Distributed Systems: Concepts and Models

• The lecture discusses distributed systems, their advantages, disadvantages, transparency, scalability, and various OS models.

Definition of a Distributed System

• A distributed system involves multiple connected CPUs working together.
• It appears to users as a single coherent system, despite consisting of independent computers.
• Examples includes parallel machines and networked machines.

Advantages and Disadvantages

• Communication and resource sharing become possible within a distributed system.
• Economics are improved via the price-performance ratio.
• Reliability and scalability are benefits.
• There is potential for incremental growth.
• Distribution-aware programming languages (PLs), operating systems (OSs), and applications are needed.
• Network connectivity is essential.
• Security and privacy are main concerns.

Transparency in a Distributed System

• Access transparency hides the differences in data representation and how a resource is accessed.
• Location transparency hides where a resource is located.
• Migration transparency hides that a resource may move to another location.
• Relocation transparency hides that a resource may be moved to another location while in use.
• Replication transparency hides that a resource may be replicated.
• Concurrency transparency hides that a resource may be shared by several competitive users.
• Failure transparency hides the failure and recovery of a resource.
• Persistence transparency hides whether a software resource is in memory or on disk.

Open Distributed Systems

• Services are described a priori
• Syntax and semantics are known via protocols.
• Services are specified via interfaces.
• Benefits include interoperability and portability.
• Open systems should evolve and be extensible over time to accommodate new functionality.
• Policy should be separate from mechanism.

Scalability Problems

• Centralized services are a single server for all users.
• Centralized data is a single online telephone book.
• Centralized algorithms perform routing based on complete information.

Scaling Techniques

• Principles for good decentralized algorithms include:
  • No machine having a complete state.
  • Decisions are made based on local information.
  • A single failure does not bring down the system.
  • No global clock.
• Techniques:
  • Asynchronous communication.
  • Distribution.
  • Caching and replication.

Distributed Systems History and OS Models

• Minicomputer Model: Early networks where each user has a local machine with local processing but can fetch remote data.
• Workstation Model: Processing can also migrate.
• Client-Server Model: Users have local workstations, and powerful workstations serve as servers.
• Processor Pool Model: Back-end processors handle processing for terminals.

Distributed System Models (contd)

• Cluster Computing Systems/Data Centers: LAN with a cluster of servers and storage used for distributed web servers, scientific applications, and enterprise applications.
• Grid Computing Systems: Clusters of machines connected over a WAN.
• WAN-Based Clusters/Distributed Data Centers: Such as Google, Amazon.
• Virtualization and data center and cloud computing are modern approaches.

Emerging Models

• Distributed Pervasive Systems: features smaller nodes with networking capabilities where computing is "everywhere".
• Home networks, mobile computing with smartphones and car-based PCs, sensor networks, and healthcare applications.
• Sustainability is another design goal.

Uniprocessor Operating Systems

• An OS acts as a resource manager or arbitrator, managing CPU, I/O devices, and memory.
• Provides a virtual interface easier to use than hardware.
• Structure:
  • Monolithic with one large kernel.
  • Layered design where functionality is decomposed into N layers.

Microkernel Operating Systems

• Microkernel Architecture: Features small kernel and user-level servers that implement additional functionality.

Distributed Operating System

• Manages resources in a distributed system seamlessly and transparently to the user.
• Presents a virtual uniprocessor to users even though it operates on multiple independent CPUs.
• Provides transparency in location, migration, and concurrency.

Types of Distributed OSs

• DOS (Distributed Operating System): Tightly-coupled OS for multiprocessors and homogeneous multicomputers to hide and manage hardware resources.
• NOS (Network Operating System): Loosely-coupled OS for heterogeneous multicomputers to offer local services to remote clients.
• Middleware: Additional layer atop of NOS implementing general-purpose services to provide distribution transparency.

Multiprocessor Operating Systems

• Acts like a uniprocessor OS and manages multiple CPUs transparently.
• Each processor has its own hardware cache, maintaining consistency.

Network Operating System

• Employs a client-server model with a minimal OS kernel and additional functionality as user processes.

Middleware-Based Systems

• General structure of a distributed system as middleware which sits between distributed applications and network OS services.