Nazmus Sakib

If a 100 Mb video gets 20k views in one month, how much bandwidth do we need to serve the video? These kinds of questions are called capacity estimation questions. They are part of system design interviews. The next version of the system design judge will have capacity estimation questions. Coming soon! #SystemDesignJudge #CapacityEstimation

319

27 Comments

𝐃𝐄𝐀𝐃𝐋𝐎𝐂𝐊 - 𝙒𝙝𝙚𝙣 𝙏𝙬𝙤 𝘾𝙤𝙤𝙠𝙨 𝘼𝙧𝙚 𝙎𝙩𝙪𝙘𝙠 𝙒𝙖𝙞𝙩𝙞𝙣𝙜 𝙛𝙤𝙧 𝙀𝙖𝙘𝙝 𝙊𝙩𝙝𝙚𝙧 🍳🍲 🧠 𝐉𝐚𝐯𝐚 𝐁𝐮𝐟𝐟𝐞𝐭 : Day 4 of #30DaysOfJava (Extending analogy of Post 3 : https://lnkd.in/gE9mb6dJ) Imagine two cooks in a kitchen. One has a frying pan and needs the spoon to finish their dish, while the other has the spoon but needs the pan. Each cook is waiting for the other to release the tool they need, but neither will budge. As a result, nothing gets done, and both dishes remain uncooked! This is what deadlock in Java looks like. In Java, deadlock occurs when two or more threads are stuck waiting for each other to release resources. Each thread holds a resource and refuses to give it up until it gets another resource from the other thread. Since neither can proceed, the program grinds to a halt—much like our cooks staring at each other in frustration, unable to finish their dishes. 📌 Key Takeaway: Deadlock in Java is like two cooks waiting for each other to give up a tool they need. Both end up stuck, and nothing gets done! Thoughts on the analogy ? If you are looking for referrals and other tech job related stuff, check out Ramit Kundu 's discord community : https://lnkd.in/g9Mupq_3 #java #Oracle #jobs #referral

21

8 Comments

Design systems as if you were the one who will maintain them. Few nuggets to keep handy, while designing a scalable system: https://lnkd.in/g_udCNm3 #systemdesign #scalability #performance

15

Observation: a lot of domains put up with ad hoc / limited programming models just to get ease of deployment and 0 serialization boilerplate. (Think: KSQL vs Kafka + normal code) You can have the best of both. Ease of deployment, etc, but with a rich, general-purpose language.

4

Chunking Revolution: Master Semantic Chunking for Powerful LLMs RSVP here -  https://bit.ly/3Vd7bYE Ever feel overwhelmed by large documents? Chunking can be your secret weapon! It's a powerful technique for breaking down complex information into manageable, meaningful pieces. 𝗧𝗵𝗶𝗻𝗸 𝗼𝗳 𝗶𝘁 𝗹𝗶𝗸𝗲 𝘁𝗵𝗶𝘀: Imagine chopping a giant tree trunk into manageable logs. Chunking does the same thing for text, making it easier to understand, analyze, and use. 𝗕𝘂𝘁 𝗵𝗲𝗿𝗲'𝘀 𝘁𝗵𝗲 𝗰𝗮𝘁𝗰𝗵: Traditional chunking methods can be limited. This is where Semantic Chunking comes in! It leverages the power of Large Language Models (LLMs) to create smarter, more context-aware chunks. 𝗪𝗵𝗲𝗿𝗲 𝗶𝘀 𝗰𝗵𝘂𝗻𝗸𝗶𝗻𝗴 𝘂𝘀𝗲𝗱? 𝗖𝗵𝘂𝗻𝗸𝗶𝗻𝗴 𝗽𝗹𝗮𝘆𝘀 𝗮 𝘃𝗶𝘁𝗮𝗹 𝗿𝗼𝗹𝗲 𝗶𝗻 𝘃𝗮𝗿𝗶𝗼𝘂𝘀 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀, 𝗶𝗻𝗰𝗹𝘂𝗱𝗶𝗻𝗴: Reading Comprehension Systems: Chunking helps computers understand the relationships between different parts of a text. Question Answering Systems: By chunking documents, systems can more easily identify relevant sections to answer specific questions. Text Summarization: Chunking allows for the creation of summaries that capture the key points of a document. RAG (Retrieval-Augmented Generation) Applications: Semantic chunking can significantly enhance the performance of RAG systems, which combine retrieval and generation techniques for tasks like summarization and question answering. Join our FREE live session on Wednesday, June 5th, to learn how Semantic Chunking can revolutionize your document processing! 𝗜𝗻 𝘁𝗵𝗶𝘀 𝘀𝗲𝘀𝘀𝗶𝗼𝗻, 𝘆𝗼𝘂'𝗹𝗹 𝗹𝗲𝗮𝗿𝗻: Chunking 101: Master the fundamentals and best practices for RAG applications. Beyond the Basics: Explore the limitations of traditional chunking methods. Semantic Chunking Power: Learn how to create high-quality, meaningful chunks for enhanced results. Unlocking Chunk Metadata: Discover how to track and cite your chunks effectively. Live Code Demo: See semantic chunking in action with a practical demonstration. Don't miss this opportunity to upgrade your chunking strategies and supercharge your RAG pipelines! Register today for FREE!  https://bit.ly/3Vd7bYE

