Jerry Perser, jperser@veriwave.com - PowerPoint PPT Presentation

About This Presentation
Title:

Jerry Perser, jperser@veriwave.com

Description:

Title: WLAN benchmarking proposal Author: Jerry Perser, Tom Alexander, Muninder Sambi Last modified by: Tom Alexander Created Date: 10/4/2000 3:44:04 AM – PowerPoint PPT presentation

Number of Views:28
Avg rating:3.0/5.0
Slides: 10
Provided by: Jerry182
Learn more at: https://www.ietf.org
Category:

less

Transcript and Presenter's Notes

Title: Jerry Perser, jperser@veriwave.com


1
IETF BMWG WLAN Switch Benchmarking
  • Jerry Perser, jperser_at_veriwave.com
  • Tom Alexander, tom_at_veriwave.com
  • Muninder Singh Sambi, msambi_at_cisco.com

67th IETF San Diego
2
Recap - Motivation and Background
  • Enterprise WLANs highly IP-centric and switched
  • WLAN switches and lightweight APs
  • Layer 3/4 aware (sometimes even Layer-7 aware)
  • Incorporates many IETF-defined functions ARP
    caching and proxying, DHCP service, firewalling,
    IPsec, etc
  • Considerable work in IETF in this area
  • CAPWAP WLAN switch protocols
  • Equipment vendors would like to use the same WLAN
    switch performance benchmarking techniques as
    their customers
  • Its all very ad-hoc today
  • Lack of accepted benchmarks gt poor equipment
    performance
  • Test vendors would like to have a common approach
    to testing too
  • IEEE 802.11T has decided not to take up this work
    (out of scope)
  • But it would like to be kept informed of progress

3
Scope of Proposed Work
  • Extend existing router/switch RFCs and drafts to
    cover WLAN switches, in support of CAPWAP work
  • RFC 1242, RFC 2285, RFC 2544, RFC 2889, RFC 3918,
    etc.
  • Adapt hash stuffing draft
  • Random MACs wireless security problems
  • Extend for specific attributes of WLAN
    switches/networks
  • Mobility, overlay WLANs, secure multicast, etc.
  • Update to handle general WLAN switch data plane
    performance
  • Zero-loss throughput not possible with wireless
    PHYs
  • Specific areas of proposed work
  • Data-plane performance throughput, latency,
    multicast, etc.
  • Control-plane performance reset recovery, etc.
  • Failover performance AP failover, controller
    failover
  • Scalability AP capacity, client capacity, etc.
  • Mobility intra- and inter-subnet roaming, etc.
  • Both terminology and methodology drafts are needed

4
Whats CAPWAP? Whats a WLAN switch?
  • CAPWAP Control And Provisioning of Wireless
    Access Points
  • Specifies architecture protocol for
    centralized, switched approach to WLAN
    infrastructure
  • General architecture WLAN clients talking to
    Wireless Termination Points (APs) talking to
    Access Controllers (WLAN switches)
  • Packets from clients to WTPs transit wireless
    medium
  • Packets are then tunneled over UDP (Ethernet)
    from WTPs to ACs
  • ACs centralize security, policy management,
    higher-level protocols and configuration
    management, and frequently also datapath
    switching
  • CAPWAP protocol specifies control plane, data
    tunneling, security
  • Supports both split-MAC (lightweight APs) and
    local-MAC
  • Provides for discovery, configuration,
    provisioning, failover functions
  • An AC is not simply a switch, bridge or router!
  • Its a combination of termination device,
    multiprotocol bridge/router and stateful
    firewall/traffic manager

5
CAPWAP System Architecture
WLAN Client
SUT
Access Controller (AC, or WLAN Switch)
WLAN Termination Point (WTP, or AP)
WLAN Client
WLAN Client
CAPWAP Tunnels Carrying Wireless Packets
WLAN Client
WLAN Termination Point (WTP, or AP)
Rest Of Network
WLAN Client
WLAN Client
Wireless Links
6
Extensions Needed To RFCs
  • ACs are not routers, but still highly involved in
    Layer 3/4 processing
  • ARP proxies, DHCP snooping, firewalling, QoS
    enforcement, etc.
  • Mobility is a significant issue
  • Wireless clients move mobility performance must
    be tested
  • WTPs support 1-16 overlay networks (logical
    WLANs)
  • Multicast performance becomes quite interesting,
    esp. with security
  • General symmetry assumptions in RFC 2544/2889 do
    not hold
  • WLAN clients dont usually talk to each other
    not an interesting case
  • General assumption that all endpoints are peers
    no longer works
  • Frame size changes as traffic transits ACs
  • WLAN frame sizes also vary with security, QoS,
    etc.
  • Other differences from wired switch benchmarking
    as well
  • ACs impose restrictions on MAC/IP address mapping
  • Intrusion detection and rogue client exclusion
  • Shared-medium on wireless side gt switched medium
    on wired side
  • WLAN PHY does not support zero-loss throughput

7
Specific Items Expected for Terminology
  • Define items not already covered by existing RFCs
  • Topology-related
  • Access Controller, Wireless Termination Point,
    client, etc.
  • Functionality-related
  • Roaming, DoS attack, rogue client, etc.
  • Metrics-related
  • Provisioning database capacity, provisioning
    rate, etc.
  • Upstream traffic, downstream traffic, etc.
  • Wireless-specific
  • SSID, BSSID, etc.

8
Specific Items Expected for Methodology
  • Adapt metrics from existing RFCs augment with
    new
  • Data-plane performance
  • Throughput, latency, etc.
  • QoS differentiation, etc.
  • Control-plane performance
  • AP client provisioning capacity, provisioning
    rate, etc.
  • DoS attack control, etc.
  • Reset recovery
  • Failover performance AP failover, controller
    failover, etc.
  • Wireless-specific
  • Intra/inter-subnet roaming, power-save capacity,
    etc.

9
Next steps
  • Start discussion on work proposals
  • Solicit help
  • Submit drafts
  • Terminology
  • Methodology
  • Comments?
Write a Comment
User Comments (0)
About PowerShow.com