BALAJI P’s Post

🏛️ Why Amazon Prime Video Opted for a Monolithic Architecture Did you know that Amazon Prime Video saved 90% of its costs by switching back to a monolithic architecture? 🚀 In a time when microservices are the gold standard, this move is surprising. But why can a monolith sometimes be the better choice? ↓ 📌 The Problem: High Costs and Performance Bottlenecks Amazon Prime Video operates a system that constantly monitors livestreams for quality issues like freezing images or audio glitches. This system was originally built on a microservices architecture using AWS Lambda. However, as the user base grew, two major challenges emerged: → Cost Explosion: More users, more streams – and rapidly rising expenses. 💰 → Performance Bottlenecks: Communication between the microservices led to delays. 📉 🛠️ The Solution: Monolithic Architecture To reduce costs and improve performance, Amazon Prime Video transitioned to a monolith. This switch simplified the architecture and made resource usage more efficient. The result? Lower latency, reduced costs, and an overall performance boost. 🚀 🏆 Conclusion: Flexibility Over Trends Microservices are not always the best solution. Especially for tightly connected systems, a monolith can be more cost-effective and efficient. Amazon’s decision wasn’t a step backward – it was a strategic move for greater efficiency. 💡 What Can You Learn from This? → Flexibility: Adapt your architecture to actual needs. → Analysis: Regularly assess if your infrastructure is still the best fit. → Economics: Ensure that your costs are aligned with the benefits. 🤔 Your Opinion Matters! Have you experienced switching between microservices and monoliths? Where do you see the biggest advantages or challenges? Share your thoughts in the comments! 👇 If you enjoyed this article, feel free to share it with your network! 😊 #Monolith #Microservices #SoftwareArchitecture #CostReduction #TechInsights #AWS #EfficiencyBoost

In 2023, Amazon Prime Video made a surprising move by replacing its microservices architecture with a monolith. This decision resulted in a staggering 90% cost reduction across the board. Typically, companies break up a monolith to scale, but Amazon took the opposite approach. Here's how they pulled it off: ♦ The Challenge Amazon Prime Video’s monitoring service for live streams, used to detect issues like freezing, clicks, and glitches, was built on a microservices architecture. However, this system couldn’t handle 90% of the anticipated load due to high costs and performance bottlenecks. The orchestration between components (via AWS Step Functions) was especially slow and expensive. ♦ The Discovery As more streams were onboarded, the team noticed the infrastructure struggled to keep up. AWS Step Functions' state transitions added significant costs and delays, while passing video frames and audio buffers between services further compounded the issue, creating network, storage, and computational overhead. ♦ The Solution Amazon Prime Video transitioned from a distributed microservices setup to a monolithic architecture. By consolidating all components (media conversion, detection, orchestration) into a single process, they kept everything in memory within the same instance, drastically cutting down on data transfer costs and eliminating the need for S3 operations. To scale, they cloned the service with multiple configurations, running several detectors in the same monolith instance rather than breaking them into microservices. This shift allowed them to cut costs by 90% and significantly improve scalability—handling thousands of streams without hitting performance limits. ♦ Key Takeaways Microservices and serverless architectures can scale well, but they’re not always the most cost-effective option. A monolith can simplify tightly coupled workflows, reducing both complexity and expenses. Scaling decisions should be based on specific needs, not just trends like microservices or serverless. #FullStackDeveloper #Microservices #Monolith #AmazonPrimeVideo #CostOptimization #SoftwareArchitecture #AWS #StepFunctions #Scalability #TechInnovation #CloudComputing #PerformanceOptimization #SoftwareDevelopment #DevOps #TechStrategy #EngineeringLeadership #FullStackDev #Coding

🚀 Tech Highlight for today's post: Amazon Simple Queue Service (SQS)! 🚀 Let's talk about Amazon Simple Queue Service (SQS) and why it's a must-have for any modern application: SQS ensures your messages are delivered and processed, even if parts of your system are down. The service handles any volume of messages, from a few per day to millions per second, and helps you build robust, scalable systems by decoupling application components. This means your services can work independently and efficiently while encrypting your data in transit and at rest, keeping your information secure. Key Features and Benefits: ✅ Standard and FIFO Queues: Offers both Standard (high throughput) and FIFO (First-In-First-Out) queues to meet different application needs. ✅ Dead-Letter Queues: Automatically handles messages that can't be processed successfully, allowing you to analyze and troubleshoot issues. ✅ Message Attributes: Attach custom metadata to your messages, providing more context and flexibility. ✅ Visibility Timeout: Temporarily hide messages from other consumers while being processed, ensuring they aren't processed multiple times. ✅ Long Polling: Reduce costs and increase efficiency by retrieving messages as soon as they become available. ✅ Message Retention: Store messages for up to 14 days, ensuring they are available until processed successfully. ✅ Batch Operations: Send, receive, and delete messages in batches, reducing network latency and improving throughput. Amazon SQS is perfect for distributed systems, microservices, and serverless applications. If you want to enhance your system's reliability and scalability, SQS is the way to go! #AmazonSQS #CloudComputing #MessagingService #Scalability #TechHighlight

