Chapter 2, Section 2.4 Josh Epps josh.epps@gatech.edu Chapter 2 Lecture Schedule Parts of this Presentation Part 1 - Basic knowledge from the Distributed Architecture Models Section from the text book Part 2 - Recent relevant knowledge on Distributed Architecture Models Part 3 - Predictions for future developments in Distributed Architectures Part 1 – Outline of Chapter 2, Section 2.4 2.4 Architecture Models 2.4.1 Distributed System Architectures 2.4.2 Communication Network Architecture What is a Distributed Architecture Model? An architectural model is an abstract view of a distributed system. Models are used to identify the specifications needed for a distributed system during the design process. This high-level abstraction describes the major components and how they interact within the system. An abstract model does not specify hardware components. For example, a distributed architecture model doesn't see computers, it sees processes and files that may need to be shared. It doesn't see routers and switches, it sees communication paths to other nodes. The Workstation-Server Model With the workstation-server model, normal user processes are performed at the workstations. Services provided by dedicated servers, such as file or database servers, respond to special requests from the workstations and return the results to the requesting workstation. User processes do not migrate to the servers for the requested activity in the workstation-server model. The Processor-Pool Model The processor-pool model attempts to provide substantial computing power for brief periods to users on-demand. For example, users may need to periodically compile a program with many files. In this model, a processor pool server accepts requests for an available processor from its pool. If a processor is available, the user process migrates to that machine until the task is completed. Each processor in the pool has its own memory to load and run a program. This is unlike the workstation-server model where processes do not migrate. [2] Communication Network Architecture Distributed system models must also include the communication network architecture. Distributed systems blur the distinction between operating systems and networking as fields of study. Considerations for the Communication Network Architecture include: • Network topologies (point-to-point, multipoint, bus-based, switched) • Network protocols (HTTP, FTP, TCP, UDP) • Interprocess communications • Process management algorithms Capacity and distance between system components will determine which applications will be supported by the network. For distributed systems, a useful metric is the ratio of propagation delay to transmission delay α = dp / dt Smaller α values indicates that the system components closer with respect to intercommunications. Lower α values are more suitable for interactive bursts of shorter messages. Larger α values lead to a more “communication oriented” system where sessions can be kept open on a semi-permanent basis. Part 2 – A Recent Study on Distributed Architecture “An Intelligent Agent-Based Distributed Architecture for Smart-Grid Integrated Network Management” [3] 1st IEEE Workshop on Smart Grid Networking Infrastructure, Denver, Colorado 2010 This paper describes a distributed software architecture designed to manage Smart-Grid networks. A little background on the Smart Grid What is “The Smart Grid?” The term “Smart Grid” refers to a more intelligent system for providing electrical power to customers. The Smart Grid will make use of Information and Communications Technologies (ICT) that will make decisions and perform operations on the electrical grid. Why do we need the Smart Grid? The Smart Grid will allow for enhancements to the electrical grid to include distributed energy sources such as wind and solar power. Smart meters at residences and businesses will allow billing data to be collected without an army of meter reader employees and can provide usage patterns for both the utility company and the customer. High-voltage transmission lines in the U.S. Smart Grid Advanced Metering Infrastructure (AMI) topology “An Intelligent Agent-Based Distributed Architecture for Smart-Grid Integrated Network Management” [3] Garcia et al. have proposed a distributed software architecture model that will take into account "massive deployments of AMI, DA, and HAN devices (meters, concentrators, sensors) from multiple vendor technologies." [3] These Smart Grid technologies must be accessed and controlled across multiple communications networks often across large geographic areas. Their proposed architecture is based on autonomous, intelligent, cooperative, distributed agents. They identified a need for Integrated Network Management (INMS) functions that provide management services for the Smart-Grid framework and identified the following functional blocks that should be present in an integrated network management system for Smart-Grids: Fault Handling, Events and alarm management Performance Management Security Management Configuration Management Device Management Integration Capabilities “An Intelligent Agent-Based Distributed Architecture for Smart-Grid Integrated Network Management” [3] The distributed architecture proposed for the Smart Grid is based on an Intelligent Agents Platform (IAP-SG) and is based on two agent management layers. Network Mediation Layer: Its main role is distributive processing of the massive data flows to and from the Smart Grid devices. Each IAP mediation device controls many disparate Smart Grid devices. These devices typically include AMI concentrators which may have hundreds of smart meters or perhaps groups of sensors at electrical substations. Management Applications Layer: This IAP layer consists of management applications running on the back-end systems. These applications perform operations over the network to implement the function blocks listed on the previous slide. Part 3 – A prediction for future developments in distributed architectures Within 5 years: ISP's will provide public cloud services to individuals as part of a premium internet bundle. These virtual machines at the ISP will allow customers using disparate host platforms to use common applications. This means Linux, Mac, and Android users could use MS Office and C#. A terabyte of persistent storage will also be included. References [1] Randy Chow and Theodore Johnson, Distributed Operating Systems and Algorithms, Adison-Wesley 1997, ISBN 0-201-49838-3 [2] A. Fedorova. (Summer 2007). Architectural Models [Online]. Available: http://www.cs.sfu.ca [3] A. P. Garcia et al,. "An Intelligent Agent-Based Distributed Architecture for Smart-Grid Integrated Network Management," 2010 IEEE 35th Conference on in Local Computer Networks (LCN), Denver, CO, 2010, pp. 1013-1018.