[Dataset] kaist/wibro (v. 2008-06-04)

top

the initial version

@MISC{kaist-wibro-2008-06-04,
  author = {Mongnam Han and Youngseok Lee and Sue B. Moon and Keon Jang and Dooyoung Lee},
  title = {{CRAWDAD} data set kaist/wibro (v. 2008-06-04)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/kaist/wibro},
  month = jun,  
  year = 2008
}
					

In order to evaluate QoS of VoIP applications over the WiBro network, 
we collected the CBR and VoIP traffic from the WiBro network in Seoul, Korea.
We conducted experiments in both stationary and mobile scenarios, e.g., on a subway train.

WiMax is a subset of the 802.16 standards whose main goal is product 
compatibility and interoperability of BWA products, just as WiFi is to 
the 802.11 standards. WiBro has been developed as a mobile BWA solution 
in Korea, and is generally considered a precursor to WiMax. It is a subset 
of consolidated version of IEEE Standard 802.16-2004 (fixed wireless 
specifications), P802.16e (enhancements to support mobility), and 
P802.16-2004/Cor1 (corrections to IEEE Standard 802.16-2004). 
The profiles and test specifications of WiBro will be harmonized 
with WiMAX Forum's mobile WiMAX profiles and test specification, 
drawing a convergence of the two standards.

Today's Internet users not only write emails and surf the web, but also make
Voice over IP (VoIP) calls, play online games, and watch streaming media. These
real-time applications have stringent Quality of Service (QoS) requirements on
delay and loss. WiMax and WiBro standards have defined multiple service types
in order to guarantee different levels of QoS. However, at the initial phase of
deployment, often only the best-effort service is made available, while users do
not limit themselves to emails and web surfing over emerging wireless technology
networks.

We conduct experiments to evaluate QoS of VoIP applications over the WiBro 
network. In order to capture the baseline performance of the WiBro network, 
we measure delay, loss, and throughput of constant bit rate streams in both 
stationary and mobile scenarios.

In Korea, KT (formerly, Korea Telecom) launched WiBro coverage for nine 
subway lines in Seoul on April, 2007. The Seoul subway system moves 
millions of people a day through an extensive network that reaches almost 
all corners within the city and major satellite cities outside. 

The maximum speed of Seoul subway trains is 90 km/h, and it takes 
about 1 - 2 minutes between two stations. We have considered measurement 
experiments in vehicles moving at or under 60 km/h, the upper limit of 
WiBro, but chosen the subway, as it presents a more popular scenario with users.
It has 38 stations over a total distance of 35.1 km and six RASs (Radio Access
Stations).

Commuters in subway are more likely to use mobile devices than those 
in moving vehicles, as the measurement experiment on a subway train is 
easier for us. We have conducted our measurement experiments on subway 
line number 6. 

We have placed a mobile node (a laptop with a WiBro modem) 
in the WiBro network and installed a stationary node (a desktop PC) 
connected to the Internet over a fixed line so that we could focus 
on the WiBro network performance.

[Traceset] kaist/wibro/seoul (v. 2008-06-04)

top

the initial version.

@MISC{kaist-wibro-seoul-2008-06-04,
  author = {Mongnam Han and Youngseok Lee and Sue B. Moon and Keon Jang and Dooyoung Lee},
  title = {{CRAWDAD} trace set kaist/wibro/seoul (v. 2008-06-04)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/kaist/wibro/seoul},
  month = jun,  
  year = 2008
}
					

We collected the CBR and VoIP traffic from the WiBro network in Seoul, Korea
in stationary and mobile scenarios, e.g., on a subway train.

We have placed a mobile node (a laptop with a WiBro modem) 
in the WiBro network and installed a stationary node (a desktop PC) 
connected to the Internet over a fixed line so that we could focus 
on the WiBro network performance. 

We refer to the laptop as the Mobile Node (MN) and the PC as the Corresponding
Node (CN). In order to place the CN as close to the WiBro network as possible, 
we use a PC directly connected to a router on Korea Research Environment Open 
Network (KREONET). It is a research network that interconnects super computing 
centers in Korea and also is used as a testbed for new networking technologies. 
It peers with KT's IP backbone network at one of KT's exchange points.

