COPY: Distributed Systems (3rd Edition) - Chapter 01: Introduction (Selected Slides)

Distributed Systems

• Primary characteristic: Loosely coupled, autonomous nodes that communicate and coordinate to achieve a common goal.
• Implication of individual node time: Each node has its own notion of time, leading to potential inconsistencies and difficulties with synchronizing events across the system.
• Node communication: Nodes typically communicate with each other through message passing over a network.
• Node relationship: Nodes are interconnected and indirectly dependent on each other; they collectively form a single system.
• Overlay Network purpose: Enhances communication and resource sharing among nodes, creating a separate virtual network structure on top of the underlying physical network.
• Challenge with autonomous nodes: Ensuring consistent data and behavior across the system while maintaining independence.
• Problem with replication: Maintaining consistency across all replicas of data, especially in the presence of failures or network issues.
• Requirement for global synchronization: Ensuring that updates to replicated data are applied consistently across all nodes, necessitating a coordinated approach.
• Preclusion of global synchronization: Precludes allowing changes to be made independently on different nodes, as it would require a global consensus to ensure consistency.
• Reducing global synchronization: Employing techniques like optimistic concurrency control to manage data consistency without needing full global synchronization for every update.
• Application dependency: The specific methods used for managing data consistency and handling failures are often dependent on the specific application's requirements and the nature of the distributed system.
• Inconsistency due to multiple copies: Multiple copies of data can lead to inconsistencies if updates are not managed carefully, especially during failures or network partitions.
• System operating independently of user interaction: A key character of a distributed system is that the system operates the same way regardless of user interaction, ensuring continuous and resilient operation.
• Not a well-known example of an overlay network: Centralized social media platforms like Facebook or Twitter.
• Middleware’s role: Provides abstractions and services to simplify the development and management of distributed applications.
• Challenge in hiding partial failures: Determining which nodes are actually experiencing failures or network problems, as opposed to being temporarily unavailable due to network latency or other transient issues.
• Not a characteristic of a structured overlay network: Decentralized and dynamic node placement, relying instead on predefined patterns or structures for node organization and communication.
• Middleware’s design goal: Simplify application development and communication by handling the complexities of distributed environments.
• Not a component of scalability in distributed systems: Network bandwidth.
• Challenge modern systems face: Balancing scalability with consistency, availability, and fault tolerance.
• Typical solution for size scalability: Scaling out by adding more nodes to the system.
• Not a key component of scalability in distributed systems: Processing speed of a single node.
• Use of 'scalable' without explanation: Many developers use the term without clearly explaining how their system actually handles size scalability, performance under load, or distribution of workload.
• Canonical example of supporting resource sharing: File systems like Hadoop Distributed File System (HDFS) or Google File System (GFS) allow multiple nodes to access and share files.
• 'The network is the computer': Refers to the idea that distributed systems treat the network as a fundamental computational resource, leveraging its capabilities for data storage, processing, and communication.
• Type of distribution transparency: Location transparency, where applications can access resources without knowing their physical location.
• Challenge with full distribution transparency: Requires complex mechanisms to hide the complexities of the distributed system, including network communication, failure handling, and data consistency.
• Disadvantage of full distribution transparency: Can lead to performance overheads and complicate debugging, as the underlying distributed nature of the system is hidden from developers and users.
• Exposure of distribution when dealing with time zones: May benefit from exposing the distribution of the system, allowing users to understand the time difference when interacting with resources in different locations.
• Cannot be hidden when aiming for full distribution transparency: Failures, as they are a fundamental aspect of distributed systems and cannot be completely masked from the system's operation.
• Challenge in distinguishing slow computers from failing ones: Requires sophisticated mechanisms to monitor and diagnose node performance, making it difficult to determine if a slow response is due to a temporary network issue or a real node failure.