MTU Aero Engines Runs on SGI Servers

By Srinivas Kodiyalam

 

The large models created in the development of aerospace engines for the German defense forces is only the beginning of the massive IT operations for MTU Aero Engines. The companymoved its SAP Advanced Planning & Optimization program over to its SGI Altix server platform.

As the lead industrial partner to the Bundeswehr, the unified armed forces of Germany,  MTU Aero Engines is responsible for nearly every type of aircraft engine operated by the German defense force. The Munich-based company also develops,  produces, sells, and maintains commercial aircraft engines as well as stationary industrial gas turbines. Its technical database and high-end computational needs are monumental.

  To help handle those tasks, MTU has relied on SGI servers to drive its complex information technology (IT) operations for more than ten years. Recently the company moved to further optimize its database environment by migrating its component for production planning, SAP Advanced Planning &  Optimization (SAP APO), to its existing Linux OS-based SGI Altix server platform.

  Step-by-Step Migration
MTU embarked on the SAP APO project in mid 2005. To manage the project, the company assembled a team that leveraged the combined expertise of representatives from MTU’s IT department, service provider T-Systems, and SGI. Working together, the team was able to design and implement the new system in just six months.

  In the first stage of the project, the team upgraded three SGI Altix 350 servers already in use at MTU. To optimize the servers for the new SAP application, they developed a configuration in which one server was used for live operation of the SAP APO component, another was converted for quality control and development, and a third was deployed as a “hot standby”  system that operated in parallel.

“By recycling these relatively new hardware platforms, we were able to prolong their operating life by four or five years,” said Norbert Diehl, IT manager of technology at MTU. “This alone represents a cost savings of 15 to 30 percent.”

  After Oracle tests were successfully completed on the SGI Altix 350 servers, the SAP APO component was installed, tested, and put into operation in parallel with the production system. The team then set up and tested the system’s high-availability functions and provided access to the new IT infrastructure for a limited number of users. In November 2006, the full range of SAP APO functions was made available. Today, around 150 MTU employees use SAP APO on SGI Altix 350 servers.

 

The ability of SGI computer technology to maximize both the performance and availability of even very large SAP applications was immediately evident to MTU in the test phases.

Impressive Results in the Test Phase
The ability of SGI computer technology to maximize both the performance and availability of even very large SAP applications was immediately evident to MTU in the test phases.

“The SGI platforms had already proven to be the best platforms for our technical IT applications,” said Diehl. “Even in extremely data-intensive environments, they run faster and are more stable than proprietary platforms — and they give us the flexibility we need to satisfy the highly complex requirements of our business areas. The combination of SAP and SGI solutions provided impressive results right from the start.”

  Driving Significantly Faster Data Transfer
SGI Altix servers combine the company’s high-performance NUMAflex connection with a global shared memory architecture running on Intel Itanium processors and open-source Linux operating systems.

  NUMAflex is uniquely designed to share system memory globally, across nodes. As a result, data records do not have to be split into smaller fragments for handling by individual processors. Instead, all processors in the system environment have one uniformly addressable memory space where they can jointly handle data. This means that data can be transferred up to 200 times faster than in traditional clusters, so that even highly complex calculation problems can be solved in record time.

  Through its global shared memory approach, the SGI Altix platform provides applications with a high level of performance, scalability,  and availability, enabling access to complex data records in real time and making records easy to manage and maintain. Globally accessible memory also allows companies to reduce their total cost of ownership and maximize their return on investment.

  Costs can be reduced further by combining SGI technology with the highly scalable Linux operating system. Using the open-source Linux platform, companies can consolidate their existing IT resources more effectively, implement new technologies more easily, and adjust their IT strategies flexibly to address shifting business requirements.

  Successful Partnership with SAP
Like MTU, an increasing number of SAP enterprise customers worldwide are opting for SGI servers to satisfy the growing need for dynamically extendable computing platforms. Both SGI and SAP share the goal of providing users with easily upgradeable, powerful, and highly available technologies. The two companies provide complementary technology that makes it possible for customers to respond to shifts in market requirements while working more profitably, faster, and systematically.

