1. Network Selection Bernard Aboba Microsoft
Month 2002
doc.: IEEE /0638r0
May 2004
Network Selection
Bernard Aboba
Microsoft
Bernard Aboba, Microsoft
John Doe, His Company

2. Outline Problem Definition Current Work Items Virtual APs Summary
May 2004
Outline
Problem Definition
Current Work Items
Virtual APs
Summary
Bernard Aboba, Microsoft

3. Challenges for Public Access WLAN
May 2004
Challenges for Public Access WLAN
Minimizing channel conflicts
In some locations (e.g. airports) multiple networks are becoming the norm.
Airlines are installing networks for use in baggage reconciliation and roving ticket counters
Multiple wireless ISPs often also want to serve airport customers
Radio interference is an issue
In the US and Europe b networks can support only 3 non-overlapping channels
In France and Japan only one channel is available
Once the channels are utilized by existing APs, additional APs will interfere and reduce performance
Minimizing capital expenditures
In this economic environment, raising capital is difficult
Undesirable to build out multiple networks in the same location - why not build one network and share it?
Attaining high utilization of deployed Access Points
Profitability enhanced by filling in periods of low usage on the “diurnal curve”
Implies a need to serve many different types of customers: business, consumers, etc.
Minimizing support costs
Desirable to support a wide variety of clients without having to preconfigure them
Bernard Aboba, Microsoft

4. Wouldn’t It Be Great If…
May 2004
Wouldn’t It Be Great If…
A single network could be shared by multiple providers?
Each provider could retain the flexibility to announce their own network, and select the services they wish to provide (rates, security mechanisms, etc.)?
Each provider could manage their own users without interfering with other providers?
Customers could discover any of the offered networks without needing to pre-configure their stations?
To get there, we need to solve the Network Selection Problem!
Bernard Aboba, Microsoft

5. Aspects of Network Selection
May 2004
Aspects of Network Selection
Access Network Discovery
Which access point to attach to?
Identifier Selection
Which identity and credentials to use in this AP?
Selection of roaming intermediaries
How to route the AAA conversation to the home network?
Payload Routing
How to route payload traffic in the right way
Filters and mandatory tunnels
Bernard Aboba, Microsoft

6. An alternative decomposition
May 2004
An alternative decomposition
Discovery
Access networks and access points
Home networks available through these
Mediating networks
Decision
Manual or automatic
Automatic is preferred
Most likely needs some pre-provisioned preference information
Indicating the selected network
Attach to the chosen network and access point
Provide the chosen identity and home domain
Possibly provide some hints about mediating networks
Bernard Aboba, Microsoft

7. When is Network Selection an Issue?
May 2004
When is Network Selection an Issue?
More than one network is available, each with different characteristics.
The user has multiple credentials, and needs to decide which to use to authenticate to a particular network.
There is more than one roaming path between the access and home network, and service parameters or pricing differs between them.
The roaming relationships between access and home networks are so complicated that current AAA protocols cannot route the requests to the home network based solely on the Network Access Identifier (NAI).
Payload packets get routed or tunneled differently, based on the roaming relationship path. This may have an impact on the available services or their pricing.
Providers share the same infrastructure, such as wireless access points.
Bernard Aboba, Microsoft

8. Some Thoughts… All four problems are relevant.
May 2004
Some Thoughts…
All four problems are relevant.
Potential need for new solutions, at least for the three first problems.
The problems are very hard if you consider them with large number of networks, fast handoffs, security, and automatic decisions.
The proliferation of multiple network selection technologies within IEEE 802, IETF, and 3GPP would be bad.
Solving all problems with current link layers and existing network access devices may not be possible. Phasing?
Bernard Aboba, Microsoft

9. Feedback from 3GPP SA2 WLAN group
May 2004
Feedback from 3GPP SA2 WLAN group
What parts of the problem does 3GPP want to solve?
Problem 1 (Access Network Discovery) and 3 (AAA routing) are relevant to 3GPP.
Problem 2 (Identifier Selection) is considered at this stage out of scope.
Problem 4 (Payload Routing) is considered a separate problem within SA2 scope and is still under discussion.
Observations [J. Arkko]
3GPP uses existing L2 mechanisms for problem 1, expects an IETF solution for problem 3
Needed for 3GPP Release 6
Bernard Aboba, Microsoft

10. Recommendations from IETF 59
May 2004
Recommendations from IETF 59
There is current interest in problem 1 (access network selection)
Problem 1 belongs to layer 2
Long-term, discovery is most efficient at layer 2
There is current interest in problem 3 (roaming intermediary selection)
Problem 3 needs at least partial IETF work, though intermediary discovery could also be done at layer 2
Any IETF intermediary discovery solution would necessarily be short-term
Other problems are lower priority…
Bernard Aboba, Microsoft

