Shift in Landscape

When considering reliability in a distributed system, considerable shifts in our environmental landscape have caused us to examine how we operate and run our systems and networks. We have had a mega change with the introduction of various cloud platforms and their services and containers, along with the complexity of managing distributed systems observability and microservices observability that unveil significant gaps in current practices in our technologies. Not to mention the flaws with the operational practices around these technologies.

Existing Static Tools

This has caused a knee-jerk reaction to a welcomed drive-in innovation to system reliability. Yet, some technologies and tools used to manage these innovations do not align with the innovative events. Many of these tools have stayed relatively static in our dynamic environment. So, we have static tools used in a dynamic environment, which causes friction to reliability in distributed systems and the rise for more efficient network visibility.

Understanding the Complexity

Distributed systems are inherently complex, with multiple components across different machines or networks. This complexity introduces challenges like network latency, hardware failures, and communication bottlenecks. Understanding the intricate nature of distributed systems is crucial to devising reliable solutions.

Redundancy and Replication

One critical approach to enhancing reliability in distributed systems is redundancy and replication. The system becomes more fault-tolerant by duplicating critical components or data across multiple nodes. This ensures the system can function seamlessly even if one component fails, minimizing the risk of complete failure.

Consistency and Consensus Algorithms

Maintaining consistency in distributed systems is a significant challenge due to the possibility of concurrent updates and network delays. Consensus algorithms, such as the Paxos or Raft algorithms, are vital in achieving consistency by ensuring agreement among distributed nodes. These algorithms enable reliable decision-making and guarantee that all nodes reach a consensus state.

Monitoring and Failure Detection

To ensure reliability, it is essential to have robust monitoring mechanisms in place. Monitoring tools can track system performance, resource utilization, and network health. Additionally, implementing efficient failure detection mechanisms allows for prompt identification of faulty components, enabling proactive measures to mitigate their impact on the overall system.

Load Balancing and Scalability

Load balancing is crucial in distributing the workload evenly across nodes in a distributed system. It ensures that no single node is overwhelmed, reducing the risk of system instability. Furthermore, designing systems with scalability in mind allows for seamless expansion as the workload grows, ensuring that reliability is maintained even during periods of high demand.

Related: Before you proceed, you may find the following post helpful:

1. Distributed Firewalls
2. SD WAN Static Network Based

Back to Basics: Reliability in Distributed Systems

Understanding Distributed Systems

Distributed systems refer to a network of interconnected computers that communicate and coordinate their actions to achieve a common goal. Unlike traditional centralized systems, where a single entity controls all components, distributed systems distribute tasks and data across multiple nodes. This decentralized approach enables enhanced scalability, fault tolerance, and resource utilization.

Key Components of Distributed Systems

To comprehend the inner workings of distributed systems, we must familiarize ourselves with their key components. These components include nodes, communication channels, protocols, and distributed file systems. Nodes represent individual machines or devices within the network; communication channels facilitate data transmission, protocols ensure reliable communication, and distributed file systems enable data storage across multiple nodes.

Distributed Systems Use Cases

Distributed systems are used in many modern applications. Mobile and web applications with high traffic are distributed systems. Web browsers or mobile applications serve as clients in a client-server environment. As a result, the server becomes its own distributed system. The modern web server follows a multi-tier system pattern. Requests are delegated to several server logic nodes via a load balancer.

Kubernetes is popular among distributed systems since it enables containers to be combined into a distributed system. Kubernetes orchestrates network communication between the distributed system nodes and handles dynamic horizontal and vertical scaling of the nodes. 

Cryptocurrencies like Bitcoin and Ethereum are also distributed systems that are peer-to-peer. The currency ledger is replicated at every node in a cryptocurrency network. To bootstrap, a currency node connects to other nodes and downloads its full ledger copy. Additionally, cryptocurrency wallets use JSON RPC to communicate with the ledger nodes.

Challenges in Distributed Systems

While distributed systems offer numerous advantages, they also pose various challenges. One significant challenge is achieving consensus among distributed nodes. Ensuring that all nodes agree on a particular value or decision can be complex, especially in the presence of failures or network partitions. Additionally, maintaining data consistency across distributed nodes and mitigating issues related to concurrency control requires careful design and implementation.

Example: Distributed System of Microservices

Microservices are one type of distributed system since they decompose an application into individual components. A microservice architecture, for example, may have services corresponding to business features (payments, users, products, etc.), with each component handling the corresponding business logic. Multiple redundant copies of the services will then be available, so there is no single point of failure.

Example: Distributed Tracing

Using distributed tracing, you can profile or monitor the results of requests across a distributed system. Distributed systems can be challenging to monitor since each node generates its logs and metrics. To get a complete view of a distributed system, it is necessary to aggregate these separate node metrics holistically. 

