By Sramana Mitra and guest author Shaloo Shalini
SM: What about integration you have done so far? Would you share your thoughts on what role integration has played, how expensive it been, and for your future plans, how are you thinking about integration costs and so forth?
PS: Hopefully, you can bear with me because this is my area of expertise, so I am very interested in it.
SM: Absolutely. We are very interested in your expertise.
PS: Now, if I may deviate for a moment. I’m pursuing my doctorate at the school of information studies, and my studies are on using service-oriented architecture (SOA) to enable unified information architecture. What’s interesting about this is these cloud or Internet-based solutions or becoming more prevalent, and there are more options that have very high value in terms of point solutions in a particular domain. If we go back to the evolution of computing with enterprise systems, in the 1990s and early 2000s people were addressing an integration problem that existed inside the organization. Enterprise resource planning (ERP) solutions solved that integration problem, but they solved it vertically and created silos. With the advent of cloud computing, we are again dis-aggregating, separating, or breaking apart monolithic enterprise systems. We are doing it more cheaply and more effectively from a point solution perspective. To go back to your point, the integration problem is back and is back bad. How do we integrate applications that are all over the globe and in different data centers? They will be on-premise and off-premise; there will always be a hybrid, I think, of systems. How do you handle integration? So I wanted to give you this backdrop.
The integration problem from a computing perspective needs to be addressed by SOA, so by that definition implies a loose coupling. In order to integrate, you need to adopt a service or architecture paradigm and to understand this concept of abstracting the technology, which is agnostic to the business functions. The level of granularity will vary. At one point it is a transportation management system; at another point it might be a Web service checking inventory. So, there is an entire domain. The integration problem for a small or medium-sized company and, I would argue, even for a larger company, is not to manage dozens or hundreds of point solutions with or without XML or Web services standards – there are too many points of connection. My philosophy is we need to embrace SOA. From an SME perspective, integration should be in the cloud. That is where the action is – don’t try to integrate inside the enterprise. Put integration in the cloud, use Integration as a Service. Allow the integration provider to be your hub and spoke using an enterprise service bus environment to essentially integrate from any format, any protocol. I can give you a lot more detail about what that looks like. But that in practice is the approach that we are taking.
SM: Tell me more about the technologies and vendors you use to accomplish this.
PS: I have been in this for 25 years, so I have seen cycles of computing. You know about EDI. EDI has been predominant for 50 years in integrating supply chains, and it has worked pretty well. The problem with EDI is that it is proprietary system and a costly system, and SMEs , which are increasingly becoming part of the value chain, can’t play too hard. If you take the EDI model and the idea of a VAN – a value-added network, the idea of hub and spoke, and you take a progressive VAN – you allow them to be cloud integration provider. That is a powerful concept that has stood the test of time. But instead of just using EDI transactions, it could be file transfer, Web services, or onboarding cloud applications. They have the deep domain expertise and infrastructure to do that.
The company that we partnered with is called Liaison Technologies. We looked at others like Hubspan. These are companies that have been doing EDI for decades and are now embracing the cloud and private public integration through service-oriented infrastructure that exists in the cloud. I can’t afford to buy TIBCO or webMethods for my dollars. I can’t afford to manage an enterprise service bus. I just can’t do it, but they can. We have a virtual private network connecting to this cloud provider – this integration as a service provider – with just one VPN, one connection. They have a Web service and we have a Web service. Web services handle all of our integration requirements. It is really a beautiful model. Does that make sense to you?
In our case we have hundreds of inputs into the cloud that go through Liaison. They do transformation, governance in terms of monitoring, and they send it to us in the desired format right into our on-premise environment – this VPN connection comes right into our environment and invokes tier 2 Web services to insert into a database to do whatever processing in our environment. The only thing we manage are our own Web services. We are abstracted and loosely coupled, we can accommodate any combination, we can have infinite scale, and we pay for what we use. All for a fixed price.