# Invisible Architectures | Anokh Kishore | TEDxPSNACET Source: https://www.youtube.com/watch?v=WJh0mtgytPw Recap page: https://rapidrecap.app/video/WJh0mtgytPw Generated: 2026-02-02T17:36:19.908+00:00 --- ## Quick Overview Anokh Kishore, a Senior Software Engineer at Akuna Capital, argues that while microservices offer benefits like independent scaling and deployment, monolithic architectures can sometimes be superior for specific needs, particularly when rapid development velocity, high test coverage (60-70%), and simple observability are prioritized, suggesting that the complexity of distributed systems like microservices often introduces unnecessary overhead like increased latency and network traffic unless managed meticulously. **Key Points:** - Microservices offer flexibility for independent scaling and deployment, but monolithic architectures can be better suited for certain problems, especially in early development stages. - The speaker, Anokh Kishore, advocates for high test coverage (60-70%) in unit tests when designing systems, regardless of architecture. - Microservices inherently increase network traffic and latency due to inter-service communication (IPC or shared memory), which monolithic systems avoid when components run in the same process. - State management in microservices requires explicit observability, metrics recording, and alerting, which is often less complex in a monolith where state is shared within one process. - Companies sometimes revert from microservices to monoliths when the operational overhead (monitoring, logging, complexity) outweighs the benefits, especially in high-frequency trading environments where latency is critical. - The best approach often involves a hybrid model where monoliths are carefully designed to allow for future decoupling if necessary, maintaining simple state management. - Unit testing becomes crucial, forcing developers to design services that are testable without relying on complex external dependencies during testing. ![Screenshot at 00:01: The opening title card displays the event branding: TEDxPSNACET, with the subtitle 'x = independently organized TED event'.](https://ss.rapidrecap.app/screens/WJh0mtgytPw/00-00-01.jpg) **Context:** The video features Anokh Kishore, identified as a Senior Software Engineer at Akuna Capital, presenting insights likely from a TEDx event (TEDxPSNACET). Kishore discusses the trade-offs between monolithic and microservices architectures in software engineering, drawing from his decade of experience across various company types, including high-frequency trading firms and Big Tech. ## Detailed Analysis Anokh Kishore, Senior Software Engineer at Akuna Capital, discusses the architectural choice between monoliths and microservices, suggesting that the prevailing trend toward microservices is not always the optimal path. He notes that monoliths, while encompassing a large functionality block in a single process, avoid the network communication overhead inherent in microservices, which can lead to higher latency and increased network traffic when services communicate across the network or via IPC. Kishore emphasizes that for certain applications, especially those demanding ultra-low latency like high-frequency trading, monoliths can offer better performance. A key takeaway is the importance of rigorous testing; he suggests aiming for 60-70% test coverage in unit tests, which forces good design practices regardless of the overall architecture. Furthermore, he points out that state management in microservices requires complex observability, logging, and alerting infrastructure to track failures across distributed components, whereas monoliths simplify this by sharing memory. He mentions observing large companies reverting from microservices back to monoliths because the operational complexity became too high. Kishore concludes that while microservices offer flexibility, a well-designed monolith that accounts for future decoupling (or a hybrid approach) is often a more pragmatic choice, especially early in development or when performance requirements are strict. ### Introduction and Experience - Working in software industry for a decade across HFT firms and Big Tech - Developed various types of software services and web applications. ### Monolith vs. Microservices Trade-offs - Monoliths run functionality in a single process, avoiding network latency/traffic; Microservices require high network communication and resource overhead. ### Testing and Observability - Advocate for 60-70% unit test coverage; Microservices necessitate robust monitoring/metrics for state management across distributed instances. ### Real-World Reversions - Some large companies reversed microservice adoption due to complexity, high latency, and operational burden. ### Conclusion on Architecture - A hybrid approach or a well-designed monolith that allows for future decoupling is often better than blindly adopting microservices. ![Screenshot at 00:03: Anokh Kishore introduces himself as a Senior Software Engineer at Akuna Capital.](https://ss.rapidrecap.app/screens/WJh0mtgytPw/00-00-03.jpg) ![Screenshot at 01:04: Kishore visually contrasts the two architectural styles, discussing Monolith versus Microservices.](https://ss.rapidrecap.app/screens/WJh0mtgytPw/00-01-04.jpg) ![Screenshot at 01:19: Kishore uses hand gestures to illustrate how monolithic components share a single process, contrasting with distributed services.](https://ss.rapidrecap.app/screens/WJh0mtgytPw/00-01-19.jpg) ![Screenshot at 04:44: Kishore emphasizes the importance of designing services to be testable, even if they depend on external factors like databases.](https://ss.rapidrecap.app/screens/WJh0mtgytPw/00-04-44.jpg) ![Screenshot at 07:01: Kishore discusses the benefit of state management in monoliths where state is shared internally, unlike microservices communicating externally.](https://ss.rapidrecap.app/screens/WJh0mtgytPw/00-07-01.jpg)