For our measurement experiments, we generate two types of traffic: constant
bit rate (CBR) and VoIP. The difference between CBR and VoIP traffic lies in
the packet sending rate and follow-up analysis. For both types of traffic, 
we take measurements when the MN is stationary and moving in a subway. We use 
iperf for CBR traffic generation, and D-ITG for VoIP traffic generation. 
We configure D-ITG to measure round-trip time (RTT) instead of one-way delay,
as we could not instrument the MN in subway and CN at an exchange point to
have access to GPS-quality clock synchronization.

Multiple types of handoff are possible in the WiBro network. An inter-ACR 
(Access Control Routers) handoff takes longer than inter-RAS (Radio Access Stations)
or inter-sector handoff. An inter-sector handoff is between two sectors 
within an RAS. An RAS typically has three sectors. Using a custom tool developed 
to monitor inter-sector and inter-RAS handoffs, we collect RAS identifiers and 
corresponding sector identifiers. By aligning the changes in RAS and sector 
identifiers with the measurement data, we can pinpoint the moments of handoffs 
in our data.

[Trace] kaist/wibro/seoul/UDP (v. 2008-06-04)

top

the initial version

@MISC{kaist-wibro-seoul-UDP-2008-06-04,
  author = {Mongnam Han and Youngseok Lee and Sue B. Moon and Keon Jang and Dooyoung Lee},
  title = {{CRAWDAD} trace kaist/wibro/seoul/UDP (v. 2008-06-04)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/kaist/wibro/seoul/UDP},
  month = jun,  
  year = 2008
}
					

We conducted CBR measurements from the WiBro network in Seoul, Korea. 
For stationary experiments, we placed the mobile node on KAIST 
Seoul campus. For mobile experiments, we rode the Seoul subway line 6.

On October 5th and 12th, 2007, we took CBR measurements in Seoul.
For stationary experiments, we placed the MN (the mobile node) on KAIST 
Seoul campus. For mobile experiments, we rode the Seoul subway line 6. 

We used iperf for CBR traffic generation. For traffic logging, we used windump 
at both the MN and CN (the corresponding node). 

In order to capture the baseline performance of the WiBro network, we first measure
the maximum achievable throughput. We generated 5 Mbps up to 6 Mbps and 1.5 Mbps 
up to 2.5 Mbps traffic in quantums of 100 Kbps for download and upload, respectively, 
and found the bandwidth capped at about 5.3 Mbps downlink and 2 Mbps uplink.

Then we set the transmission rate of our CBR traffic at 5.3 Mbps for downlink
and 2 Mbps for uplink with the packet size of 1460 bytes and saturated the link.
We conducted 10 sets of 300-second-long uploads and downloads.

The "UDP" directory contains UDP wibro pcap files and bsid (Base Station ID) log files. 
It has two subdirectories "stationary" and "subway":

stationary - collected with MN (the mobile node) placed at a stationary location,
subway - collected with MN placed in subway.

In each subdirectory, the format of file name is as follows: 
kw_seoul_<SM>_udp_cbr_<SRATE>_iperf_<UD>_<CONF>_<CS>_<DATE>.<EXT>

SM: where to place the mobile node. "st" (stationary) or "mb" (mobile).
SRATE: sending rate, e.g., "1500~2500kbps".
UD: "up" (upload) or "dn" (download).
CONF: experiment configuration. For example, "12x5x120sec" indicates 
	that 12 different CBR throughputs are used, and the experiment 
	repeats 5 times with the duration of 120 seconds for each experiment.
	"10x300sec" indicates that the experiment repeats 10 times with the
	duration of 300 seconds.
CS: logging location. "cl" (the mobile node) or "sv" (the stationary node). 
DATE: YYYYMMDD.
EXT: "pcap" or "bsid.log".

You can find the following files:

1. pcap files (*.pcap) : pcap files captured by windump

2. bsid log files (*.bsid.log) : the log of bsid (Base Station ID)

Only "subway" directory contains bsid log files. 
Each line contains a UNIX timestamp, a human readable time (KST), and BSID, 
which are seperated by a tab. This data is used to locate when handoff has 
occurred.