89

6 Comments

Interesting to read actual evidence of what one could expect as an inevitability of scale!

21

3 Comments

#apache #kafka 👉Apache Kafka is an open-source distributed event streaming platform used for building real-time data pipelines and streaming applications. 👉Here's an overview covering various aspects of Kafka:- 👉Architecture:- Kafka follows a distributed architecture comprising multiple components such as producers, consumers, brokers, topics, partitions, and consumer groups. 👉Topics and Partitions:- Topics are categories to which messages are published by producers. Topics are divided into partitions for scalability and parallelism. Each partition is an ordered, immutable sequence of messages. Partitions allow messages to be distributed across multiple brokers for load balancing. 👉Producers:- Producers are applications that publish messages to Kafka topics. Producers are responsible for choosing which topic and partition to publish a message to. 👉Consumers:- Consumers are applications that subscribe to Kafka topics and consume messages. Consumers can be part of a consumer group, where each consumer in the group reads messages from a subset of partitions. Consumers maintain their offset in each partition, allowing them to control message consumption. 👉Brokers:- Brokers are Kafka servers responsible for storing and managing topic partitions. Brokers handle message storage, replication, and distribution across partitions. 👉Consumer Groups:- Consumer groups are sets of consumers that jointly consume a topic's messages. Each consumer in a group is assigned to one or more partitions of the topic. 👉Scalability and Fault Tolerance:- Kafka is designed for horizontal scalability and fault tolerance. Data replication ensures that partitions are replicated across multiple brokers for fault tolerance. Kafka brokers can be added or removed dynamically without interrupting message delivery. 👉Retention and Compaction:- Kafka allows configuring retention policies for topics to control how long messages are retained in a topic. 👉Streams Processing:- Kafka Streams API enables building real-time stream processing applications on top of Kafka. It provides high-level DSL and low-level Processor API for stream processing. 👉Integration and Ecosystem:- Kafka has a rich ecosystem with connectors for integrating with various systems like databases, message queues, and analytics platforms. 👉Use Cases:- Kafka is widely used for real-time data ingestion, event sourcing, stream processing, log aggregation, and data integration. Feel free to follow Moti Lal Kumar ❤️ for such kind of informative content!! #kafka #learning #softwareengineer #developer #professionals #software#product #development #community

5

With NVIDIA AI Foundry, enterprises and nations can supercharge #generativeAI with a service spanning curation, synthetic data generation, fine-tuning, retrieval, guardrails, and evaluation to deploy custom AI at Meta Llama 3.1 NVIDIA NIM Microservices.

3

After taking 100+ Mock Online System Design Interviews, I saw few generic problems. - People say to use NoSQL DB for scale. Which? How? - Is missing? - People say they will use cache - Which one? Why not Memcached? This is gone... - People replace LBs with API Gateway or BFF 🤦‍♂️ - People use MongoDb most often at scale (no matter whether we need transactional systems or write-heavy systems). - Kafka is a peanut. Use it with any drink. But yes, we don't know how to solve head-of-line blocking 🤦‍♂️ - Auto-scaling is for free (even in persistent systems) 🤦‍♂️ - Redux is required even for a simple app. - Node JS is single-threaded and what not ... What will you add?

