Conclusion
The proof of concept was successfully implemented over several weeks and demonstrated a throughput many times in excess of the requirements, so the full production system was scheduled to move forward. The production application was built over the course of a quarter to be fully tested and implemented in time for the 2011 tax season. The SharePoint 2010 farm provides the ability to scale out as needed to handle messages. A unique handler architecture developed by Atmosera provided the means to build distinct handlers for messages that were registered with SharePoint and used to process incoming messages. To handle new messages, the team simply writes a new handler and registers it with the SharePoint system and it is ready to begin processing.
Nearly 100 reports were converted from the legacy Oracle system to SSRS. These reports are hosted on SSRS nodes. The partitioned SQL Server 2008 architecture allows SFS to specify new nodes as needed based on demand. If a new or existing customer starts requesting a high load of reports, the team can specify a new node and replicate data to that node for dedicated processing of reports. The customers are able to request a report online, which is routed to the correct node through the SharePoint 2010 server and then delivered in CSV or PDF format to the end customer.
The original system processed 20 million returns a year. It peaked at 110,000 returns per hour and reports were 24 hours old. The converted system achieved up to 360,000 returns per hour and all reports are now generated near real-time.
Atmosera and Microsoft were able to demonstrate the power and scalability of the SharePoint 2010 and SQL Server 2008 technologies. The developer tooling enabled the team to write the entire system over a single quarter in time for the customer to face the 2011 tax season. The use of the Entity Framework not only accelerated development by providing a seamless integration with the database backend, but also demonstrated the maturity of the technology through the high performance and throughput it was able to achieve. The architecture will allow SFS to meet the increasing demands for tax returns in subsequent years by providing multiple tiers to scale out. The final result was over 300% improvement throughput of message processing and the elimination of the old batch-based 24-hour reporting system in favor of the new on-demand reports that are generated in real-time.