Engineering
Global

Engineering Experts

Tal Lavian

Scientist
Industrial Engineering
University of California, Berkeley (UCB)
United States of America

Biography

Dr. Tal Lavian focuses his research on telecommunication, wireless, and Internet technologies. He investigates applications that incorporate computation, communications, and mobile devices. He translates cutting edge research into practical products while integrating science, engineering, and innovation. Dr. Lavian has over 25 years of experience in the fields of computer science, electrical engineering, and telecommunications. He is currently a Berkeley CET Industry Fellow and lecturer at the UC Berkeley College of Engineering. Dr. Lavian has been listed as co-inventor on over 80 patents, issued and pending, and he has co-authored over 25 scientific peer-reviewed papers. Additionally, he is a Principal Scientist at TelecommNet Consulting and a Senior Member of IEEE. Dr. Lavian also participates as an active member in ACM and IEEE-CNSV. Previously, Dr. Lavian led—as the Principal Investigator—three US Department of Defense (DARPA) projects. He also served as a visiting scientist at UC Berkeley's RAD Lab and as a Technical Co-Chair for IEEE Interconnects at Stanford University. Dr. Lavian successfully spearheaded development of the first network resource-scheduling service for Grid computing, and he was the first to demonstrate Transatlantic dynamic allocation of 10Gbs Lambdas as a grid service. He created the first demonstrated wire-speed active network device on commercial hardware. Dr. Lavian graduated from U.C. Berkeley with a Ph.D. in Computer Science with a specialization in networking and communications. He received his Master of Science in Electrical Engineering and Bachelor of Science in Math and Computer Science from Tel Aviv University.

Research Interest

GPAN - The Grid Proxy Architecture of Networks GPAN (the Grid Proxy Architecture of Networks) aims to deploy Grid-aware networking technology onto existing commercial network devices. Those network devices—such as switches and routers—are not equipped with Grid-ware or the Grid-enabled software that support computing-intensive and data-intensive Grid services. GPAN exposes the unique proxy architecture through integrating DRAC network services and Grid/OGSA/WSRF Web Services under service coordination and resource policy. The outcome is that Grid applications such as GridFTP and GRAM can effectively make reservations on network resources such as bandwidth on demand, dynamic optical links, Ethernet VLANs and IP routes—just like reservations on computer and storage resources. GPAN has been successfully demonstrated in SuperComputing and Globus World. DWDM-RAM DWDM-RAM is a DARPA-funded research project that aims to support data-intensive Grid computing services through advanced optical networking. The primary goal of this Grid-enabled architecture is to enable efficient support for data-intensive Grid applications, which may require moving—without prior notice—Terabytes or even Petabytes of data among multiple sites. The DWDM-RAM solution is to orchestrate data flows with many different types of characteristics, including those requiring exceptionally high bandwidth for sustained periods of time. These high performance data flows are provided by dedicated optical lightpaths (lambdas or wavelengths), which are dynamically established. The DWDM-RAM system implementation leverages the state of the art DWDM technology with dynamic wavelength switching, which enables the creation of Grid services that allocate and release these light-paths either on-demand or by advance reservation. The DWDM-RAM architecture relies on a new network service that provides options for allocating dedicated network resources. This service discovers the network topology, explores the availability of network resources, and optimizes the schedule and availability of the optical network resources. It also presents a standardized, high-level, and network-accessible interface. A natural choice for implementing this interface is the Open Grid Service Interface (OGSI). Such interfaces are compliant with the GGF's OGSA specification, and they conform to widely used Web Services standards (WSDL, SOAP, and XML). DRAC - The Dynamic Resource  Allocation Controller Application-aware network technology creates a high degree of coupling between user applications and transport networks in order to improve user experience of networking and optimize equipment investments and operational expenses. This allows an application to obtain a share of network resources under service level agreement and to drive its share of network resources within a policy-based control flexibility. Network resources are on-demand, and can include bandwidth, quality of service (QoS), security, acceleration appliances, sensors, and various communication services. DRAC (Dynamic Resource Allocation Controller) exposes this kind of application-aware network system for dynamic allocations of network resources in advance and in real-time. DRAC is the middleware between data applications and data transport networks. Typical data applications include large data transfer, data-intensive Grid computing, sensor-triggered disaster data evacuation and restoration, and emergent data processing for healthcare. Data transport networks involve carrier and enterprise networks over L1 (DWDM/OXC, ASTN) optical, L2 (Ethernet switch, RPR and SONET, VLAN, VPLS) and L3 (IP routing and VPN router). DRAC is released as a Nortel product that has control and allocation capabilities with Nortel and other network products, including optical cross-connect, Ethernet switch and IP routers. DRAC has been applied in major backbone networks such as Internet2 and SURFnet6. CO2 - Content Over Optics CO2 (Content Over Optics) is a cutting-edge technology that combines content delivery directly over the optical transport network. Its goals are to provide content-aware networking capability and to transport content leveraging to the bandwidth-abundant fiber channels. CO2’s technological advantages include content processing and redirection, media streaming and data replication at network edge, packet/protocol inspection, secure network transport, network resource allocation, and dynamic setup of optical links. The CO2 system achieves content-aware network resource control and user management for data dissemination and storage applications. ORE - Oplet Runtime Environment for Open Programmable Networking Bringing programmability to a network makes the data transport network accessible to every network user, from an application developer to an Internet service provider or a business service provider. ORE (Oplet Runtime Environment) is an open, neutral, and programmable platform by which a user can execute network service development and deployment. ORE provides a software development toolkit for customized network service creation and development and a runtime environment for dynamic network service deployment onto commercial network devices. ORE is implemented in Java and C/C++, and has its development environments on Linux and Windows. ORE runtime has been deployed on commercial network devices such as Nortel Networks Passport 8600 and Accelar 1000 Routing Switches. ORE supports dynamic service compositions, and it provides common network service features such as routing, forwarding, filtering, QoS/Diffserv and policing. ORE supports customer network services such as Active Networks and ANTS. Active Networks On Commercial Network Devices Active Networks is a DARPA-funded research effort that aims to explore the next generation of inter-network break-through technologies in terms of a programmable and user-oriented approach. The key technological concepts behind Active Networks are active packets and dynamic network protocol composition. As a result, a user can define a new network protocol with its own data format and processing. An ISP can inject network services onto the network "on the fly". One key challenge is to enable the support of Active Networks on commercial network devices. Our investigation leads to two improvements on a commercial network device. One is that the device has the network programmability to provide dynamic service APIs, and the other is that the device installs an Active Networks core to dispatch active packets. Finally, an Active Networks-enabled Gigabit Ethernet switch is successfully built up and demonstrated in the DARPA Active Networks Conference and Exhibition (DANCE). This project is a collaboration effort with other research institutions including Columbia University, University of California, Los Angles, University of California, Berkeley, University of Utah, University of South California and University of Washington at St. Louis.

Global Experts from United States of America

Global Experts in Subject

Share This Profile
Recommended Conferences