Distributed Computing

Overview

There are two important aspects in distributed computing which affect implementation choices:

• CAP
• BASE

CAP Theorem

The CAP theorem states that “Though it’s desirable to have Consistency, High-Availability and Partition-tolerance in every system, unfortunately no system can achieve all three at the same time”.

• Consistency - A consistent service ensures that “what you read is what you get”, based on the rule of ordering updates - i.e. if Client A writes old_value and then new_value to a certain location, then another Client B will read old_value then new_value. This means that when updates occur, then any client will see the same change immediately after the change occurred. This is also known as strong consistency.

• Availability - An available service is one where its guaranteed that for every request there will be a response about whether it succeeded or failed. This means the system is responsive, i.e. that the request will be dealt with. It doesn’t require that a specific server be available (they can fail) but nonetheless there will be others that can fulfil the request.

• Partition tolerance - A partition-tolerant service continues to operate despite failure of a part of the system. This means that some sequence of communication events might be lost between the parties.

ACID Property

ACID provides strong consistency (synchronous transactions). It gives up availability.

In traditional relational databases, ACID transactions guarantee the following properties, and these are guaranteed by a two-phase commit (which ensures this across multiple database instances when performing a transaction):

• Atomicity - Either all operations in the transaction complete successfully or none, i.e. there is no possiblity for partial success and partial failure in a transaction; if one part fails then the entire transaction fails

• Consistency - Database is consistent before a transaction and it is consistent at the end of the transaction, i.e. it’s always in a valid state

• Isolation - Transactions execute in isolation, as if it is the only operation performed upon the database

• Durability - Operations are not reversed when the transaction completes.

BASE Property

BASE provides eventual consistency (generally asynchronous transactions), i.e. weaker consistency. It gives up consistency to improve availability.

• Basically Available
• Soft State
• Eventual Consistency

References

• https://ivoroshilin.wordpress.com/2012/12/13/brewers-cap-theorem-explained-base-versus-acid
• https://medium.com/@marton.waszlavik/demystifying-cap-theorem-eventual-consistency-and-exactly-once-delivery-guarantee-ed20cf7cc877