Case Study

Seamless Operational Data Migration: A Genentech Story

By Zach Wagner, Robosys Automation, and April Phillips, Genentech

Data Network Business

Genentech’s South San Francisco manufacturing site ran four separate instances of its legacy historian to cover various manufacturing functions and plants. Three out of four of these instances included collection of batch data, using 15 different batch models to collect that data.

In 2018, Roche (Genentech’s parent company) inked an enterprise agreement with OSIsoft to facilitate phasing out of its legacy historian, but the transition included numerous obstacles. The legacy historian contained more than 20 years of data, comprising over 75,000 tags, and had been integrated with multiple systems across the plant, including several manufacturing execution systems (MES). It also was integrated with legacy control systems, in some cases filling gaps in those control systems.

Further, Genentech had developed a suite of custom tools to support and monitor its system, some of which had become integrated in different business processes across the manufacturing plant — thus, the tools required migration, as well. Finally, 1,000+ users would have to be trained to use the new historian following migration.

Following discussions with OSIsoft, Genentech implemented two PI Systems.1 Each System includes a PI Collective,2 a PI Asset Framework3 (AF) server, an Analytics4 and Notifications5 server, and interfaces configured over failover. Both Systems use a single vision instance to allow users to more easily connect and view data. Data collectors on the legacy historian were upgraded to be redundant and failover — so, if any issues occurred during data migration, Genentech had multiple connections. The automation server level also was retained.

Figure 1 – Mapping from legacy system (L) to PI system (R)

Key objectives for the data migration included completely divesting the facility from its legacy historian — no more licensing costs and no need for users to operate two different systems. Genentech also needed to be able to show 20+ years of data had been moved with 100-percent data integrity.

Additionally, Genentech wanted to retain its suite of custom tools, along with business processes dependent on these tools, and reuse its existing OPC servers. The company had established OPC connections with the more than 550 different control systems from which data was collected and it would not have been possible to completely duplicate this OPC layer.

Genentech also needed to ensure no data gaps existed due to data migration since, without the data to support production of goods manufactured, the company would not be able to release the products for sale. Finally, Genentech needed to migrate existing batch data, as well as continuous data, because batch data could not be regenerated.    

Using the aforementioned custom tools, Genentech was able to map its legacy historian tags to PI Tags.6 The company utilized its existing OPC server layer to start simultaneous data collection on both the legacy system and the PI System. Once simultaneous data collection was confirmed, Genentech performed an archive shift. From that point forward, PI was considered the system of record.

Genentech’s enterprise agreement with OSIsoft allowed the company to stand up/license an intermediate PI System, facilitating data migration on the intermediate system, rather than the live production system. Thus, when data migration was complete, Genentech could simply register the archives and see all historical data without having to do archive remapping via the piconfig7 utility.

Figure 2 – Data migration process facilitated by a PI intermediate system

Genentech built a Windows Form C# application — an ODBC connection to the legacy historian — and used PI AF SDK8 to connect to the intermediate proxy system. This tool could run in three ways: in backfill, in validate, or in both backfill and validate.

Figure 3 -- PI system validation (L-R): time of execution, tag migrated, account in PI, count in legacy historian, start time of data "chunk," end time, confirmation that counts matched

To migrate historical batch data, Genentech again used a Windows Form C# application and an AF SDK connection to the PI System, leveraging its own custom tools to facilitate the migration. The company exported batches in XML format from the legacy database, then parsed those XML files and auto-generated the Event Frames9 (allowing us to generate batch reports identical to the legacy system). Additionally, some of the existing batch functions use XML exports, allowing use of the migration tool to generate live batches moving forward.

Conclusions

Genentech was able to migrate 20+ years of continuous and batch data from multiple legacy historian instances to the PI System, as well as validate, with no data loss. Benefits included greater system stability, reduced disc space necessary to store the same data, reliable access to all data for users, and the enabling of operation in a web-based client with Asset Relative displays. All of this was supported by vendor-provided training10 and thorough documentation of applications.

Resources

  1. PI System
  2. PI Collective
  3. PI Asset Framework (AF)
  4. PI Analytics
  5. PI Notifications
  6. PI Tags
  7. piconfig
  8. PI AF ADK
  9. PI Event Frames (EF)
  10. OSIsoft Learning (YouTube channel)