Distributed Systems and Technologies

A Distributed system is a group of independent computers (called nodes) working together to achieve a collective goal. This goal could be a task or providing a service. They work together as a unit in such a way that they must appear as a single system to the user. They communicate and work together by distributing state, information and resources amongst themselves in a highly coordinated manner. These nodes can be located together or spread across different geographical locations.

Examples of distributed systems include distributed computing, distributed storage, distributed databases etc.

The biggest example of a distributed system is the internet. The internet is powered by thousands of computers all over the world, receiving and processing millions of requests every second.

Why would we want to build distributed Systems?

Distributed systems exist for many reasons, some of which include:

Scaling: Scalability is the term used to describe a system’s ability to cope with increased load.

This is one of the most important reasons for the existence of distributed systems, some workloads are just too big for a single computer. There are two ways systems can deal with increasing workload: vertically (scaling up)and horizontally(scaling out).

Vertical scaling involves upgrading a computer’s hardware with more powerful processors, increased CPU, and RAM. However, this approach has its limitations, and there is a cap on how much a single computer can scale. Eventually, the system encounters a hard ceiling beyond which further vertical scaling becomes impractical or impossible. This is also expensive as this implies continuous investment in hardware.

Horizontally scaling on the other hand means continuously adding more computers/nodes as workloads increase.

Here’s where distributed systems come into play. By adopting a horizontal scaling strategy, we can expand the system’s capacity by adding more machines or nodes to the network. Each node contributes its processing power, memory, and resources to the overall workload, collectively enabling the system to scale beyond what a single machine could achieve.

Fault tolerance/High availability: All or part of systems fail eventually. The ability of a system to carry on operating in the event of a failure is called fault tolerance. Let us imagine we have our entire infrastructure on a single system in a system and then the system fails. This means that the entire service goes down. This one system becomes a single point of failure. Distributed systems by nature are designed to be fault tolerant, when systems go down as they inherently will, the failure of a single system or part of the system does not imply total outage of the service. Distributed systems use concepts like redundancy, replication to ensure fault tolerance. Redundancy involves having backup nodes that can take over the tasks of failed nodes, while replication ensures that data is stored on multiple nodes, ensuring its availability even if one node fails. They ensure that if one node fails, another can seamlessly take over the task without causing service disruption. This is important for services that need to be available continuously, such as online streaming platforms like Netflix or Public Cloud platforms. Users expect uninterrupted access to such services, and the fault tolerance of distributed systems ensures that disruptions are minimised and service downtime is avoided.

Performance and Latency:Latency refers to the time delay between initiating a request and receiving a response. It represents the time taken for data to travel from the sender to the receiver and encompasses various factors such as network communication, processing time, and queuing time. By distributing tasks across different nodes in different geographical locations, distributed systems achieve higher performance, efficient processing, and reduced latency.

Cloud Computing and Distributed Systems

Cloud computing is the delivery of IT resources over the internet on demand. These resources include data storage, servers, databases, networking, and software. The public cloud (AWS, Google Cloud , Azure, etc) are distributed. They provide a vast array of services and resources that operate across multiple geographically distributed data centres.

By leveraging the distributed nature of the public cloud, people can deploy distributed technologies without having to buy the infrastructure.

In traditional on-premises setups, deploying and managing distributed systems can be complex, time-consuming, and costly due to the need to purchase, configure, and maintain hardware, networking equipment, and data centres. You can rent the infrastructure or use the providers managed services for the technologies.

Distributed Technologies

Distributed technologies refer to a set of tools, frameworks, and software components that enable the development, deployment, and management of distributed systems. These technologies facilitate the coordination and communication between multiple nodes or components within a network, allowing them to work together to achieve a common goal. Distributed technologies play a crucial role in building scalable, fault-tolerant, and high-performance distributed systems. Some common examples of distributed technologies include:

• Message Queue Systems: Message queue systems facilitate asynchronous communication between distributed components, decoupling the sender and receiver of messages. Examples include Apache Kafka, RabbitMQ, and Amazon Simple Queue Service (SQS).
• Distributed Databases: Distributed databases store data across multiple nodes and locations, providing high availability, scalability, and fault tolerance. Examples include Apache Cassandra, Amazon DynamoDB, and Google Cloud Spanner.
• Distributed File Systems: Distributed file systems enable data storage across multiple servers and allow clients to access data as if it were on a single file system. Examples include Hadoop Distributed File System (HDFS) and Google File System (GFS).
• MapReduce Frameworks: MapReduce frameworks allow distributed processing of large datasets across multiple nodes. Examples include Apache Hadoop MapReduce and Google Cloud Dataflow.
• Microservices Architectures: Microservices architectural style decomposes applications into small, independent services that communicate through APIs, enabling easy scaling and independent deployment of each service.
• Containerization Technologies: Containerization enables the packaging of applications and their dependencies into lightweight, portable containers, making it easier to deploy and manage distributed applications. Examples include Docker and Kubernetes.
• Service Meshes: Service meshes provide a dedicated infrastructure layer for managing service-to-service communication within a distributed application. Examples include Istio and Linkerd.
• Distributed Computing Frameworks: Distributed computing frameworks facilitate the execution of distributed algorithms and parallel processing across multiple nodes. Examples include Apache Spark and Apache Flink.
• Consensus Algorithms: Consensus algorithms help achieve agreement among distributed nodes in the presence of failures. Examples include Paxos and Raft.
• Load Balancers: Load balancers distribute incoming network traffic across multiple servers, ensuring even resource utilization and preventing individual nodes from being overloaded
• Content Delivery Networks (CDNs): CDNs cache and deliver content from multiple distributed servers located in different geographical regions, reducing latency and improving the performance of web applications.

However, like all technologies, distributed systems are not without downsides, some of which include high level of complexity, difficulty in debugging and troubleshoting, network dependency and so on. They require careful planning, effective design, and proper use of distributed technologies and best practices.