11. IETF Work In Network Selection
May 2004
IETF Work In Network Selection
Access Network Discovery
No chartered work yet
Identifier Selection
draft-ietf-pkix-wlan-extns-05.txt
Product of the PKIX WG
Selection of roaming intermediaries
draft-arkko-roamops-rfc2486bis-00.txt
Included in RADEXT WG Charter
Payload Routing
Improvements to AAA filtering and redirection capabilities, proposed in RADEXT WG Charter
Bernard Aboba, Microsoft

12. Thinking About the Access Netowrk Discovery Problem
May 2004
Thinking About the Access Netowrk Discovery Problem
Why are existing solutions not adequate?
WFA: Public Access MRD
Why are recently proposed standards not adequate?
IEEE 802.1ab, IEEE 802.1af
Bernard Aboba, Microsoft

13. What is a Virtual Access Point?
May 2004
What is a Virtual Access Point?
A “Virtual Access Point” is a logical entity that exists within a physical Access Point (AP).
Each Virtual AP appears to stations (STAs) to be an independent physical AP.
Virtual APs emulate the operation of physical APs at the MAC layer.
Virtual APs provide partial emulation of the IP and Application Layer behavior of physical APs.
Emulating the operation of a physical AP at the radio frequency layer is typically not possible unless multiple radios are available.
Bernard Aboba, Microsoft

14. Is It “Virtual” Or Is It Real? Only Your Radio Knows For Sure!
May 2004
Is It “Virtual” Or Is It Real? Only Your Radio Knows For Sure!
Physical APs
Channel 6
Channel 6
SSID: Foo
BSSID: A
Rates: 5.5,11
Security: WPA
SSID: Bar
BSSID: B
Rates: 1,2,5.5,11
Security: Open
Beacon/Probe Response
AP B
STA
AP A
Virtual APs
Channel 6
SSID: Foo
BSSID: A
Rates: 5.5,11
Security: WPA
SSID: Bar
BSSID: B
Rates: 1,2,5.5,11
Security: Open
AP A
Bernard Aboba, Microsoft

15. What Is Required for a Virtual AP?
May 2004
What Is Required for a Virtual AP?
Multiple SSIDs.
Support for multiple SSID advertisement by APs
Support for STA discovery for advertised SSIDs.
Multiple capability advertisements.
Each Virtual AP can advertise its own set of capabilities.
Pre-authentication routing.
Determination of the target SSID prior to Association (for routing of pre-authentication traffic).
Multiple VLANs.
Allow a unique VLAN (and unique default key) to be assigned to each Virtual AP.
Multiple RADIUS configurations.
Multiple RADIUS configurations, one for each virtual AP.
Multiple virtual SNMP MIBs.
A virtual MIB instance per Virtual AP.
Bernard Aboba, Microsoft

16. The State of “Virtual APs” Today
May 2004
The State of “Virtual APs” Today
IEEE does not provide guidance on required MAC-layer behavior of “Virtual APs”
Multiple approaches taken by AP vendors
Different assumptions made by NIC vendors
Interoperability, reliability problems abound
Solution: WFA addressing the issue in the Public Access Group
MRD near completion; requires separate BSSID for each advertised configuration
Test plan to come
Bernard Aboba, Microsoft

17. How Well do Virtual APs Scale?
May 2004
How Well do Virtual APs Scale?
Velayos & Karlsson “Techniques to Reduce IEEE b Handover Time”, Swedish Royal Institute of Technology
Bernard Aboba, Microsoft

18. Problems with IEEE 802.1 Solutions
May 2004
Problems with IEEE Solutions
IEEE 802.1ab only supports “announcements”, not request/response
Designed as an IEEE “slow protocol”
IEEE 802.1af Discovery not oriented towards wireless
Does not address rate negotiation issue
May not support existing Information Elements
In pre-auth, APs from other networks are not reachable
Unlikely to be implemented on APs
Bernard Aboba, Microsoft

19. Recommendations Work needed on enhanced scaling for network discovery
May 2004
Recommendations
Work needed on enhanced scaling for network discovery
Existing Beacon/Probe mechanism does not scale
Problem with increasing dynamic range of rates
Problem with large number of Virtual APs
Problem with increasing Beacon/Probe Response size
Work in progress does not address discovery issues
Discovery of APs whose Beacons cannot be heard
IEEE k Site Report only supports discovery of APs within the same administrative domain
Existing IEEE 802 work does not address the problem
IEEE 802.1ab, IEEE 802.af oriented toward wired, not wireless networks
Bernard Aboba, Microsoft

20. May 2004
Feedback?
Bernard Aboba, Microsoft