5

6 Comments

Engineers often mention only 1 out of 4 techniques to scale database in a system design interview. That is "Shard the data". But sharding is not the best solution in every case. And, then they are unable to justify it.   e.g. small data size (few GBs to 1-2 TB) being queried 100K+ times per seconds by multiple processes. This is a problem of IOPS (Input/Output Operations Per Second), not sharding. Yes, sharding may help, because then you can have more machines to handle the IO, but it also increase the complexity and maintainability. And, then when interviewer puts these constraints, like imagine this data won't grow to PBs, they find it difficult to justify their choices. So, here are 4 ways, you can answer scaling databases in a systems design interview. First, look for what kind of scalability issue it is ("why"): [1] Query Performance - is your database schema comply to user access patterns or not? - are you storing data in a heavily normalised form in a nosql db? - while reading the data, are you doing too many code level joins? If yes, then: - optimise your queries using right keys and indexes - check your user access patterns to avoid joins [2] Data Volume - is the size of individual tables have gone beyond a data node's size - is the rate of data growth for next 5 years breach the data node's size if yes, then: - archive old data, and apply data compression to reduce size - implement table sharding; table partitioning is mostly not useful in SD interviews [3] Concurrent Users - are you handling a huge TPS resulting into multiple concurrent connections (IOPS) - are you seeing a possibility of lock contentions, deadlocks, transaction isolation levels if yes, then: - implement connection pooling, and use read replicas (cheapest and the best) - optimize transaction isolation levels and Implement dead lock prevention strategies [4] Infrastructure Constraints - is your interviewer taking you in the direction of OS constraints like CPU utilization, memory limitations - or you are being challenged for IO constraints like network I/O bottlenecks If yes, then: - go for vertical scaling (bigger machines), horizontal scaling (more machines) - and/or SSD/faster storage, network optimisations (TCP/IP) level Thats all. No matter the solution you propose, discuss trade-offs: - Cost vs Performance - Complexity vs Scalability - Immediate vs Long-term solutions Engineers find DB questions to be the most difficult one. Only because they apply the surface level knowledge. If you understand the "why" behind the questions, Then you can find the "how" behind it. If you want to feel confident for System design interviews in 4 hours, then DM me COACH.

99

21 Comments

In system design interviews, discuss cache eviction strategies like LRU (Least Recently Used), LFU (Least Frequently Used), and FIFO (First In, First Out). These methods determine which data to remove when the cache is full, balancing memory usage and ensuring high cache hit rates for optimal performance in high-traffic systems. _ Subscribe to my weekly newsletter. Learn a software engineering concept simplified like you're five, in just 5 minutes, every week! (Link in bio)

2

Amazon S3 now supports conditional writes!! .... but why should you care? Because of these 5 reasons.. 1. Prevents Overwriting and Ensures Data Integrity: Conditional checks prevent accidental overwrites, maintaining data integrity while allowing successful creation of new objects. 2. Simplified API Usage and Server-Side Validation: Clients can perform writes and existence checks in a single operation, with Amazon S3 handling all conditional logic, reducing API calls and offloading responsibility from clients. 3. Efficient Parallel Processing for Large-Scale Workloads: Multiple clients can write concurrently without risking data loss, making it suitable for distributed operations like analytics or machine learning tasks. 4. Eliminates Client-Side Coordination and Improves Performance: By removing the need for complex client-side consensus mechanisms and reducing API calls, overall system performance is improved. 5. Consistent Behaviour Across Operations: The same conditional logic applies to both PutObject and CompleteMultipartUpload operations, providing a uniform approach to data management. Biggest one for me is the simplification of client logic. Follow Kaivalya Apte if you like #SoftwareEngineering #DistributedSystems #Databases #StreamProcessing

46

4 Comments

🚀 **Exciting News!** 🚀 MuleSoft just dropped the **new SFTP Connector (v2.4)** on **November 13, 2024**! 🌟 Say goodbye to KEYBOARD_INTERACTIVE auth and hello to **PUBLIC_KEY, PASSWORD, or GSSAPI_WITH_MIC** methods. Get ready to streamline your file transfers like never before! 📈✨ #MuleSoft #SFTP #TechUpdate

