1. CONSERN Concepts and Techniques in Action – A Real Life Experiment P. Magdalinos, P. Spapis, G. Katsikas, K. Chatzikokolakis, P. Makris Department of Informatics and Telecommunications National and Kapodistrian University of Athens Athens, Greece {panagis, pspapis, katsikas, kchatzi}@di.uoa.gr, petmakris@phys.uoa.gr Obtained Results Self-evolving Cognitive & Autonomic Networking Laboratory http://scan.di.uoa.gr Internal Architecture and Expected Behaviour Experimental Assessment Goals and Setup Assessment Goals Assess the validity and viability of key CONSERN concepts and techniques. Extract results from a real world environment (i.e. Small Office) Implemented Algorithms Access Points Clustering: APs select a Cluster Head which manages the network. Self-Organization based on topology information derived from local sensing modules. System Idle Time Estimation: APs autonomously adjust their scanning rate following the rate of appearance of events. Cooperative Power Control: APs collaboratively decide the optimal transmission rate in order to reduce overall interference. Access Points On/Off: The Cluster Head periodically evaluates the overall usage of network resources and directs APs to switch on or off. Access Points Channel Selection: Each AP autonomously decides its transmission channel. Experiment Setup Four linux-based router-APs (Fig.2) deployed in the building (Fig. 1) Two wireless interfaces, One is deployed as an Access Point, The other is used to monitor the environment, CONSERN algorithms, implemented in Java, are running in the background and adjust the operation of the device. Network Setup APs are calibrated at 5.5Mbps, APs are interconnected through core network (Fig. 3). ~15 researchers accessing the Internet through these APs (Fig 4) Commodity laptops – no special equipment, Mobile devices and tablets, Aggregated network traffic raged from 1 to 10 Mbps, User needs varied, from internet browsing to video streams and torrent downloads. Duration Experiment was initiated on May 28 th, 11:00 CET, Experiment ended on May 28 th, 23:00 CET. Fig. 1Fig. 2 Fig. 3Fig. 4 Detection of new TxPower Communication with all APs Carrier Sensing, AP Operational status (associated terminals, used & available capacity) Algorithm for Channel Selection during Activation RF (De)Activation, Tx Power Adjustment, HostAP, DHCP initiation Routing rules, Channel Deployment Local status storage Deactivation & Reactivation Thresholds Update of AP On/Off Scheme Periodic communication with APs CE (Receive Monitoring Data, Send Conf. Actions) AP Topology graph & Operational Status Detection and Selection of AP for On/Off Red Color signifies common blocks of cluster-member and cluster-head APs Blue Color signifies functionalities implemented only by a cluster-head AP All-in-One 1.Device initiates CONSERN modules Monitoring with the use of System Idle Time Estimation, AP autonomously selects transmission channel and avoid interfering other APs. 2.Device is fully active Fuzzy logic based decision making exploits monitoring measurements and identifies potential problematic situations (high load, high interference), Access Points Clustering takes place, Cooperative Power Control is initiated trying to minimize interference, The cluster head periodically evaluates network conditions and directs access points to switch On/Off. 3.CONSERN modules are de-activated All threads are suspended and the RF is shut down. Expected Behaviour Access Points Clustering: Selection of a device which will undertake the task of network organization and its holistic management – Deployed during 2. System Idle Time Estimation: APs will scan based on the rate of events appearance; Deployed during 1, 2. Cooperative Power Control: Tx Power will be adjusted according to network conditions. Deployed during 2. Access Points Switch On/Off: APs will be directed to switch their RFs off or on according to network conditions and coverage requirements. Deployed during 2 on the cluster head. Access Points Channel Selection: APs are expected to operate in different, non-overlapping –if possible- channels. Deployed during 1. AP 1AP 2AP 3AP 4 ECP47%36% 37% SDP91%96%98%97% Requirement to switch on device – None available An AP is switched off Requirement to switch on a device People start to leave. Devices are progressively switched off A Single Device (the Cluster Head) serves all network System Idle Time Estimation Main concept Adjustable thread sleep time. Assessment Metric SDP: Success Detection Percentage = #Events Detected / #Events Detected without sleep time adjustment ECP: Energy Consumption Percentage = #Loops Performed / #Loops Performed without sleep time adjustment Perform 50% less loops while detect more than 90% of events StateEnergy Requirements All interfaces down7.05 W Platform deactivated7.79 W Platform fully operational 9.50 W Platform fully operational and AP serves 500Kbps traffic 9.73 W AP Switch On/Off Main Concept Switch devices on - off (network requirements). A network interface induces a 10.5% energy load when activated. Platform initiation and functioning results in a 34.75% energy overhead (JVM and 2 WiFi interfaces) Deactivating the platform and the interfaces results in 20% energy saving on each device. Aggregated gain because of RF de-activation: 4 APs from 18:00 to 23:00 require 700.56 KJoules 1 AP and 3 with their RFs suspended require 595.80 KJoules Network requires 15% less energy Channel Selection Main concept APs are autonomously deployed on channels that impose minimum interference Result AP #1: Channel 1 AP #2: Channel 6 AP #3: Channel 11 AP #4: Channel 3 The maximum possible separation AP Clustering Main concept APs based on monitoring information, autonomously elect a device which will manage the whole network and trigger the initiation of Cooperative Power Control and AP Switch On/Off algorithms Result AP #1: Channel 1 AP #2: Channel 6 AP #3: Channel 11 AP #4: Channel 3 AP #4 is elected as cluster head Cooperative Power Control Main concept Asynchronous and collaborative Tx Power adjustment in order to minimize interference Assessment Metric The TxPower of each APs antenna Maximum attainable TxPower by each AP is 27dbm (500mW) AP Operates with a 15%-35% reduction in its TxPower from 12:30-14:30 and 15:30-19:00 AP Operates with a 25% reduction in its TxPower from 20:00-23:00 (single device) Device is directed to switch off (i.e. deactivate CONSERN modules and wireless interfaces) TxPower adjustment from 11:15-12:30. AP operates with a 15% - 45% reduction in its TxPower Adjusting the TxPower of an AP at the 75% of its maximum capacity for 5 hours (modest estimation for AP #4) can save 250KJoules AP #1 AP #4