Heating, ventilation and air-conditioning (HVAC) is the largest consumer of electricity in commercial buildings. Consumption is impacted by group activities (e.g. meetings, lectures) and can be reduced by scheduling these activities at times and locations that minimize HVAC utilization. However, this needs to preserve occupants' thermal comfort and be responsive to dynamic information such as new activity requests and weather updates. This paper presents an online HVAC-aware occupancy scheduling approach which models and solves a joint HVAC control and occupancy scheduling problem. Our online algorithm greedily commits to the best schedule for the latest activity requests and notifies the occupants immediately, but revises the entire future HVAC control strategy each time it considers new requests and weather updates. In our experiments, the quality of the solution obtained by this approach is within 1% of that of the clairvoyant solution. We incorporate adaptive comfort temperature control into our model, encouraging energy saving behaviors by allowing the occupants to indicate their thermal comfort flexibility. In our experiments, the integration of adaptive temperature control further generates up to 12% of energy savings when a reasonable thermal comfort flexibility is provided.

Energy consumption in commercial and educational buildings is impacted by group activities such as meetings, workshops, classes and exams, and can be reduced by scheduling these activities to take place at times and locations that are favorable from an energy standpoint. This paper improves on the effectiveness of energy-aware room-booking and occupancy scheduling approaches, by allowing the scheduling decisions to rely on an explicit model of the building's occupancy-based HVAC control. The core component of our approach is a mixed-integer linear programming (MILP) model which optimally solves the joint occupancy scheduling and occupancy-based HVAC control problem. To scale up to realistic problem sizes, we embed this MILP model into a large neighbourhood search (LNS). We obtain substantial energy reduction in comparison with occupancy-based HVAC control using arbitrary schedules or using schedules obtained by existing heuristic energy-aware scheduling approaches.

SIGAI Doctoral Consortium. The Doctoral Consortium (DC) provides an opportunity for a group of Ph.D. students to discuss and explore their research interests and career objectives with a panel of established researchers in artificial intelligence. SIGAI provides travel funding to support students from institutions outside the U.S., while the NSF provides funding for U.S. students. This year, SIGAI supported four international students to attend and present at the 20th AAAI/SIGAI Doctoral Consortium. These students provided testimonials reporting on their experiences at the DC.

One of the main inefficiencies in building management systems is the widespread use of schedule-based control when operating heating, ventilation and air conditioning (HVAC) systems. HVAC systems typically operate on a pre-designed schedule that heats or cools rooms in the building to a set temperature even when rooms are not being used. Occupants, however, influence the thermal behavior of buildings. As a result, using occupancy information for scheduling meetings to occur at specific times and in specific rooms has significant energy savings potential. As shown in Lim et al. [15], combining HVAC control with meeting scheduling can lead to substantial improvements in energy efficiency. We extend this work and develop an approach that scales to larger problems by combining mixed integer programming (MIP) with large neighborhood search (LNS). LNS is used to destroy part of the schedule and MIP is used to repair the schedule so as to minimize energy consumption. This approach is far more effective than solving the complete problem as a MIP problem. Our results show that solutions from the LNS-based approach are up to 36% better than the MIP-based approach when both given 15 minutes.

My research focuses on developing innovative ways to control Heating, Ventilation, and Air Conditioning (HVAC) and schedule occupancy flows in smart buildings to reduce our ecological footprint (and energy bills). We look at the potential for integrating building operations with room booking and meeting scheduling. Specifically, we improve on the effectiveness of energy-aware room-booking and occupancy scheduling approaches, by allowing the scheduling decisions to rely on an explicit model of the building's occupancy-based HVAC control. From computational standpoint, this is a challenging topic as HVAC models are inherently non-linear non-convex, and occupancy scheduling models additionally introduce discrete variables capturing the time slot and location at which each activity is scheduled. The mechanism needs to tradeoff minimizing energy cost against addressing occupancy thermal comfort and control feasibility in a highly dynamic and uncertain system.

Cloud resource provisioning is crucial to assure timely deliverable of delay-sensitive cloud services. Today, virtual machine (VM) reservations are done mainly based on cloud resource availability. Often, maximum VM resources are preserved to assure service response time, resulting in a waste of resources. While various techniques have been proposed to perform cloud response time measurement, most of these methodologies involve deploying standard target applications on selected cloud infrastructure, gathering, and analyzing each individual dataset collected. Such methods are useful for offline analysis, but incur high overhead and are not useful for real-time performance measurement for delay-sensitive application. In this demo, we present a light-weight real time service latency prediction mechanism based on Euclidean Steiner Tree (EST) model for optimum VM resource allocation in delay-sensitive cloud services. Our aim is to derive a highly accurate service latency prediction mechanism in a short time reflecting timely information of the actual cloud resources conditions, while imposing minimum overheads to the cloud service itself. We shall present a fast response cloud resource estimation system FARCREST which integrates the prediction model with cloud front-end server for VM services latency prediction and deployment with production cloud experiment results.

