Documents ->2008
Software + Services for Engineers.
*Steven Johnston, †Phillip Marsh, *Simon J. Cox and *Kenji Takeda
*Microsoft Institute for HPC, School of Engineering Sciences, University of Southampton, United Kingdom.
†Microsoft, Thames Valley Park, Reading, United Kingdom.
Email:sjj698@zepler.org; sjc@soton.ac.uk; ktakeda@soton.ac.uk

Abstract

In this paper we describe how novel Software + Services approaches can significantly improve productivity for scientists and engineers. In partnership with Rolls-Royce, Airbus and BAE Systems we demonstrate how to accelerate the understanding of engineering systems and product development. We approached different levels of the enterprise engineering process using a series of user-driven Proof of Concept (PoCs) projects. By exploiting workflow tools and HPC capabilities we show how to formalise the engineering design and optimisation process, thus providing accountability and Quality Assurance. Using SharePoint, we demonstrate advanced engineering search capabilities and data management, seamlessly integrating existing data across multiple locations. Locating and retaining knowledge within an enterprise infrastructure is often difficult. Much of the knowledge is retained within people and social networks within organisations. Through the natural cycle of people and the difficulty in locating relevant information, much knowledge is lost or not utilised. We demonstrate how to capture and retain knowledge, exposing it through an extensible framework, providing a persistent audit trail detailing the thought process behind decisions. By abstracting infrastructure, audit and QA from engineering design we can improve the understanding of design processes thus improving overall engineering process. In combination, these approaches pave the way for providing significant reduction in time to insight and execution of scientific investigation and engineering design in real-world applications.

Introduction
Advanced product design relies on a deep scientific understanding of underlying systems to improve performance. For example, due to compressed timescales from market pressure and for instance environmental goals, Airbus aims to halve the time to market for new airliner designs. In addition, engineers need to be accountable and have to demonstrate that decisions based on firm foundations. Even once an employee has left there needs to be a trail which shows how and why decisions are made. Key decisions can only be signed-off once they have demonstrated accountability. For example, in the event of an incident, the aerospace industry requires that a decision-making audit trail is retained for many decades.

Today the computational framework is core to the engineering design process and management. By obtaining a better understanding of the design process we can improve designs, thus meeting ambitious design goals. Microsoft and the University of Southampton’s School of Engineering Sciences have developed three Proof of Concepts demonstrators with Airbus, Rolls-Royce and BAE Systems showing how current and future Microsoft tools, technologies and platforms can be exploited to make the engineering design process faster, cheaper and better. These Proof of Concepts (PoCs) aim to develop a means to dramatically increase simulation power, revolutionise the design process and transform user interface in order to provide tools to exploit multi-disciplinary optimisation and so bring products to market in extremely short timescales. Each PoC was user driven, starting with a multiday Architecture Design Session to gain an understanding of the overall process, resulting in a Vision and Scope document. Each PoC development took 2 to 3 weeks and resulted in a working completed system, demonstrated using example scenarios.

The Rolls-Royce PoC focused on supporting the end-to-end process of design optimisation coupling large scale heterogeneous computing and data handling resources to tackle challenging engineering problems. We approached an existing design optimisation problem, currently handled by a series of large scripts, offering no accountability, progress or audit capabilities and poor HPC utilisation. This was then transferred into a visually easy to understand robust workflow using Windows Workflow Foundation. Each engineering tool is wrapped as an activity which is capable of running on an HPC cluster. Each activity performs a single task, for example parametric meshing, flow solving and optimisation. The activities are then connected together in workflows. The workflow defines a business process which is customisable by setting activity properties. Process compliance and accountability is accomplished by versioning, locking and signing off workflows. Designed for resilience, the workflow is executed using the Microsoft Workflow Runtime, which stores data using SQL Server. This design provides complete tracking of workflow progress, checkpointing for long running queries, persistence of state for failures, thus providing a comprehensive audit trail.