6

𝐁𝐮𝐢𝐥𝐝𝐢𝐧𝐠 𝐨𝐧 𝐎𝐩𝐞𝐧𝐀𝐈'𝐬 𝐠𝐩𝐭-4𝐨-𝐫𝐞𝐚𝐥𝐭𝐢𝐦𝐞 𝐀𝐏𝐈.  Here are few things I learned about it today. 📌The underlying model is actually 𝐠𝐩𝐭-4𝐨-𝐚𝐮𝐝𝐢𝐨-*. It is optimized to operate at lower latency with a WebSocket interface. 📌It can respond to audio inputs at an average of 320 𝐦𝐢𝐥𝐥𝐢𝐬𝐞𝐜𝐨𝐧𝐝𝐬, which is similar to human response time in a conversation. 📌It requires either audio output or audio input to be used at this time. Acceptable combinations of input and output are: ➡️text in → text + audio out ➡️audio in → text + audio out ➡️audio in → text out ➡️text + audio in → text + audio out ➡️text + audio in → text out 📌You can also access gpt-4o-audio-* models via 𝐜𝐡𝐚𝐭 𝐜𝐨𝐦𝐩𝐥𝐞𝐭𝐢𝐨𝐧𝐬 𝐞𝐧𝐝𝐩𝐨𝐢𝐧𝐭. Just that it may not be suitable for real-time conversations due to the latency added by http protocol. 📌Pricing seems to be the same with Audio 𝐢𝐧𝐩𝐮𝐭 𝐜𝐨𝐬𝐭𝐬 ~6¢ 𝐩𝐞𝐫 𝐦𝐢𝐧𝐮𝐭𝐞 and 𝐨𝐮𝐭𝐩𝐮𝐭 𝐜𝐨𝐬𝐭𝐬 ~24¢ 𝐩𝐞𝐫 𝐦𝐢𝐧𝐮𝐭𝐞 . Found this useful? If you know someone planning to use this API, share this with them! #ConversationalAI #VoiceAI #MachineLearning #generativeai

39

12 Comments

A very useful detailed blog by Georgios - Nikolaos Palaiologopoulos which explains spring caches with code samples on demo MVC application. A good refresher on application data caching! https://lnkd.in/gy8AxV8D #spring #caches

17

Highly recommended for senior engineers.

10

Did you know there are no "roles" (SDE, SDET, TPM, …) when you get to L8 IC (Senior Principal) at Amazon? It’s one of Amazon’s peculiarities. [Side note: in general, Amazon SDEs tend to be "jack of all trades" where we pragmatically do whatever is necessary to make our organization successful at that given moment (within reason). In fact, I believe being well-rounded and able and willing to wear a number of hats is one of the things that makes me valuable to an employer. We don’t over-index on "that’s not part of my role doc." Other companies where I worked (eg Google) have very clear swimming lanes and people stick to them pretty religiously, eg. "oh no I don’t talk to customers, I’m an SDE and talking to customers is for TPMs." Whereas at Amazon I never once had a manager tell me, "what you did (although impactful, important and necessary) was not really your job." ] My little side note about this critical piece of amazon culture explains why as you get to L8 the concept of a "role" no longer really makes sense. The boundaries between roles get fuzzier and fuzzier as you go from Senior Engineer (L6) to Principal Engineer (L7), and they become unimportant when you get to L8. We’re all just “Senior Principals” - whether you came via the SDE path, the TPM path, or in my snowflake case, the SDET path. In fact I think most L8s are snowflakes in that their path to L8 was unique and the value they bring to the company is also unique. In phonetool (our amazon directory) you can have whatever unofficial title you want (also within reason...), so I chose to keep the "SDET" title to reflect my background and passion for testing, engineering and operational excellence, and developer tools. I also hope that by keeping the title, others in my (old?) role can see that SDETs at Amazon do not have a "ceiling" in their career. And lastly, it's a fun conversational piece ("you're an L8 SDET???") (The image here is "Peccy" - a beloved little icon that has represented Amazon's Peculiar Ways for years) [Edit: Strictly speaking there are a few L8 IC roles that do have a different job code, like PMT or Scientist - but I think we generally view all L8 ICs in the same lens]

929

41 Comments