Distributed Transactions: 2PC vs 3PC vs Saga It's a complex topic in itself. Sometimes even seasoned professionals misunderstand. ▢ 𝟮𝗣𝗖 - atomic but blocking, commit or abort in two steps. ▢ 𝟯𝗣𝗖 - splits commit into two, reduces blocking and handles partial failures. ▢ 𝗦𝗮𝗴𝗮 - a sequence of local transactions that breaks a transaction into multiple steps. 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝟮-𝗣𝗵𝗮𝘀𝗲 𝗖𝗼𝗺𝗺𝗶𝘁 → When strict consistency is needed, all participants commit or abort. → For simple, low-latency systems with minimal crash risk. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: 𝗕𝗮𝗻𝗸 𝗙𝘂𝗻𝗱 𝗧𝗿𝗮𝗻𝘀𝗳𝗲𝗿 𝗕𝗲𝘁𝘄𝗲𝗲𝗻 𝗔𝗰𝗰𝗼𝘂𝗻𝘁𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗦𝗮𝗺𝗲 𝗕𝗮𝗻𝗸 • Transfer involves debiting one account and crediting another. • If one fails, the entire transaction must rollback. • Atomicity is a must, and latency is acceptable. 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝟯-𝗣𝗵𝗮𝘀𝗲 𝗖𝗼𝗺𝗺𝗶𝘁 → When minimizing blocking is key, and partial failures must be avoided. → Where fault tolerance takes priority over message overhead and complexity. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: 𝗖𝗿𝗼𝘀𝘀-𝗥𝗲𝗴𝗶𝗼𝗻 𝗟𝗲𝗱𝗴𝗲𝗿 𝗦𝘆𝗻𝗰𝗵𝗿𝗼𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻 Synchronizing transaction records between European and Asian data centers. • Each region prepares and pre-commits. • Final commit is sent when all regions are ready. • Handles network partition or coordinator crash more gracefully than 2PC. 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝗦𝗮𝗴𝗮 → Performance matter more than strict consistency → Long-running distributed transactions where full rollback isn't practical 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗙𝘂𝗻𝗱 𝗧𝗿𝗮𝗻𝘀𝗳𝗲𝗿 (𝗦𝗪𝗜𝗙𝗧) Transferring funds from a bank in Germany to one in Singapore via SWIFT or similar clearing systems. • Debit sender → Local transaction • Notify intermediary → Local transaction • Credit receiver → Local transaction • If the final step fails, compensation (e.g., refund sender) is triggered.

Distributed Transactions: 2PC vs 3PC vs Saga It's a complex topic in itself. Sometimes even seasoned professionals misunderstand. ▢ 𝟮𝗣𝗖 - atomic but blocking, commit or abort in two steps. ▢ 𝟯𝗣𝗖 - splits commit into two, reduces blocking and handles partial failures. ▢ 𝗦𝗮𝗴𝗮 - a sequence of local transactions that breaks a transaction into multiple steps. 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝟮-𝗣𝗵𝗮𝘀𝗲 𝗖𝗼𝗺𝗺𝗶𝘁 → When strict consistency is needed, all participants commit or abort. → For simple, low-latency systems with minimal crash risk. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: 𝗕𝗮𝗻𝗸 𝗙𝘂𝗻𝗱 𝗧𝗿𝗮𝗻𝘀𝗳𝗲𝗿 𝗕𝗲𝘁𝘄𝗲𝗲𝗻 𝗔𝗰𝗰𝗼𝘂𝗻𝘁𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗦𝗮𝗺𝗲 𝗕𝗮𝗻𝗸 • Transfer involves debiting one account and crediting another. • If one fails, the entire transaction must rollback. • Atomicity is a must, and latency is acceptable. 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝟯-𝗣𝗵𝗮𝘀𝗲 𝗖𝗼𝗺𝗺𝗶𝘁 → When minimizing blocking is key, and partial failures must be avoided. → Where fault tolerance takes priority over message overhead and complexity. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: 𝗖𝗿𝗼𝘀𝘀-𝗥𝗲𝗴𝗶𝗼𝗻 𝗟𝗲𝗱𝗴𝗲𝗿 𝗦𝘆𝗻𝗰𝗵𝗿𝗼𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻 Synchronizing transaction records between European and Asian data centers. • Each region prepares and pre-commits. • Final commit is sent when all regions are ready. • Handles network partition or coordinator crash more gracefully than 2PC. 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝗦𝗮𝗴𝗮 → Performance matter more than strict consistency → Long-running distributed transactions where full rollback isn't practical 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗙𝘂𝗻𝗱 𝗧𝗿𝗮𝗻𝘀𝗳𝗲𝗿 (𝗦𝗪𝗜𝗙𝗧) Transferring funds from a bank in Germany to one in Singapore via SWIFT or similar clearing systems. • Debit sender → Local transaction • Notify intermediary → Local transaction • Credit receiver → Local transaction • If the final step fails, compensation (e.g., refund sender) is triggered.

Discover how transaction processing systems (#TPS) support real-time business transactions, the differences between real-time and batch processing, and how #OLTP compares to #OLAP systems.

aerospike.com/blog/transac...

#Aerospike #Database #DistributedDatabase #DistributedTransactions

#Windows2022
#SQLServer2019 CU30
#MSDTC
#DistributedTransactions fail !
#SQLHelp
www.sqlservercentral.com/forums/topic...

Distributed Transactions at Scale in Amazon DynamoDB This paper appeared in July at USENIX ATC 2023. If you haven't read about the architecture and operation of DynamoDB, please first read my s...

Another great write up on #distributedtransactions / #DynamoDB at $AMZN #AWS @muratdemirbas.bsky.social 👏! #SoftwareDevelopment #Databases

Transactions in a microservice world
#microservices #distributedtransactions

wso2.com/whitepapers/tr…