PI scalability in hub-and-spoke model

[Former Member]

Hi,

I have a fundamental question regarding the technical architecture for exposing and consuming services in ESOA. I fully understand that PI is the core of ESOA by hosting the ESR, Service registry, BPM capabilities etc. However, what is the recommended architectural model for leveraging PI? Should one use the hub-and-spoke model and have PI handle all the service requests? Has anybody used PI to expose performance-critical services with heavy workload like the ones that support an OLTP application? Will PI scale?

Otherwise, what are the potential drawbacks of using PI at design-time and not at run time - i.e. consume services using a point-to-point model?

I appreciate any inputs on this.

-Suresh

[Former Member]

Hi Suresh,

It is always better to go for point to point model instead of going for middleware like XI only if your landscape and interfaces are small.

If you have huge no. of integrations , then XI can serve better in terms of reusability, monitoring and scalibility.

Yes, you can achieve a good performance through XI by fine tuning and enhacing hardware .Take the example of PI 7.1 which would contain adapter to adapter communication bypassing the integration engine. Enhancements are coming and will continue in future.

But, if u compare JCO connections with Web Services, no doubt that JCO is fast.But the complexity involving coding, maintainace, scalibility is a big headache. Technology and processes are changing daily. So it is better to go for loosely coupled architecture and also scalibility.

It depends on your need whether to go for XI or not.

If you see future fafter two years, SAP is converting everything to ES. So, it is always better to go with the river's direction.

Dont worry with the XI performance. All needed is true and good consultants who can tune the server.

Regards,

Piyush