𝗥𝗲𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗦𝗵𝗮𝗿𝗶𝗻𝗴 𝗶𝗻 𝗗𝗮𝘁𝗮𝗯𝗮𝘀𝗲𝘀 – 𝗪𝗵𝗲𝗻 𝗮𝗻𝗱 𝗪𝗵𝘆?

𝙃𝙖𝙣𝙙𝙡𝙞𝙣𝙜 𝙡𝙤𝙖𝙙𝙨 𝙤𝙛 𝙧𝙚𝙖𝙙𝙨 𝙖𝙣𝙙 𝙬𝙧𝙞𝙩𝙚𝙨 𝙤𝙣 𝙖 𝙨𝙞𝙣𝙜𝙡𝙚 𝙙𝙖𝙩𝙖𝙗𝙖𝙨𝙚 𝙘𝙖𝙣 𝙗𝙚𝙘𝙤𝙢𝙚 𝙤𝙫𝙚𝙧𝙬𝙝𝙚𝙡𝙢𝙞𝙣𝙜. 𝙏𝙤 𝙢𝙖𝙣𝙖𝙜𝙚 𝙩𝙝𝙞𝙨, 𝙬𝙚 𝙘𝙧𝙚𝙖𝙩𝙚 𝙘𝙤𝙥𝙞𝙚𝙨 𝙤𝙛 𝙤𝙪𝙧 𝙙𝙖𝙩𝙖𝙗𝙖𝙨𝙚 𝙩𝙤 𝙝𝙖𝙣𝙙𝙡𝙚 𝙢𝙤𝙧𝙚 𝙧𝙚𝙦𝙪𝙚𝙨𝙩𝙨 𝙨𝙞𝙢𝙪𝙡𝙩𝙖𝙣𝙚𝙤𝙪𝙨𝙡𝙮. 𝙏𝙝𝙞𝙨 𝙘𝙤𝙣𝙘𝙚𝙥𝙩 𝙢𝙤𝙨𝙩𝙡𝙮 𝙖𝙥𝙥𝙡𝙞𝙚𝙨 𝙩𝙤 𝙉𝙤𝙎𝙌𝙇 𝙙𝙖𝙩𝙖𝙗𝙖𝙨𝙚𝙨.

𝗟𝗲𝗮𝗱𝗲𝗿-𝗙𝗼𝗹𝗹𝗼𝘄𝗲𝗿 𝗠𝗲𝗰𝗵𝗮𝗻𝗶𝘀𝗺 (𝗠𝗮𝘀𝘁𝗲𝗿-𝗦𝗹𝗮𝘃𝗲)

• One leader handles all 𝘄𝗿𝗶𝘁𝗲𝘀. • Followers are copies of the leader and handle reads. • Followers may not always be fully up-to-date; this is called 𝗲𝘃𝗲𝗻𝘁𝘂𝗮𝗹 𝗰𝗼𝗻𝘀𝗶𝘀𝘁𝗲𝗻𝗰𝘆. • Data syncing from leader to followers is called 𝗿𝗲𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻, which can be 𝘀𝘆𝗻𝗰𝗵𝗿𝗼𝗻𝗼𝘂𝘀 or 𝗮𝘀𝘆𝗻𝗰𝗵𝗿𝗼𝗻𝗼𝘂𝘀. • In synchronous replication, writes are slower because the leader waits for followers to update before completing the write. • Reads are usually routed to followers to reduce latency and spread load. • If the leader fails, a follower can be promoted to leader.

𝗟𝗲𝗮𝗱𝗲𝗿-𝗟𝗲𝗮𝗱𝗲𝗿 𝗥𝗲𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 (𝗠𝘂𝗹𝘁𝗶-𝗟𝗲𝗮𝗱𝗲𝗿)

• 𝙈𝙪𝙡𝙩𝙞𝙥𝙡𝙚 𝙣𝙤𝙙𝙚𝙨 𝙖𝙘𝙩 𝙖𝙨 𝙡𝙚𝙖𝙙𝙚𝙧𝙨, each handling both reads and writes independently. • This allows scaling write capacity by distributing writes across leaders. • Because nodes sync asynchronously, the system may have lower immediate consistency (𝙖𝙡𝙨𝙤 𝙠𝙣𝙤𝙬𝙣 𝙖𝙨 𝙡𝙤𝙤𝙨𝙚𝙡𝙮 𝙘𝙤𝙣𝙨𝙞𝙨𝙩𝙚𝙣t). • Conflicts can arise if two leaders make conflicting writes, needing conflict resolution. • Works well for distributed systems with users across regions, offering high availability and fault tolerance. • Compared to synchronous replication, it offers lower latency but weaker consistency.

𝗦𝗵𝗮𝗿𝗱𝗶𝗻𝗴 • Splits a large database into smaller parts called shards to handle huge datasets. • Data is partitioned based on a key, which can be:

1. Range-based (e.g., A-L, M-Z)
2. Hash-based (data distributed by a hash function)
3. NoSQL databases often have built-in sharding; SQL databases usually require manual sharding.
4. Manual sharding in SQL databases can compromise ACID guarantees making transactions less reliable.

𝘜𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥𝘪𝘯𝘨 𝘵𝘩𝘦𝘴𝘦 𝘤𝘰𝘯𝘤𝘦𝘱𝘵𝘴 𝘩𝘦𝘭𝘱𝘴 𝘵𝘰 𝘥𝘦𝘴𝘪𝘨𝘯 𝘥𝘪𝘴𝘵𝘳𝘪𝘣𝘶𝘵𝘦𝘥 𝘥𝘢𝘵𝘢𝘣𝘢𝘴𝘦 𝘴𝘺𝘴𝘵𝘦𝘮𝘴 𝘵𝘩𝘢𝘵 𝘣𝘢𝘭𝘢𝘯𝘤𝘦 𝘱𝘦𝘳𝘧𝘰𝘳𝘮𝘢𝘯𝘤𝘦, 𝘳𝘦𝘭𝘪𝘢𝘣𝘪𝘭𝘪𝘵𝘺, 𝘢𝘯𝘥 𝘥𝘢𝘵𝘢 𝘤𝘰𝘯𝘴𝘪𝘴𝘵𝘦𝘯𝘤𝘺 𝘢𝘤𝘤𝘰𝘳𝘥𝘪𝘯𝘨 𝘵𝘰 𝘴𝘱𝘦𝘤𝘪𝘧𝘪𝘤 𝘢𝘱𝘱𝘭𝘪𝘤𝘢𝘵𝘪𝘰𝘯 𝘯𝘦𝘦𝘥𝘴.