Press Release

SGI and Ontoprise Deliver Blazing Performance Results On Business Intelligence Applications Benchmark

SGI Altix and Ontoprise OntoBroker Inference Engine Enable High-Performance Reasoning for Semantic Web Applications

Sunnyvale, Calif. (April 8, 2008) — SUNNYVALE, Calif. (April 8, 2008)—SGI (NASDAQ: SGIC) and Ontoprise GmbH in Karlsruhe, Germany, announced today that an SGI® Altix® 4700 server has achieved blazing results with the OntoBroker® benchmark.

Together, SGI Altix and Ontoprise OntoBroker Inference Engine achieved results that underscore how the uniquely scalable, shared memory SGI Altix platform, combined with the OntoBroker professional inference engine, makes Semantic Web applications a reality. Unlike typical Web applications that require direct and specific instruction from human users, Semantic Web applications involve a vast range of information that is understandable by computers, so they can perform more of the tedious work required to find, share and combine information on the Web.

The OntoBroker benchmark evaluates query performance using intelligent reasoning — an essential requirement for Business Intelligence — and the BEA® JRockit® 5.0 Java Virtual Machine. The test suite comprises nine test scenarios derived from real-world applications involving deep reasoning, knowledge management and intelligent query processing.

The tests compared performance of an in-memory data model (to leverage the shared memory architecture of SGI Altix) with a disk-based model. The benchmark also includes a series of client-server stress tests to demonstrate query-intensive throughput with minimal response time — an important requirement for large-scale, mission-critical applications in a heavily distributed environment like the Web.

The OntoBroker® engine can process queries of varying sizes of ontologies, or representations of related concepts, that can be grouped into a larger category. A major advantage of an "intelligent" business application is the ability to use ontologies to connect and make available datasets from distributed sources to the information consumers in an understandable format.

Altix and OntoBroker Scale with Business Intelligence Load
The tests performed demonstrate that the SGI® Altix® 4700 server and the OntoBroker software offer a commercial environment for intelligent data processing that can scale along with the complexity of the ontology, the size of the ontology, and the number of users.

• Complex graph traversal. The benchmark includes a query to find the possible traversable paths between the nodes in a graph comprising of 1 million nodes. Test results show that a graph traversal with 1 million nodes took only 15.7 seconds using the OntoBroker inference engine on an SGI Altix server. The result proves that queries using complex reasoning can be completed only in a few seconds.
• Lehigh University Benchmark. This test scenario simulates querying university domain ontology of large, complex facts in the order of 106. Results show that the most complex query executed on SGI Altix completed in only 130 milliseconds using 40 threads, while using 212 seconds in a single-threaded mode.
• SemanticMiner®. This knowledge retrieval application test involved semantic retrieval and query processing with 104 facts. Five classes of 195 queries with varying complexity ran in less than 18 milliseconds on the SGI Altix server.
• SmartWeb®. This test executed queries on multimedia Web content. It contained 95 queries and more than 60 rules, with a strong dependency between rules. The queries for multimedia Web content took an average of 17 seconds from a disk-based database but only an average of 73 milliseconds from a non-materialized in-memory model.
• Automotive test. To characterize load and query times with large ontologies, an automotive ontology was tested on the SGI Altix. The in-memory load of 1.4 million facts in file sizes of 41MB took just 133 seconds, and the database load took 381 seconds. The query times showed that large result sets benefit most from an in-memory model.
• Wikipedia® knowledge base search. This test simulated knowledge searches in large documents. On SGI Altix, the most complex query with average result sets from Wikipedia took less than 35 milliseconds with a trade-off of load time of 48 million wiki facts in 138 minutes.
• Client-server test. This scenario of up to 1,000 clients showed a query throughput of up to 102.4 requests per second on SGI Altix. The client-server tests demonstrate that OntoBroker on SGI Altix can handle many concurrent clients permanently sending requests to just one OntoBroker.

"The same uniquely scalable shared memory architecture that enables SGI Altix to power technical and scientific breakthroughs also has been proven to excel at high-performance reasoning for enterprise Business Intelligence applications," said Christian Tanasescu, Sr. Director, Applications Engineering Group, SGI. "Jointly with Ontoprise, we have demonstrated that OntoBroker Inference Engine can load and process large ontologies in memory, with query times generally growing linearly with the number of facts."

"With a professional inference engine like OntoBroker running on the high-performance, scalable SGI Altix platform, executing new Semantic Web applications is now a reality — and proves the readiness of Semantic Web technology for even large-scale and mission-critical applications," said Prof. Dr. Jürgen Angele, CTO and co-founder, Ontoprise. "High-performance reasoning by OntoBroker can also leverage a persistence layer, using an open source database system, without increasing TCO."

The tests were run on an SGI® Altix® 4700 server powered by 40 Intel® Itanium® 2 1.6GHz/18M dual-core processors and 240GB of globally shared memory. The tested system ran Ontoprise® OntoBroker® 5 Inference Engine software, BEA® JRockit® 5.0 Java Virtual Machine (JVM) and Novell SuSE Linux Server 10 SP1 and SGI® ProPack 5 SP2. The tested server was supported by a 2.3TB SGI® InfiniteStorage 4500 system with 15,000 RPM drives and two RAID controllers.

ONTOPRISE | know how to use Know-how™
Ontoprise® GmbH is the leading provider of industry-proven SemanticWeb infrastructure technologies and products used to support dynamic semantic information integration and information management processes at the enterprise level. Ontoprise has developed a mature and industry-proven product suite designed to support the deployment of semantic technologies in the enterprise. The company is headquartered in Karlsruhe, Germany and can be found at www.ontoprise.com.

SGI - Innovation for Results™
SGI (NASDAQ: SGIC) is a leader in high-performance computing. SGI delivers a complete range of high-performance server, visualization and storage solutions along with industry-leading professional services and support that enable its customers to overcome the challenges of complex data-intensive workflows and accelerate breakthrough discoveries, innovation and information transformation. SGI helps customers solve significant challenges whether it's enhancing the quality of life through drug research, designing and manufacturing safer and more efficient cars and airplanes, studying global climate change, providing technologies for homeland security and defense, or helping enterprises manage large data. With offices worldwide, the company is headquartered in Sunnyvale, Calif., and can be found on the Web at sgi.com.

© 2008 SGI. All rights reserved. Silicon Graphics, SGI, Altix, the SGI cube and the SGI logo are registered trademarks, and The Source of Innovation and Discovery is a trademark of Silicon Graphics, Inc., in the United States and/or other countries worldwide. OntoBroker®, SemanticMiner®, and Ontoprise® are registered marks of Ontoprise GmbH. Parts of the technology used in OntoBroker®, and SemanticMiner® are patent pending or patented. Linux is a registered trademark of Linus Torvalds in several countries. Intel and Itanium are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. JRockit is a registered trademark of BEA Systems, Inc. All other trademarks mentioned herein are the property of their respective owners.