Minka is a Java library for distribution and balance of backend systems through sharding of user-defined duties.

What’s for ?

Used to horizontally scale up usage of physical resources by splitting processing workload into application instances, and keeping a dynamic distribution and balancing of input data.

Usage and backend model

Duties	Pallets	Delegates	Shards	Leader	Followers

user-defined distributable entities, representing input data or signal to user functions.	group duties, with specific balancing strategies based on lifecycle, costs and performance.	the API client applying the contract of reacting on distribution events sent to the shard	fragments of the duty universe, each machine in the cluster.	coordinates client CRUD requests. Controls everything.	receives duties and pallets from the leader, to invoke delegate’s contract.

Features

Highly available service of distribution of duties and resilient to failure
Several balancers to achieve continuous task distribution
Mutipurpose abstract model with an Inversion of Control non-intrusive integration
Isolation of scaling concerns out of the business layer by encapsulation

Implementation flow

Thru the inversion of control pattern, the host application yields the creation, update, deletion, shard selection, and transportation of duties to Minka. The following implementation flow takes place:

request	reception	transport	available
client builds a duty and the CRUD request thru API	minka routes request to the current leader	leader ensures consistency and distributes duty to the shard	follower process invokes delegate’s contract onCapture or onRelease