book

Designing Distributed Systems

by Brendan Burns

February 2018

Intermediate to advanced

162 pages

4h 8m

English

O'Reilly Media, Inc.

Read now

Unlock full access

Who Should Read This BookWhy I Wrote This BookThe World of Distributed Systems TodayNavigating This BookConventions Used in This BookOnline ResourcesUsing Code ExamplesO’Reilly SafariHow to Contact UsAcknowledgments
A Brief History of Systems DevelopmentA Brief History of Patterns in Software DevelopmentFormalization of Algorithmic ProgrammingPatterns for Object-Oriented ProgrammingThe Rise of Open Source SoftwareThe Value of Patterns, Practices, and ComponentsStanding on the Shoulders of GiantsA Shared Language for Discussing Our PracticeShared Components for Easy ReuseSummary
MotivationsSummary
An Example Sidecar: Adding HTTPS to a Legacy ServiceDynamic Configuration with SidecarsModular Application ContainersHands On: Deploying the topz ContainerBuilding a Simple PaaS with SidecarsDesigning Sidecars for Modularity and ReusabilityParameterized ContainersDefine Each Container’s APIDocumenting Your ContainersSummary
Using an Ambassador to Shard a ServiceHands On: Implementing a Sharded RedisUsing an Ambassador for Service BrokeringUsing an Ambassador to Do Experimentation or Request SplittingHands On: Implementing 10% Experiments
MonitoringHands On: Using Prometheus for MonitoringLoggingHands On: Normalizing Different Logging Formats with FluentdAdding a Health MonitorHands On: Adding Rich Health Monitoring for MySQL
Introduction to Microservices
Stateless ServicesReadiness Probes for Load BalancingHands On: Creating a Replicated Service in KubernetesSession Tracked ServicesApplication-Layer Replicated ServicesIntroducing a Caching LayerDeploying Your CacheHands On: Deploying the Caching LayerExpanding the Caching LayerRate Limiting and Denial-of-Service DefenseSSL TerminationHands On: Deploying nginx and SSL TerminationSummary
Sharded CachingWhy You Might Need a Sharded CacheThe Role of the Cache in System PerformanceReplicated, Sharded CachesHands On: Deploying an Ambassador and Memcache for a Sharded CacheAn Examination of Sharding FunctionsSelecting a KeyConsistent Hashing FunctionsHands On: Building a Consistent HTTP Sharding ProxySharded, Replicated ServingHot Sharding Systems
Scatter/Gather with Root DistributionHands On: Distributed Document SearchScatter/Gather with Leaf ShardingHands On: Sharded Document SearchChoosing the Right Number of LeavesScaling Scatter/Gather for Reliability and Scale

Determining When FaaS Makes SenseThe Benefits of FaaSThe Challenges of FaaSThe Need for Background ProcessingThe Need to Hold Data in MemoryThe Costs of Sustained Request-Based ProcessingPatterns for FaaSThe Decorator Pattern: Request or Response TransformationHands On: Adding Request Defaulting Prior to Request ProcessingHandling EventsHands On: Implementing Two-Factor AuthenticationEvent-Based PipelinesHands On: Implementing a Pipeline for New-User Signup
Determining If You Even Need Master ElectionThe Basics of Master ElectionHands On: Deploying etcdImplementing LocksHands On: Implementing Locks in etcdImplementing OwnershipHands On: Implementing Leases in etcdHandling Concurrent Data Manipulation
A Generic Work Queue SystemThe Source Container InterfaceThe Worker Container InterfaceThe Shared Work Queue InfrastructureHands On: Implementing a Video ThumbnailerDynamic Scaling of the WorkersThe Multi-Worker Pattern
Patterns of Event-Driven ProcessingCopierFilterSplitterSharderMergerHands On: Building an Event-Driven Flow for New User Sign-UpPublisher/Subscriber InfrastructureHands On: Deploying Kafka
Join (or Barrier Synchronization)ReduceHands On: CountSumHistogramHands On: An Image Tagging and Processing Pipeline

Content preview from Designing Distributed Systems

Chapter 9. Ownership Election

The previous patterns that we have seen have been about distributing requests in order to scale requests per second, the state being served, or the time to process a request. This final chapter on multi-node serving patterns is about how you scale assignment. In many different systems, there is a notion of ownership where a specific process owns a specific task. We have previously seen this in the context of sharded and hot-sharded systems where specific instances owned specific sections of the sharded key space.

In the context of a single server, ownership is generally straightforward to achieve because there is only a single application that is establishing ownership, and it can use well-established in-process locks to ensure that only a single actor owns a particular shard or context. However, restricting ownership to a single application limits scalability, since the task can’t be replicated, and reliability, since if the task fails, it is unavailable for a period of time. Consequently, when ownership is required in your system, you need to develop a distributed system for establishing ownership.

A general diagram of distributed ownership is shown in Figure 9-1. In the diagram, there are three replicas that could be the owner or master. Initially, the first replica is the master. Then that replica fails, and replica number three then becomes the master. Finally, replica number one recovers and returns to the group, but replica three remains as ...