A distributed system generally doesn’t access its entire set of nodes but rather a path through those nodes. With distributed tracing, teams can analyze and monitor commonly accessed paths through a distributed system. The distributed tracing is installed on each system node, allowing teams to query the system for information on node health and performance.

Benefits and Applications

Despite the challenges, distributed systems offer a wide array of benefits. One notable advantage is enhanced fault tolerance. Distributing tasks and data across multiple nodes improves system reliability, as a single point of failure does not bring down the entire system. Additionally, distributed systems enable improved scalability, accommodating growing demands by adding more nodes to the network. The applications of distributed systems are vast, ranging from cloud computing and large-scale data processing to peer-to-peer networks and distributed databases.

Distributed Systems: The Challenge

Distributed systems are required to implement the reliability, agility, and scale expected of modern computer programs. Distributed systems are applications of many different components running on many other machines. Containers are the foundational building block, and groups of containers co-located on a single device comprise the atomic elements of distributed system patterns.

The significant shift we see with software platforms is that they evolve much quicker than the products and paradigms we use to monitor them. We need to consider new practices and technologies with dedicated platform teams to enable a new era of system reliability in a distributed system. Along with the practices of Observability that are a step up to the traditional monitoring of static infrastructure: Observability vs monitoring.

Knowledge Check: Distributed Systems Architecture

Client-Server Architecture

A client-server architecture has two primary responsibilities. The client presents user interfaces and is then connected to the server via a network. The server handles business logic and state management. Unless the server is redundant, a client-server architecture can quickly degrade into a centralized architecture. A truly distributed client-server setup will consist of multiple server nodes that distribute client connections. In modern client-server architectures, clients connect to encapsulated distributed systems on the server.

Multi-tier Architecture

Multi-tier architectures are extensions of client-server architectures. Multi-tier architectures decompose servers into further granular nodes, which decouple additional backend server responsibilities like data processing and data management. By processing long-running jobs asynchronously, these additional nodes free up the remaining backend nodes to focus on responding to client requests and interacting with the data store.

Peer-to-Peer Architecture

Peer-to-peer distributed systems contain complete instances of applications on each node. There is no separation between presentation and data processing at the node level. A node consists of a presentation layer and a data handling layer. Peer nodes may contain the entire state data of the system. 

Peer-to-peer systems have a great deal of redundancy. Peer-to-peer nodes discover and connect to other peers when they are initiated and brought online, thereby synchronizing their local state with the system’s. As a result of this feature, nodes on a peer-to-peer network won’t be disrupted by the failure of one. Additionally, peer-to-peer systems will persist. 

Service-orientated Architecture

A service-oriented architecture (SOA) is a precursor to microservices. Microservices differ from SOA primarily in their node scope, which is at the feature level. Each microservice node encapsulates a specific set of business logic, such as payment processing. Multiple nodes of business logic interface with independent databases in a microservice architecture. In contrast, SOA nodes encapsulate an entire application or enterprise division. Database systems are typically included within the service boundary of SOA nodes.

Because of their benefits, microservices have become more popular than SOA. The small service nodes provide functionality that teams can reuse through microservices. The advantages of microservices include greater robustness and a more extraordinary ability for vertical and horizontal scaling to be dynamic.

Reliability in Distributed Systems: Components

Redundancy and Replication:

Redundancy and replication are two fundamental concepts distributed systems use to enhance reliability. Redundancy involves duplicating critical system components, such as servers, storage devices, or network links, so the redundant component can seamlessly take over if one fails. Replication, on the other hand, involves creating multiple copies of data across different nodes in a system, enabling efficient data access and fault tolerance. By incorporating redundancy and replication, distributed systems can continue to operate even when individual components fail.

Fault Tolerance:

Fault tolerance is a crucial aspect of achieving reliability in distributed systems. It involves designing systems to operate correctly even when one or more components encounter failures. Several techniques, such as error detection, recovery, and prevention mechanisms, are employed to achieve fault tolerance.

Error Detection:

Error detection techniques, such as checksums, hashing, and cyclic redundancy checks (CRC), identify errors or data corruption during transmission or storage. By verifying data integrity, these techniques help identify and mitigate potential failures in distributed systems.

Error Recovery:

Error recovery mechanisms, such as checkpointing and rollback recovery, aim to restore the system to a consistent state after a failure. Checkpointing involves periodically saving the system’s state and data, allowing recovery to a previously known good state in case of failures. On the other hand, rollback recovery involves undoing the effects of failed operations and returning the system to a consistent state.

Error Prevention:

To enhance reliability, distributed systems employ error prevention techniques, such as redundancy elimination, consensus algorithms, and load balancing. Redundancy elimination reduces unnecessary duplication of data or computation, thereby reducing the chances of errors. Consensus algorithms ensure that all nodes in a distributed system agree on a shared state despite failures or message delays. Load balancing techniques distribute computational tasks evenly across multiple nodes to prevent overloading and potential shortcomings.