[Trace] kaist/wibro/seoul/VoIP (v. 2008-06-04)

top

the initial version

@MISC{kaist-wibro-seoul-VoIP-2008-06-04,
  author = {Mongnam Han and Youngseok Lee and Sue B. Moon and Keon Jang and Dooyoung Lee},
  title = {{CRAWDAD} trace kaist/wibro/seoul/VoIP (v. 2008-06-04)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/kaist/wibro/seoul/VoIP},
  month = jun,  
  year = 2008
}
					

We conducted VoIP measurements from the WiBro network in Seoul, Korea. 
For stationary experiments, we placed the mobile node on KAIST 
Seoul campus. For mobile experiments, we rode the Seoul subway line 6.

On October 5th, 6th, 7th, 12th and 13th 2007, we took VoIP measurements in Seoul.
For stationary experiments, we placed the MN (the mobile node) on KAIST 
Seoul campus. For mobile experiments, we rode the Seoul subway line 6. 

We use D-ITG for VoIP traffic generation. You can download D-ITG at
http://www.grid.unina.it/software/ITG/

For logging, the MN and CN (the corresponding node) also dumped log files 
including sequence numbers, packet departure times, acknowledgement arrival 
times, and calculated round trip time. 

We have generated voice traffic that has the same characteristics of the G.711 
voice codec without Packet Loss Concealment (PLC). The payload size is set 
to 160 bytes and the sending interval to 20 ms in G.711 codec without PLC. 
The resulting throughput of VoIP traffic is 64 Kbps. We collected 10 300-second-long 
data sets after transmitting voice packets between the MN and the CN. 

Because the clocks on the MN and CN were not synchronized, we could not measure 
the one-way delay accurately. Instead, we took round-trip measurements of VoIP 
traffic, and halved the delay. Due to the difference in uplink and downlink bandwidths, 
half the round-trip delay is likely to be larger than the one-way uplink delay. 
However, the WiBro link was very lightly loaded and thus we assume the difference in
transmission delay to be minimal.

The "VoIP" directory contains D-ITG log files for VoIP traffic, pcap trace 
for VoIp traffic, and bsid (Base Station ID) log files. It has two subdirectories 
"stationary" and "subway":

stationary - collected with MN (the mobile node) placed at a stationary location.
subway - collected with MN placed in subway.

In each subdirectory, the format of file name is as follows: 
kw_voip_<CODEC>_<UD>_<CS>_<EXPNO>.<EXT>

CODEC: type of codec used. 
UD: "up" (upload) or "dn" (download).
CS: logging location. "cl" (the mobile node) or "sv" (the stationary node). 
EXPNO: experiment number ("all" or 1-10).
EXT: extension. "log", "voip", "pcap", or "bsid.log" 

You can find the following files:

1. D-ITG binary files (*.log) : binary log files collected with D-ITG software.

2. D-ITG text files (*.voip) : text version of D-ITG log files converted using D-ITG software.

Each line contains  seq #, send time, recv time, rtt, size(=168).

3. pcap files (*.pcap) : pcap files captured by windump

4. bsid log files (*.bsid.log) : the log of bsid (Base Station Identification)

Only "subway" directory contains bsid log files. 
Each line contains a UNIX timestamp, a human readable time (KST), and BSID, 
which are seperated by a tab. This data is used to locate when handoff has 
occurred.

[Author] Mongnam Han

top

[Author] Youngseok Lee

top

[Author] Sue B. Moon

top

[Author] Keon Jang

top

[Author] Dooyoung Lee

top

[Paper] han-wibro

top

In this work, we have conducted experiments to evaluate QoS of VoIP 
applications over the WiBro network. In order to capture the baseline 
performance of the WiBro network we measure and analyze the characteristics of 
delay and throughput under stationary and mobile scenarios. Then we evaluate 
QoS of VoIP applications using the E-Model of ITU-T G.107. Our measurements 
show that the achievable maximum throughputs are 5.3 Mbps in downlink and 2 
Mbps in uplink. VoIP quality is better than or at least as good as toll quality 
despite user mobility exceeding the protected limit of WiBro mobility support. 
Using RAS and sector identification information, we show that the handoff is 
correlated with throughput and quality degradation.