  SGI and SAP are both committed to open-source technology,  which can bring about significant savings in both direct and indirect IT costs compared to closed operating systems. SGI is a long-time member of the SAP LinuxLab, where SAP collaborates with selected partners to provide enterprise customers with the best Linux performance and functionality.

  With SAP certification, SGI Altix enterprise customers can leverage the benefits of the Linux OS-based server line to operate even the most advanced SAP system landscapes.

“SGI is an important hardware partner in the SAP LinuxLab,”  said Helge Deller, development manager at the SAP LinuxLab. “Altix servers provide enterprise customers with powerful computers that ensure the performance, high availability, and stability of SAP systems, while also allowing for dynamic expansion.”

  The result is that MTU has cut operating costs by more than 20 percent compared to its previous architectures, while still fulfilling the high resource requirements demanded by SAP APO. To further leverage the potential of SGI for the SAP solutions,  MTU, which uses Oracle 9i/10i as its database system, now plans to consolidate additional SAP components on SGI Altix platforms.

 

Platform LSF Keeps Priority Jobs On Track MTU Aero Engines’ IT department’s other challenge is to optimize its high server use so that it does not preclude processing priority jobs as needed. It’s a big task when you consider the HPC infrastructure handles an average of 3,000 jobs per day and up to 15,000 during peak periods. The company uses Platform Computing’s Platform LSF, the flagship product in the Platform Accelerate Suite that enables the management and acceleration of batch workload processing for mission-critical compute-intensive applications. The solution intelligently schedules parallel and serial workloads submitted from both commercial and in-house applications like analysis packages, while using available computing resources at maximum capacity and capability. “With Platform LSF, we can achieve high utilization while still getting free slots for those who really need them,” says Axel Philipp,  MTU’s Platform LSF Administrator. Platform LSF’s fairshare scheduling functionality allows Philipp to reserve resources for users who submit high-priority jobs by pre-empting lower-priority jobs, resuming them when compute resources are again available. Platform LSF also allows MTU Aero Engines to defer the acquisition of more hardware until absolutely necessary. Philipp estimates that MTU Aero Engines would need “at least 10 percent more hardware capacity without the scheduling possibilities of Platform LSF.” That translates into some 150 fewer cores, a tangible financial benefit. Joe Cieslak, HPC commercial market manager,  Platform Computing

Meeting the Demands of the Enterprise Market
Originally designed for high-performance computing (HPC) in scientific applications, SGI Altix servers are now increasingly being deployed to meet the needs of enterprise customers such as MTU.

“As we expand internationally, SGI platforms enable us to adapt our IT infrastructures to meet shifting and growing global market requirements,” said Diehl. “Our IT systems have to be able to process millions of transactions every day and have the power and scalability to handle complex monitoring, reporting, and analysis tasks. SGI delivers the performance we need to meet the demands of our worldwide enterprise.”

MTU Opts for Linux
MTU plans to further leverage the benefits of open-source technology in the future. As a result of the successful migration of its technical applications, the company now intends to convert other business IT systems to Linux OS-based hardware platforms, migrating additional SAP components to SGI Altix servers.

“Our first major SAP migration project has shown us that Linux is ready for deployment in the enterprise market,” said Friedemann Ost,  head of IT services and technology at MTU.

“SGI enables us to consolidate and integrate our SAP processes along the entire value chain using a standard platform on Linux,”  added Diehl. “This makes it possible for us to provide high-performance systems in the enterprise environment and consolidate know-how between engineering and enterprise areas. By doing so, we can boost the quality of our system landscape considerably.”

More Info:
Intel Corporation
Santa Clara,CA
intel.com

Linux
Ogdensburg,NY
linux.org

MTU Aero Engines
Munich, Germany
mtu.de/en

Platform Computing
Markham, ONT
platform.com

SAP AG
Waldorf,Germany
sap.com

SGI
Sunnyvale,CA
sgi.com

---

Srinivas Kodiyalam, Ph.D., is an HPC business development manager at SGI. Kodiyalam has his doctorate in mechanical engineering from UC, Santa Barbara, and is an associate fellow of AIAA. To comment on this article, send e-mail to [email protected].