Lack of Connective Event: Traditional Monitoring

If you examine traditional monitoring systems, they look to capture and investigate signals in isolation. The monitoring systems work in a siloed environment, similar to that of developers and operators before the rise of DevOps. Existing monitoring systems cannot detect the “Unknowns Unknowns” that are familiar with modern distributed systems. This often leads to disruptions of services. So you may be asking what an “Unknown Unknown” is.

I’ll put it to you this way: the distributed systems we see today lack predictability—certainly not enough predictability to rely on static thresholds, alerts, and old monitoring tools. If something is fixed, it can be automated, and we have static events, such as in Kubernetes, a POD reaching a limit.

Then, a replica set introduces another pod on a different node if specific parameters are met, such as Kubernetes Labels and Node Selectors. However, this is only a tiny piece of the failure puzzle in a distributed environment.  Today, we have what’s known as partial failures and systems that fail in very creative ways.

Reliability In Distributed System: Creative ways to fail

So, we know that some of these failures are quickly predicted, and actions are taken. For example, if this Kubernetes POD node reaches a specific utilization, we can automatically reschedule PODs on a different node to stay within our known scale limits.

Predictable failures can be automated in Kubernetes and with any infrastructure. An Ansible script is useful when these events occur. However, we have much more to deal with than POD scaling; we have many partial and complicated failures known as black holes.

In today’s world of partial failures

Microservices applications are distributed and susceptible to many external factors. On the other hand, if you examine the traditional monolithic application style, all the functions reside in the same process. It was either switched ON or OFF!! Not much happened in between. So, if there is a failure in the procedure, the application as a whole will fail. The results are binary, usually either a UP or Down.

With some essential monitoring, this was easy to detect, and failures were predictable. There was no such thing as a partial failure. In a monolith application, all application functions are within the same process. A significant benefit of these monoliths is that you don’t have partial failures.

However, in a cloud-native world, where we have broken the old monolith into a microservices-based application, a client request can go through multiple hops of microservices, and we can have several problems to deal with.

There is a lack of connectivity between the different domains. Many monitoring tools and knowledge will be tied to each domain, and alerts are often tied to thresholds or rate-of-change violations that have nothing to do with user satisfaction. User satisfaction is a critical metric to care about.

System reliability: Today, you have no way to predict

So, the new, modern, and complex distributed systems place very different demands on your infrastructure—considerably different from the simple three-tier application, where everything is generally housed in one location.  We can’t predict anything anymore, which breaks traditional monitoring approaches.

When you can no longer predict what will happen, you can no longer rely on a reactive approach to monitoring and management. The move towards a proactive approach to system reliability is a welcomed strategy.

Blackholes: Strange failure modes

When considering a distributed system, many things can happen. A service or region can disappear or disappear for a few seconds or ms and reappear. We believe this is going into a black hole when we have strange failure modes. So when anything goes into it will disappear. Peculiar failure modes are unexpected and surprising.

There is certainly nothing predictable about strange failure modes. So, what happens when your banking transactions are in a black hole? What if your banking balance is displayed incorrectly or if you make a transfer to an external account and it does not show up? 

Highlighting Site Reliability Engineering (SRE) and Observability

Site reliability engineering (SRE) and observational practices are needed to manage these types of unpredictability and unknown failures. SRE is about making systems more reliable. And everyone has a different way of implementing SRE practices. Usually, about 20% of your issues cause 80% of your problems.

You need to be proactive and fix these issues upfront. You need to be able to get ahead of the curve and do these things to stop the incidents from occurring. This usually happens in the wake of a massive incident. This usually acts as a teachable moment. This gives the power to be the reason to listen to a Chaos Engineering project. 

New tools and technologies: Distributed tracing

We have new tools, such as distributed tracing. So, what is the best way to find the bottleneck if the system becomes slow? Here, you can use Distributed Tracing and Open Telemetry. The tracing helps us instrument our system, figuring out where the time has been spent and where it can be used across distributed microservice architecture to troubleshoot problems. Open Telemetry provides a standardized way of instrumenting our system and providing those traces.

SLA, SLI, SLO, and Error Budgets

So we don’t just want to know when something has happened and then react to an event that is not looking from the customer’s perspective. We need to understand if we are meeting SLA by gathering the number and frequency of the outages and any performance issues.

Service Level Objectives (SLO) and Service Level Indicators (SLI) can assist you with measurements. Service Level Objectives (SLOs) and Service Level Indicators (SLI) not only help you with measurements but also offer a tool for having better reliability and forming the base for the reliability stack.