Smart Routing of Request across Muti Datacenter to Optimize resource latency and Data consistency (30-Sep-2011)

Document Author (alias)

Anurag Handa (ahanda)

Defensive Publication Title 

Smart Routing of Request across Muti Datacenter to Optimize resource latency and Data consistency.

Anurag Handa (ahanda)

Don Hemme   (donhemme)

Forest Gower (forestg)

Jyothish Nair (jnair)

Nayab Khan (nayabk)

Toji Thomas Oommen (tojio)

Summary of the Defensive Publication/Abstract

Currently in a typical Service Oriented Architecture (SOA) based applications deployed in multiple geographically diverse data centers with one write data center and multiple read data centers, client applications are  penalized with high network latency to ensure data consistency after write operations. Our solution optimizes the client calls to route them to their closest data center and tries to keep them there as long as possible without compromising data or resource consistency.

Description:  Include architectural diagrams and system level data flow diagrams if: 1) they have already been prepared or 2) they are needed to enable another developer to implement your defensive publication. Target 1-2 pages, and not more than 5 pages.  

Service oriented architecture applications are generally hosted on multiple geographically diverse datacenters to have minimum latency impacts for Client applications. A typical configuration consists of common data centers where all the updates  happen and local read data centers which can serve data from the closest location. Some of the challeges faced in this architecture are

a)             Data Consistency : Client applications who have just performed an update operation should be able to fetch the newly updated data.

b)             Transparent Read and Read/Write logic : Client applications shouldn’t have to be aware of multiple database pairs sitting

behind the service

c)             Optimized Read Operations: Client applications just performing simple read operations should not be penalized with high

network latency issues by not going to the local datacenter

d)             High Scalability: Since there is only one datacenter which is performing write operations which is normally the biggest

bottleneck to achieve high scalability.

The solution involves creating an intermediary router service which sits in front of all the service instances and is responsible for routing the call to either the closest read instance or remote write instance. The routing service maintains a list of methods and it’s corresponding kind r...