The Need for a High-Performance Integration Platform in the Cloud Era
Dipanjan Sengupta, Principal Architect, Cognizant Technology Solutions, US
Integration strategy in the cloud-era primarily centers on managing a distributed application portfolio that spans not just across locations but are also on different deployment models and how such an application portfolio coexist, complement each other and interoperate with considerations on security, flexibility of choice and performance. The challenges are considerable, ranging from performance and latency of transactions over a public network, near real time synchronization of application data and business entities, governance of this heterogeneous portfolio and the time to adopt or get on-boarded. Likewise, the integration strategy has impacts on operational efficiency, time to market, operational cost and overall governance of the distributed application portfolio.
Dipanjan proposes to present the need of a high-performance, service-oriented, integration platform that can address the challenges of cloud-era integration. Users of this platform should be in a position to author their specific integration flows for connecting any of their source and target applications or systems, expose these flows as services, specify the conditions under which the services are to be invoked, consume these services at run time and of course govern and manage these services and the flow configurations.
But are the traditional wire formats namely XML or JSON the most optimum formats for exchange of payloads over a public network? Dipanjan argues that faster formats like Protostuff, Google Protobuff, Fast Infoset, etc. reduces the payload size significantly and offers a high-performance, low-latency payload exchange capability.
Another aspect is the data model. Most of the cloud-to-cloud, on-premise-to-cloud and cloud-to-on-premise integration scenarios today, involve entity (like say, Customer, Order, Invoice, etc.) information & state synchronization across the different participating systems.
To reduce the on-boarding time for any tenant hence, the integration platform can offer Entity Services with entity data models compliant with common industry standards or generic data models based on the Universal Data Element Framework (UDEF). Users should be able to subscribe to these Entity Services and configure them for specific use.
HTTP is the de facto transport protocol over a public network. But many traditional on-premise applications may not be in a position to exchange data on this protocol. Exchange of data through file systems, message queues and relational databases continue to be very popular. Keeping in mind that any integration solution should be in a position to support both real-time and batch integration requirements, there is a need for a lightweight, on-premise side adapter component that can be installed at the users' on-premise data center and can be configured for interactions with the on-premise systems on the one hand and with the integration platform on the other (over HTTP).
One other requirement from an integration platform is the support for all the major integration patterns. Typical enterprise scenarios may not only require the basic patterns but also the composite patterns combining the basic ones.
During the session, Dipanjan will present and demonstrate some of the interesting real-life use cases of a high-performance, service-oriented integration platform developed by Cognizant Technology Solutions using best of breed open source technologies and frameworks.
1. Challenges of Cloud-Era Integration and Solutions
2. Use of Faster Wire Formats, Low-Latency Processing and Optimized Serialization and Deserialization for Achieving High Performance Integration
3. Cloud Application Interoperability
Dipanjan Sengupta is a Principal Architect with the Global Technology Office of Cognizant Technology Solutions with extensive experience in service-oriented integration, integration of cloud-based and on-premise appl