In my previous post i talked about building distributed consensus using paxos. In continuation of that lets talk about Raft protocol in this post. This is “the Raft paper”, which describes Raft in detail: In Search of an Understandable Consensus Algorithm (Extended Version) by Diego Ongaro and John Ousterhout. …

This is my old blog post, moving it to medium A simple sorted map. The keys of this map correspond to (RowKey,ColumnName,Timestamp) pairs. Update to row is atomic. It’s not the cell but the complete row. So two different clients can’t concurrently update different cells of the same row. …

This is an article from my old blog post and i am moving everything to medium. Dynamo’s way of handling write conflicts A simple key value store that’s eventually consistent. It gives precedence to the AP (Availability an partition tolerance) in the CAP. Availability is achieved by replicating the data. In case of network partitions that…

This paper starts with the limitations of using ACID for distributed databases. Primarily it worries about the 2PC protocol wherein the failure of a individual database participating in the protocol results in the failure of the complete system affecting the availability of the system. The author proposes that we separate…

Paxos is a consensus algorithm that widely finds it’s use in lot of consensus systems. The google’s chubby lock service made paxos very famous. Paxos in general is considered as one of the tough algorithms to understand. …

At the heart of every distributed system is a consensus algorithm. Consensus is an act wherein a system of processes agree upon a value or decision. In this post let’s look at two famous consensus protocol namely two phase and three phase commits widely in use with the database servers. …

The concept of time is fundamental to our way of thinking about ordering of events in a system. …

A simple way to distribute pairs among a set of servers is to hash(key) and locate it in a server at (hash(key) MOD NUM_SERVERS). If the hash function used for hashing is uniform and randomly distributed, the data get’s distributed properly. However, when the number of servers changes suddenly, most…

In continuation with the previous post on consistent hashing, today we will look at a distributed lookup protocol by name Chord. The protocol was published as part of paper “Chord : A Scalalable Peer To Peer lookup service for Internet Applications” by MIT Laboratory of Computer Science. …