The Airbus PoC focused on data searching and retrieval, integrating existing business applications and heterogeneous infrastructure.  The engineering process produces large volumes of data which need to be managed across worldwide networks in a secure, fault tolerant scalable infrastructure. For this PoC, each ‘data item’, called a business object (BO) has a small set of metadata which is used to locate particular BOs.  BO’s are related to one another to produce a network of related data. This produces two key problems i) how to securely search for a BO and ii) how to move the large volumes of data. All of PoC features had to be accessible from a platform independent application, called FlightPad, currently utilised for existing engineering tasks.
In this PoC we used a multi site SharePoint index to store all the metadata, consuming user and group credentials from Active Directory. This provided us with secure and rich user/group permission capabilities utilising an existing Active Directory infrastructure.
Using Search Server we created custom capabilities to allow rich searching for a BO by specifying metadata of a related BO. All of the search capabilities were exposed through WCF services, integrating into FlightPad using Python. Administration and object manipulation was also made available through SharePoint Portal webpages. 
Moving the large volumes of data was managed as a separate in-house feature of the PoC. Data transfer requests are queued and pulled from the destination, reducing network overloading. Multiple sources are possible helping to loadbalance network traffic.
This PoC made searching of complex metadata accessible to platform independent clients which was previously not possible. By utilising mainstream applications like SharePoint we were able to secure user and group access to BO data, down to a BO level. This work also supported the concept of promoting a BO from personal use, to group access, to project access, bases on a secure group promotion system, using AD groups and group owners. This ensures that promotion only takes places by authorised users and non promoted data is not accessible by other users.
The BAE PoC addressed how to support and inexperienced engineer undertaking a new task, enabling communication, capturing knowledge and decisions as they are made.
The engineer begins by researching his task using reference documents, previously executed analysis and chatting to experts in the area. As the engineer searches for related information the system can show a list of expert users who are available on line, through Office Communicator. Any documents, analysis results or conversations are recorded as possible events relevant to the task.  Based on the research, the engineer has enough information to perform his own analysis using HPC capable workflows, which will in turn generate data. Using this data the engineer can demonstrate which design option is best suited. Once engineer considers the task completed it can be signed-off as an accepted decision. Capturing this knowledge is difficult and often results in an overhead which employees dislike. This PoC imposed a minimal overhead to employees by monitoring their normal work behaviour to build up a knowledge repository since the value of a given engineering tool lies in ‘applying the tool’ rather than the tool itself. 
As the engineer reaches milestones in the task she can select which of the recorded events were used to make her decision and associate them with the milestone. This collecting of events is automatic but the association is manual to provide the engineer the ability to remove unsuitable events.  These items then become searchable in the knowledgebase for other users to access.
The system provides a clear and concise audit trail in databases which are easily queried to display decisions and related data, for example by Reporting Services. Three databases were implemented, one to support the overall tracking and updating of tasks with associated events and milestones, another to record and hold associated data (such as conversations and documents), the third supports the Windows Workflow Foundation running HPC based computations. All the data stored is indexed using Full Text search in SQL server 2008 and filtered with ifilters supporting over 40 data types. Each events and milestone builds up a timeline clearly showing how a task is progressing and can also be used from a management prospective to report on task progress.
Summary
Each of the PoCs addressed a different level in of an organisations product development, providing an end-to-end solution. The Rolls-Royce PoC provided a resilient method to formalise engineering processes, supporting complete accountability. The Airbus PoC focused on storing and retrieving the data produced by the engineering processes. It also provided advanced searching across complex data relationships whilst maintaining access permissions determined by user credentials. The BAE PoC concentrated on knowledge retention, providing engineers with resources, documents and information related to an engineering process.
While the problems addressed here focus around aerospace engineering processes, they tackle issues that are faced by many other scientific and engineering users and organisations. Other domains are experiencing increased expectations to deliver science quicker, improve knowledge re-use and come under more regulation and scrutiny, for example, in climate science, drug design, life sciences and biomedical applications. The approaches and core technologies used are industry standard and the problems investigated are general cases, making this approach applicable across many domains.

Sources and further information

Download presentation (PDF)