Summary of meetings Chris, Dorian, and Ian have with Maria. The meetings are about wireless power management. The meeting minutes document what we discuss the next action items.

Meetings

• next: Jan 28, 2003 @2:30-3:00, with Ian 3:00-3:30
• Jan 21, 2003
• Jan 16, 2003

Several points were made about the power management protocol.

• The protocol should stimulate cooperation among hosts.
• The protocol should use hints from the applications, such as checking the fullness of the buffer. What will the effect of buffer be on the application delay? 
• We should also consider the impact of packet loss when a receiving host walks out of range. The lost packets are wasted energy.
• The power management protocol will function at the application layer. 
• We have borrowed a laptop from Mark for the power consumption experiment.

We discussed some security schemes. 

• There are two possible schemes (1) Trusted third party and (2) implicit trust. 
• Private key system is more difficult to implement. 

Jan 16, 2003 [top]

Chris and Dorian showed the results from the wireless power measurements. We showed that the current consumption of the “DELL true mobile 1150” is similar to the “2.4GHz DSSS Lucent IEEE 802.11 WaveLAN” wireless card, as shown in Feeney’s results. We measured power consumption (V, I) for transmitting, receiving, and broadcast mode.

A suggestion to improve the experiment is to use UDP instead of TCP transactions. With UDP the sends and receives are isolated from each other because they are controlled by the programmer. With TCP the send and receive transactions are mixed as for every received packet an ACK is sent.

To better estimate power consumption, we should also investigate the difference in the wireless unicast and multicast protocols. The protocol that requires more sends and receives will consume more power.

We started to discuss the SPAN protocol. We want to contribute to reducing wireless power consumption by enhancing the SPAN protocol. The hypothesis is to modify the on interval to adapt power consumption. One question is if the on interval can be changed depending on the application.

With Ian we discussed securing the wireless protocol to protect from DoS (denial of service) attacks. Our solution is to encrypt the on/off intervals, which requires a security key. The question is how to distribute it among the members of the group?

An experimental question is how the on interval correlates to packet loss. UDP packets could be sent from a device to the wireless device. At different bit rates and on intervals, the number of dropped packets can be counted. It is also a question if the ATIM (Ad-hoc traffic indication message) window can be changed through the wireless-card driver API. 

A question to answer is if it is possible for a wireless user to belong to multiple groups. Belonging to multiple groups impacts the wireless power saving protocol.

Action items listed by priority (as understood by Dorian)

1. Gather ideas of how the SPAN algorithm can be enhanced. Can the algorithm be application dependent? Dorian thinks the on interval can be adapted according to the amount of network traffic, e.g. more traffic means longer and more frequent on-intervals. 
2. Investigate a secure method to communicate the variable on-interval pattern to all valid users of the MANET (Mobile ad-hoc network).
3. Continue the wireless power consumption experiments with UDP traffic. 
4. Measure the correlation between on-interval length, bit rate, and packet loss. The experiment would use a similar program to the power consumption experiment. 
5. In linux you can change the transmission power dynamically (i think there are four or five power transmission levels -- fixed). I wonder if you can do some experiments and investigate the actual energy spendings in each of these levels.