New Extensible, Parallel Processing Database Technology
Is the Next Generation in Enterprise Data Warehousing
Contact | Introduction | Hitachi Meets the Challenge of Information Management |
THOR SQL objectManager | THOR MPP Data Server
| More Information
FOR IMMEDIATE RELEASE
Contact:
Harris Hall
Hitachi Computer Products (America) Inc.
h_hall@hitachi.com
CHICAGO (September 9, 1996) Hitachi Computer Products (America), Inc., today introduced a new database system technology for the enterprise data warehouse market capable of high-performance, parallel processing of customer data. Hitachi introduced THOR, The Hitachi Object Relational Database System, which brings increased processing power to resource-intensive decision support applications, such as data mining and on-line analytical processing (OLAP).
At the Commercial Parallel Processing (CPP) Conference being held here today through Wednesday, Hitachi unveiled the architecture for the THOR Database System, which combines Massive Parallel Processing (MPP) server hardware with the THOR SQL objectManager, a high-performance object-relational database management system. Together they create a powerful, open database subsystem optimized for processing the complex queries created by decision support applications on very large databases (VLDBs).
Hitachi developed its THOR SQL objectManager software system as a powerful, high-performance object-relational database solution optimized to handle complex, multi-layered queries that typically have high I/O activity, and long searches with varying response times. THOR SQL objectManager also is an open database platform, using standard SQL and supporting gateways such as the Microsoft Open Database Connectivity specification (ODBC) and the Sybase Open Server to connect to popular DSS tools and middleware.
Because THOR is designed and implemented as an object oriented application, it represents the next generation in high performance database technology. THOR is the only truly parallel object-relational system that is intrinsically object oriented. The allows new function such as row-level locking, triggers is supported in parallel and cleanly. While many vendors have grafted on object support such as binary large objects (BLOBs), THOR will allows the user to create and manipulate new user objects like this in a non-disruptive and secure fashion.
Hitachi's THOR MPP Data Server will offer powerful parallel computing hardware with next-generation, object-relational data server software for performance, scalability, and function beyond the capabilities of other SQL-based MPP systems. Hitachi's THOR MPP Data Server hardware combines a unique interconnect architecture with a high-performance modular design of proven components. This means database managers can add data processing capacity incrementally -- scaling to hundreds of processors and terabytes of data -- so the processing capacity can expand to match the workload.
“Modern decision support tools used in data mining and multi-dimensional analysis produce complex queries that are a challenge to optimize and can cause many relational systems to run out of gas. With THOR we designed a next-generation, object-based, open data warehouse system for customers who want high performance processing of complex queries for large databases. Now customers can leverage more of their transaction and production data for competitive advantage,” noted Harris Hall, Product Manager for Hitachi Computer Products (America) Inc.
THOR is optimized for complex queries on large data warehouses. It handles data sizes from a few gigabytes to terabytes. Its inherent DataFlow object-design architecture combined with a high-performance, shared-nothing MPP platform provides reliable, high-performance processing economically. “Hitachi's announcement of THOR clearly indicates the company's intention to be an active participant in the rapidly growing market for data intensive applications such as the data warehouse.”, said Howard Richmond II, Vice President of Gartner Group, a market research company based in Stamford, Conn.
THOR has a scaleable architecture, and delivers server-based data security and integrity that scales with the needs of the user. THOR's SQL objectManager technology supports such features as row-level concurrency control, user-defined data types and business rules, stored procedures and triggers, transaction logging and recovery, database backup and recovery, and comes with a full suite of database administration utilities.
In today's complex computing environments, IT managers must cope with increasing demands for getting information from their data. Three trends are heightening this problem:
To meet the data warehousing challenge posed by these trends, Hitachi has designed THOR as a Massively Parallel Processing (MPP) database system capable of powering an entire enterprise data warehouse within a single physical database. Unlike Symmetrical Multiprocessing (SMP) data warehouse architectures that rely on shared resources, THOR utilizes a “shared-nothing” model made up of stand-alone data processing modules. Where SMP solutions can become overloaded by complex queries on VLDBs (e.g. too much information or too many users), MPP solutions use a shared-nothing model, are more scaleable, and can grow to handle larger amounts of data efficiently.
With MPP, multiple processing elements are configured in parallel so each CPU processes a different part of a complex query or the data. Data is spread across the parallel nodes for simultaneous access. So the heavier the workload and the greater the size of the database, the more computing resources can be applied to increase the capacity and performance; e.g. a query being processed by four nodes in parallel ideally will take half as long to answer as when processed by two nodes and so on. The key to achieving this is a database system that can optimize the query to operate efficiently in parallel and allocate data across the nodes for greater efficiency (i.e. no bottlenecks).
