Capacity (AC)(Mbps)
Free Air
Time
(%)
X
=
13Slide14
Probe Request
Probe Request
Probe Request
Free air time = 0.35
DAP2
DAP1
DAP3
RSSI = 20
RSSI = 10
RSSI = 30
Free air
time = 0.45
Free air
time = 0.22
DAP
Free Air-Time
RSSI
DAP1
0.35
20
DAP2
0.22
10
DAP3
0.45
30
Accept Client
Probe Response
DAP
Free Air-Time
RSSI
Ex.
Tx
-Rate
AC
DAP1
0.35
20
18
6.3
DAP2
0.22
10
6
1.32
DAP3
0.45
30
48
21.6
14Slide15
RateMap: Estimating Expected
Transmission-RateCorrelation betweenRSSI of Probe Request packets Avg. throughput between a DAP-client pairRough approximation - ordering of DAPsOnline profiling method that builds RSSI to data-rate estimates15
Upload and RSSI correlation = 0.71
Download and RSSI correlation = 0.61Slide16
Estimating Free Air Time
Estimate how busy is the medium around at a DAPTechnique similar to ProbeGap*Measure time taken to finish a packet transmissionEstimates match up closely with offered traffic load16*Lakshminarayan et al., 2004
*
Vasudevan
et al., 2005Slide17
Channel Assignment
Integrated into the association process DAPs not discovered by clients don’t need channelsA DAP is assigned a channel only when it goes from being passive (no clients) to active (services at least one client)Central controller assigns channel with
least
load
17Slide18
Re-evaluating Associations
So far, associations when a new client joins the networkNo association is perfectClient traffic demands changeLocal hotspots created 18Slide19
Load Balancing
Central controller monitors load on every DAP When channel load on a DAP crosses a certain thresholdClient causing most load is determined Moved to less loaded DAP nearbyEnsure client continues to get at least as much available capacity at the new DAPLoad balancing achieved via handoffsUse association control; manipulate ACLs on DAPs19Slide20
Results
20Slide21
Testbed
21
1 Corp AP
24 DAPs
24 Clients
802.11 a/
bgSlide22
Results: Roadmap
PerformanceDensityChannelsIntelligent AssociationLoad Balancing22Slide23
Overall DenseAP Performance: 802.11a
Gains due toMore channels
DAP density
Intelligent associations
23
12
50% gain
Why?Slide24
Exploring the impact of density
Put all DAPs on the same channelFactors outChannelsIntelligent Associations: same load on all DAPsSingle out impact ofDensity24Slide25
Impact of Density: Using only 1 channel
Higher density provides better performance25Slide26
Is intelligent association control necessary?
26Slide27
Why does intelligent association matter?
Client-DrivenDisable intelligent association controlLet clients pick DAP to associate with (conventional WLANs)Compare with DenseAPFactors outChannelsDensitySingle out impact ofIntelligent association
27Slide28
Necessity of the Association Policy
Intelligent association policy is necessary28
160
% gainSlide29
Load Balancing
29Slide30
Load Balancing
Client 1 moved
Client 1 improves
Clients 2 & 3 improve
Client 2 moved
30Slide31
Other Details and Results in the Paper
Load balancing algorithm and mechanismMobilityPerformanceFewer DAPsFewer channels802.11g…..Scalability31Slide32
Related Work
Plenty of prior work on static channel assignment, power control and associationsEach studied each aspect in isolationRequire client modifications [Ramani and Savage, Infocom 2005]SMARTA [Ahmed et al., CoNext 2006]Examines channel and power controlIncrease overall network capacityDoes not consider associations, load balancing
MDG [
Broustis
et al., MOBICOM 2007]
Identified tuning channel, power and associations
Studies the order in which these knobs must be tuned
Requires client modifications
32Slide33
Overall Contributions
Practical systemHow do density, intelligent association, and more channels affect capacity?Adaptive systemFuture directionsImpact of hidden terminalsHeterogeneous mix of client traffic patternsOther backhauls: e.g. Wireless, powerline33