Microservices, Monolith, and Massive Savings. The Three 'M’s that transformed Amazon Prime Video Discover how Amazon Prime Video reduced its infrastructure costs by 90% by migrating from microservices to a monolith. 🚀 Let’s get into the details 🔥 ⏩Initial Amazon Prime Video Architecture The initial version of the Amazon Prime Video architecture consisted of distributed components orchestrated by AWS Step Functions. Theoretically, it would have allowed the team to scale each component independently. But instead: ➜ A few components hit a hard scaling limit at 5% of the expected load. ➜ Video splitting and data transfer to the S3 bucket increased the costs of video rendering. ➜ Orchestration management was expensive since the service performed multiple state transitions every second. ⏩ Moving from Microservices to Monolith: To address the problem, the engineering team decided to re-architect the entire infrastructure. ➜ The team packed all the components into a single process. ➜ Eliminated the need for an S3 bucket as the intermediate storage. ➜ Implemented orchestration that controlled components in a single instance. ⏩ Performance Engineering for Better Customer Experience: The new architecture only allowed the detectors to scale vertically because they all ran within the same instance. The team had to add more detectors to the service, so they exceeded the capability of a single instance. To overcome this: ➜ The team cloned the service multiple times and parameterized each copy with a different subset of detectors. ➜ They implemented a lightweight orchestration layer to distribute customer requests. 🔄 The result: ➜ Moving to a monolith architecture reduced infrastructural costs by up to 90%. ➜ It improved scaling capabilities and allowed them to stream 1000s of videos. ➜ The changes allowed them to observe and monitor all streams viewed by their customers. The transformative journey highlights the need to rethink choices before blindly following trends. The focus must be on building evolving systems strategically instead of doing it religiously. Now a quick question - Is the monolith the solution to your engineering challenges? Let us know in the comments. 👇 #aws #microservices #softwarearchitecture #softwareengineering | Prime Video & Amazon Studios

Today in one of interesting conversation some one asked will microservice going to stay in future given that Amazon prime itself ditched it in favor of monolithic way. I feel that not all can be rolled back to monolithic way , ofcourse microservice comes with its benefits of easy auto scaling at expense of orchestration..ur views

If you haven't started deploying applications to Amazon EKS Auto Mode and are looking for before and after changes, a sample application, and more information about key considerations, then this post is for you: https://lnkd.in/egNyvybx. Check out this blog co-authored by Sheetal Joshi Alex Kestner Ashley Ansari Robert Northard as we dive deep into all these aspects.

Optimizing Global Content Delivery with Amazon CloudFront I’ve finished up a project aimed at delivering faster and more reliable content globally! 🌍 💡 Utilized Amazon CloudFront with multiple origins and failover configurations to ensure high availability and improved performance. 🔄 Built a distribution that intelligently routes traffic to the optimal origin for reduced latency and seamless user experience. This project pushed my skills in designing for both performance and resiliency. Dive into the details here: [https://lnkd.in/gvrEeyFP] #AWS #CloudFront #ContentDelivery #ResilientArchitecture #PerformanceOptimization #SolutionArchitect

If you're using SQS FIFO take a look at these new metrics. They give you a lot more visibility into the behavior of your publishers.

During my initial days at Amazon, in a KT session I asked what's Amazon SQS! Yup, I didn't know what it was and why it was used! Somebody replied it's a normal queue that stores messages for you! Later on I googled and understood messaging queues and the need for it! In today’s distributed systems, message queues play a critical role in ensuring smooth communication between different services or components. But how do they actually work? 🔄 Asynchronous Communication: Message queues decouple the sender and receiver, enabling asynchronous messaging. This allows the sender to push a message to the queue and move on to other tasks without waiting for a response, ensuring better efficiency. 📥 Buffering: If a service is too busy, a message queue acts as a buffer. It stores messages until the service is ready to process them, preventing overloading and allowing for better system resilience. ⚙️ Scalability: As load increases, message queues can distribute messages across multiple consumers, improving the overall scalability of the system. It helps systems handle peaks in traffic without degradation in performance. 🔐 Reliability: Message queues provide durability by ensuring that messages are delivered even if the system or network temporarily fails. They can persist messages to disk and guarantee delivery once the receiver is ready. 📈 Use Cases: Whether it's handling requests in a microservices architecture or orchestrating tasks in a cloud environment, message queues are essential for building robust, scalable, and fault-tolerant systems. I am still amazed on how a simple queue can ensure all of the above things 😮!! #DistributedSystems #AmazonSQS #MessageQueues #CloudComputing #Microservices

Amazon Aurora Serverless v2 supports scaling to zero capacity now. If you have used this in the past you maybe thought it might of, but it didnt, even though it was serverless so likely caused some cost for some people or some frustrations, but its now here! I know it will save me a bit of money in some accounts, so thanks :-) Theres also been a nice addition to #FCS for #amazonConnect now supporting new languages which I know could help adoption in some of our European customers. New languages now supported include: Canadian French, Chinese (Simplified and Traditional), French, German, Italian, Japanese, Korean, Portuguese (Brazilian), and Spanish. Connect also enhanced its reporting Dashboards further today alongside other enhancements this week, with ability to now create custom dashboards Now, you can further customise these dashboards by changing widgets to create the view that best fits your specific business need. For example, if you want to monitor self-service, queue, and agent performance, you can add all three types of widgets to your dashboard to have a single end to end view of contact center performance.