By applying this massively parallel processing approach, Hitachi is able to offer an open, highly scaleable, object-relational data warehouse technology through its THOR Database System. The Hitachi THOR Database System consists of the THOR SQL objectManager, an optimizing RDBMS engine that presents a single logical image of the database, and the THOR MPP Data Server, a massively parallel hardware platform with a shared-nothing architecture.
THOR is the only parallel object relational system in the market today:
The THOR SQL objectManager object-relational database software provides all the features and functions necessary to take full advantage of the THOR MPP Data Server or any other MPP platforms. The THOR SQL objectManager can handle very large databases and complex queries, viewing the entire data warehouse infrastructure as a single data source. The data is distributed across nodes using sophisticated hashing distribution techniques, but the information is accessed as if it were a single database. And thanks to Hitachi's exclusive application of DataFlow techniques on the software architecture, complex data queries are decomposed and optimized for efficient parallel processing. This data model allows more data operations to be performed in parallel.
The THOR SQL objectManager supports standard API interfaces based on the Sybase SQL Server. Because of its object-oriented design that isolates machine dependencies, the SQL objectManager can easily be ported to popular database platforms. The external APIs also are compatible with Sybase Open Server and Sybase Transact-SQL, and compatible with the Open Database Connectivity (ODBC) standard. Hitachi's RDBMS is a plug-and-play solution that can interface with most popular DSS tools and middleware.
An important advantage offered by the SQL objectManager is its ability to handle objects. All data, even traditional relational data, is handled internally as objects. This tightly integrated design ensures that application programs will be able to include and reference objects easily and reliably. THOR is inherently object-oriented and handles all data, even traditional relational data, in the form of objects encapsulating data and its related rules, triggers, and other operations, making it easy to support complex data types.
In Hitachi's THOR MPP Data Server architecture, the basic building block is a processing node (brick): a complete RISC-based computer system equipped with a processor and I/O subsystem that interconnects to other such nodes. THOR's building block structure bundles nodes in a single data processing module. Individual modules can be configured from one to six modules in a single tower, and up to 12 towers or 288 nodes currently can be interconnected for maximum MPP processing power and terabytes of data. There is no architectural limit to the number of nodes that may be coupled together.
Hitachi's THOR MPP Data Server is a “shared-nothing,” massively parallel hardware platform based on powerful industry standard components—such as the PowerPC, PCI bus, and fast/wide SCSI II disk drives—chosen for superior performance, cost-effectiveness, and reliability. The THOR MPP Data Server is a self-contained computer with its own RISC-based CPU, memory, bus, and disk storage. Each node is interconnected with other nodes to form a highly connected toroidal mesh.
The design of the interprocessor communications system (IPC) is that of a torus, a doughnut-shaped surface that enables a near linear scaleable system for communication among points. Each processing brick interconnects with the other bricks in this configuration in a “nearest neighbor” fashion. This connection employs special communication processors to offload the CPUs and is more cost-effective than other MPP systems that rely on multistage or crossbar switches or require the CPU to participate in the data routing. It also provides enhanced system availability through multiple path routing.
The database system and the micro-kernel operating system operate together to distribute user data, manage each brick's devices, process user workloads, and provide communications. This means that all the components in the module operate at near linear scalability as the bricks balance the workload for data processing. To increase the processing power, simply add more modules to the system and the workload capacity processing power will increase proportionally.
Hitachi Computer Products is working closely with partners to refine the THOR Database System technology. For more information, please contact Hitachi by telephone at 1-800-588-THOR (1-800-588-8467), 408-588-3300, or by FAX at 408-988-1279. Information is also available on the World Wide Web at [live URL used to be here]..
Hitachi Computer Products (America), Inc. (HICAM), a subsidiary of Hitachi America, Ltd., develops hardware and software for high performance computing, internet and networking applications.
Hitachi America, Ltd., a wholly owned subsidiary of Hitachi, Ltd., Japan, markets and manufactures a broad range of electronics, computer systems and products, and provides industrial equipment and services throughout the United States.
Hitachi, Ltd. (NYSE:HIT) with headquarters in Tokyo Japan is one of the world's largest electronics companies, with fiscal 1995 consolidated sales (ending March 31, 1996) of 76.6 billion. The company markets and manufactures a wide range of products including computers, semiconductors, consumer products, and industrial equipment.
THOR, THOR MPP Data Server, and THOR SQL objectManager are trademarks of Hitachi Computer Products (America), Inc. All other trademarks are the property of their respective owners.