Network latency is often used as an optimization parameter for network path construction over the Internet for various real-time applications. This paper proposes a high accuracy prediction tree method for latency estimation minimizing the need for intrusive mesh measurements. The network overlay of communication nodes is represented as a tree structure, called a prediction tree, with the latency of unmeasured network links predicted based on selected measured network links. We describe three novel heuristics that are the foundations of this high accuracy prediction tree, assisted by optimal target node selection and elimination of imprecise prediction steps. We have examined the proposed method based on publicly available data to ensure accuracy of high precision latency estimation process. Experiment results show that with 50% measurement, our proposed algorithm obtains 82% accuracy of latency prediction over a 120-node network.

Application Layer Multicast (ALM) enables packet replication and forwarding at the application layer (as opposed to other techniques, for example, Internet Protocol (IP) layer multicast) through an overlay tree. This packet forwarding capability is suitable for applications such as real-time audio-video (AV) content distribution since it provides independence from the lower layer (e.g. IP layer) while maintaining better end-to-end packet delivery times. ALM requires all the participating nodes in a given active session to be collectively responsible for forwarding the group's streaming data among the members via a set of chosen paths which make up the overlay tree.This chapter discusses the unique characteristics of ALM when used in consumer electronics, the design considerations that need to be thought about when creating such a system, and the considerations required when testing the system.

In this paper we present a testbed for the functional and performance verification and validation of product level overlay multicast (or ALM in short) applications which is a complex task due to humongous test patterns. Structured, systematic and simplified test environment is vital for product level quality validation. The key features of this testbed are network emulation, automated test scenario execution, log collection and testbed/real environment interconnectivity. The testbed uses StarBED as its mother testbed and netem as the network emulator. The key contributions of this work are calibration of netem in term of network emulation and establishing an architecture to use netem and StarBED for ALM system verification. The calibration of netem was carried out up to 30 pipes (logical links) and the results show the network emulation can be done at maximum 1.2% packet loss. One of the simple ALM verification experiments executed on the testbed took only 43 hours for execution completion compared to human execution which would take 15 weeks.

In this paper we propose bandwidth fair N-Tree algorithm for ALM distribution tree construction and a new protocol for ALM packet replication and distribution, namely Almcast. Both the tree construction algorithm and packet replication/distribution protocol were implemented as proof-of concept by modifying an existing multi-party video conference application. The results show that N-Tree algorithm takes less than 3ms to construct ALM distribution tree for 12 nodes. Almcast implementation enables the intermediate relay node to lookup for next destination, replicate and forward packets as fast as its receiving rate at application layer.

Universal plug and play (UPnP) is a widely accepted and most popular candidate for Consumer Electronics' (CE) critical function of devices and services discovery, eventing, access and control at home network (HNet), The impact of bursty and high bandwidth audio video (AV) streaming in HNet which is to be the common scenario is not much explored. Timely UPnP message advertisements and delivery are critical to ensure actual device or service status is propagated to all users or devices in HNet for optimal user experience. Home CE devices are critical in terms of processing power and memory. It is also common to have display devices such as television to be the monitoring and control device for other tiny devices which are incapable of displaying information. Therefore, this paper aims to briefly analyze the challenges of HNet and study the impact of AV streaming on UPnP advertisements by running an AV client and an AV server in a Ethernet-based HNet for AV content streaming while multiple UPnP devices are simulated in the same network to understand further the effectiveness of the message delay. Selected messages from UPnP 1.0 specification were used for this purpose on the UPnP AV testbed

With the growing use of broadband Internet, the demand for hardware-based intrusion detection system (IDS) is exploding. Network processor is poised to be the future platform for hardware-based IDS and firewall due to its programmability and capability to process packets at wire speed. In this paper, we explore the practical implementation of statistical-based SYN-flooding detection system in a network processor-based router. An embedded architecture, called synmon is proposed. We employ an instance of change-point detection, non-parametric Cumulative Sum (CUSUM) algorithm, for SYNflooding detection. It performs per-flow attack detection based on SYN and ACK packets exchanged in TCP friendly flow. A prototype of synmon embedded forwarder is developed and the performance of synmon under different attack patterns, network loads, sampling interval and tuning parameters is investigated. We demonstrate that the synmon architecture seamlessly integrates with common forwarding tasks while providing cost-effective service for SYN-flooding detection on network processor platform.

Distributed denial-of-service attacks remains inflict damage to the Internet services, after almost five years since its large-scale explosion. The demand for robust and high-speed firewall has led to the advent of hardware-based DDoS defense system. Network processor is becoming the cornerstone of many new firewall designs due to its programmability and high performance packet processing ability. In this paper, we propose an innovative and practical syn-flooding defense system built on network processor. An embedded architecture, called synmon is proposed. We characterize our solution as a source-based autonomous system which resides in upstream border routers. It detects wide-range of attacks and blocks large portion of attack traffic before flooding into core network. Change-point detection algorithm is employed to detect occurrence of syn-flooding attack. It performs per-flow attack detection based on SYN and ACK packets exchanged in TCP friendly flow. A fuzzy-based adaptive rate-limiting mechanism is proposed to restrict intensity of outgoing SYN packets. Under the per-flow mitigation scheme, while the attacker is penalized with limited outgoing connection, the legitimate clients in the same subnet are free from collateral damage. A hardware prototype of synmon embedded router is developed. We demonstrate that the synmon architecture seamlessly integrates with common routing tasks while providing cost-effective service for SYN-flooding defense system on network processor platform.