Replicated Key Value storage using Android

-
This was my fourth project in Android. This project involves designing a simple distributed replicated-value storage based on Dynamo. It is a simplified version of Dynamo, which includes partitioning, replicating and handling failures. The Simple Dynamo is built on top of the android content provider.

 System Architecture

     Design Overview

The design is pretty simple as seen from the above diagram. The Dynamo (Content Provider) has a server thread and client threads which communicate by sending messages. Quorum manager at the client side will check whether reader quorum size and writer quorum size is satisfied in the respective cases. There are different types of message for Put, insert, query, recovery etc, which simplifies the application logic.

Handling Failures

Case: Coordinator fails
   
    If the coordinator node fails, then its first successor is contacted to see if the operation can be     continued with the first successor acting as a make shift coordinator; if at this point, even the first successor has failed, then we proceed to contact the second successor which will now take on the role of make shift coordinator. If this whole process fails, we retry after a reasonable delay.


Case: Successor fails




Since only one node failure is assumed in the system at any given point in time, the quorum for read/write will be satisfied by getting two votes, one from the coordinator and the other from a successor, and the operation will continue seamlessly.


Node Recovery

Case: Get values from first predecessor and first successor | second successor
Case: Get values from second predecessor | second predecessor and first successor

In either of the above case the system can recover all its values successfully. These are the ideal cases given our system constraints. 

Comments

Post a Comment

Popular posts from this blog

Difference between "diff" and "sdiff" commands in Unix

-
Image
Hi everyone, in this blog post, I will be discussing about the difference between "diff" and "sdiff" commands in Unix. This is my first post about Unix. diff command diff command will compare two files and prints the difference between the 2 files. The format of diff command is diff file1 file2 . sdiff command sdiff command will compare two files side by side. The format of sdiff command is sdiff file1 file2. The contents of files are displayed side by side with a marker denoting the difference between each files. Some of the important markers used in sdiff are as follows: | - denotes the 2 files differ in the line completely. < - only the first file contains the line. > - only the second file contains the line. The better way to depict the difference is by showing examples. Here, I am comparing 2 files which are named as " ipl_teams_2008.txt " and " ipl_teams_2015.txt ". Let's see the contents of the files. ...
Read more

Important annotations used in Hibernate

-
Hi everyone, in this blogpost, I am discussing about some of the important annotations used in Hibernate. Following are some of the important annotations: - @Entity - @Entity specifies that the class is marked as an entity bean . @Table -  @Table is used to specify the table for persisting data. The optional attributes of the table annotation is ' name ' and ' schema ' which specifies both the name and schema of the table respectively. Other attributes are ' unique constraints ' which specifies the unique constraints that are to be placed on the table and ' catalog ' which specifies the catalog of the table. @Id - @Id specifies the identifier property of the entity bean. It is like a primary key of the table. The mapped column for primary key of the entity is assumed to be primary key of the primary table. If no column annotation is specified, the primary key column is assumed to be the name of primary key property or field. @Column - @C...
Read more