Title: SCinet2000 The Networks at the SC2000 Conference Dallas, Texas 610 November 2000 Wireless Network Sl
1SCinet2000The Networks at the SC2000
ConferenceDallas, Texas 6-10 November
2000Wireless Network Slides
- 29 November 2000
- David Koester, Ph.D.
- dkoester_at_mitre.org
2Outline
- The Networks of SC2000
- SCinet2000
- XNet
- commodity network
- wireless network
- Photos
- The SC2000 Networking Challenge
- Network monitoring
- tools
- statistics
- Details on wireless
- SCinet2001
3Commodity Network
4Wireless Network OverviewUpper Level
5Wireless Frequency AssignmentsUpper Level
6Wireless Frequency AssignmentsGround Floor
7Network Access Point (NAP) Ceiling Mount
8Outline
- The Networks of SC2000
- SCinet2000
- XNet
- commodity network
- wireless network
- Photos
- The SC2000 Networking Challenge
- Network monitoring
- tools
- statistics
- Details on wireless
- SCinet2001
9Wireless Networking Details Hardware and
Configuration
- Network hardware
- A total of 38 network access points (NAPs)
- As many as 360 simultaneous wireless clients
- Configured NAPs via DHCP
- Installation process
- modify DHCP scripts
- install on the network while watching boot over
console port - check parameters via web server in NAP
- Modify setup
- manually using web server interface
- using a script and http push
10Wireless Networking DetailsNetwork Performance
- 100 coverage in required areas
- Able to communicate on the move and reassociate
to other NAPs automatically - Supported 220 Kbps RealVideo download stream when
stationary - Could not support similar RealVideo uploads
- Supported only 34 Kbps RealVideo download stream
when moving - Load balancing during associations appeared
problematic - Limited WAN TCP/IP throughput performance over
wireless - as great as 3 Mbps throughput to ORNL over wired
meanwhile less than 100 Kbps throughput to ORNL
over wireless - problems caused by additional errors encountered
over shared media same route could not
maintain large widow
11Wireless Networking DetailsFrequency Assignment
- Want to optimize spatial separation
- Use a Greedy Algorithm with Rollback
- Examine closest three
- Results are not always ideal
- Be sure NAPs are setup correctly!
START
12Wireless Networking PerformanceTotal NIC
Associations
13Wireless Networking PerformanceNIC Associations
at the Education Meeting
14Wireless Networking PerformanceNAP Throughput
15Wireless Networking Performance100 Wireless
Users in a Room
- Had the opportunity to examine wireless network
performance - 2 NAPs
- 100 wireless users
- Users were K-12 educators with Compaq laptops and
Cisco NICs - Observations
- 90 NICs associated with the NAP at the back of
the room - over time 70 NICs associated with the NAP at the
front of the room - teachers were asked to download images from a
server - aggregate load peaked at about 1 Mbps
- server was on a T1 line!
- users experienced significant delays
- was able to download a video stream from MITRE
over wireless, through VPN at 200kbps without
loss at the same time as image downloads - hoped for more stress on the link!
16Wireless Networking ObservationsReal-time Mobile
Connectivity Wednesday
- Watched Webcast at 220Kbps in Ballroom
C-Four - Walked back to the SCinet NOC
- Webcast dropped to 34 Kbps
- Automatically reassociated
- Stopped in D320
- No audio dropouts
17Wireless Networking ObservationsReal-time Mobile
Connectivity Thursday
- Watched Webcast at 220Kbps in Ballroom
C-Four - Walked back to the SCinet NOC
- Webcast dropped to 34 Kbps
- Automatically reassociated
- Stopped in D320
- No audio dropouts
18Wireless Networking DetailsNAP Coverage
Experiments
- Reduced power from 30 to 15 mw
- performance improved on floor
- performance degraded between rooms
- Hypothesized that there were too many NAPs
- shut down three NAPs
- examined coverage
- performance degraded
- turned NAPs back on
Channel 1
86
82
Channel 6
81
83
Channel 11
85
84
72
30
76
73
D322
71
74
D321
75
D324
D320
D327
35
B4
B3
B2
B1
19Wireless NetworkingElectromagnetic Interference
(EMI)
- EMI seen as very large numbers of CRC errors
- 802.11b wireless Ethernet operates in the 2.4 GHz
non-licenced frequency band along with - telephones
- wireless television cameras
- wireless closed captioning devices
- other 802.11 networking hardware
- microwave ovens
- EMI was a significant problem on the conference
showfloor - booths wanted their own wireless networks
- the company doing attendee registration brought
wireless - numerous wireless appliances
- EMI was not a problem downstairs in another part
of the building
20Wireless Network Monitoring Lessons Learned
- Be prepared to bring your own server
- preinstall any software
- test the application before deployment
- A good network monitoring package will assist
during network deployment - autodiscovery tie it to the DHCP server
- monitor actual system parameters
- Virtual topology is not enoughespecially for
wireless need spatial location information - Require other information in readily available
format to monitor wireless - transmitter setup frequency and power setting
- trap on several errors radio statistics and CRC
errors - links to web server in each NAP
- Need to fix network configuration when fully
deployed to test for - unexpected additions
- hardware failures
21Wireless NetworkingLessons Learned
- SCinet2000 offered a unique testbed to examine
wireless - large numbers of NAPs (38) and users (360) (100
in a room) - able to learn from staff that have deployed
wireless networks - 802.11b wireless Ethernet is nearly ubiquitous
- more than 250 attendees brought their own NICs
- should have even greater exposure in the future
- EMI is a serious problem
- will need exclusive control of the 2.4 GHz
frequency band - need gear to be able to search out noise sources
- Throughput performance had potentially
significant problems - acceptable performance for reading email, some
multimedia applications, and light web browsing - unacceptable for large file transfers over WANs
and some multimedia applications serious work
required a switched, wired connection