Presentation: Building a High Performance, Large Scale Enterprise PI System

Within Shell many PI Systems are deployed from a local perspective and use case. This approach is changing as the operating model is changing to a global model. This global model will benefit from a ‘one-stop-shop’ for data to improve business efficiency. To support this business model the PI System architecture is redesigned to leverage the latest hosting technologies and latest OSIsoft versions. Local PI Systems are combined into so called Super Bucket systems. These Super Bucket systems hold all global real time data per business (like Upstream). The amount of tags will scale-up to 2.5 million.

The technology used is based on ‘In-memory’ processing of archives. The system will be equipped with 1 TB of internal memory to hold at a minimum six months of data. The current version of the Windows OS provides file caching. A start-up script will push all PI System archives of the last six months into the OS file cache. Once the PI System is retrieving data, processing is done in internal memory instead of ‘slow’ SAN systems. In combination with the high performance capabilities of PI Server, read throughputs have been measured in test scenarios at up to 2.5 million values per second.

This approach is a significant simplification of the PI System landscape, infrastructure cost and support cost. Due to simplification reliability will go up. A ‘one-stop-shop’ for data means the connected portfolio only has to deal with one form of interfacing and will also be simplified.

Users are less dependent on local knowledge of data structures and naming standards. As data access is standardized it becomes easier to compare production processes across locations. This will trigger significant additional business value as global learning and sharing best practices becomes a fact.