N_gine LLC Real-time Parameterized Parallel Processing of
Financial Transactions
6354 Harvard Lane Telephone: (303) 995-7675
N_gine Introduction
Given that global economic environments are now being restructured in terms of new rules and regulations and financial institutions are now being redefined in terms of new roles and responsibilities, the need for new automated financial transaction processing systems has never been greater. Financial transaction processing systems represent the underlying foundation of every commercially-oriented society, enterprise, or person operating domestically or globally.
Fortunately, innovation is now emerging to process all relevant financial reporting upon entry of each financial transaction with dramatic reductions in equipment, personnel, space, and electrical costs. This innovation is based on new multicore computer chips and new software architecture employing the time-honored concepts of parameterization (user-definability) and massively parallel processing (execution of numerous simultaneous processors to perform a particular job).
Parameterization enables users to define their own respective financial transaction processing environments in terms of financial reports, financial transactions, and financial processes and then process all financial transactions through the financial environment. Parallel processing means decomposing each financial transaction into a set of separate and independent processes that may be allocated across any number of separate and independent processors, thereby creating faster response times and assuring concurrency for all financial reporting.
During Q1-2005, Intel, IBM, and AMD announced that future advances in computing speeds would be achieved not by one faster processor per chip but two or more slower processors per chip (now known as “multicore” chips). Chip vendors are now delivering multicore chips (dual-core and quad-core) chips en masse. Their usage in scientific and graphics applications is exploding. However, application to commercial or business applications remains an enigma. There are several reasons why.
First, the simple addition of a second processor to a single chip has rendered legacy serial processing obsolete as there is no logic in today’s software to utilize the additional processor(s). Second, today’s serial processing consists of large amounts of executable code that may exceed the amount of memory available and must be continuously reloaded into memory in segments to complete the processing of a single transaction (“memory thrashing”). Third, business applications require large amounts of data stored on slower disk drives where response times degrade when users access the same information (“disk contention”). Therefore, all executable code must be small in size and all disk operations must be processed in parallel mode to achieve maximum results - unprecedented competitive advantage and operating efficiency.
Recognizing the need for business
solutions, Intel and Microsoft granted $20 million ($10 million each) to the
Software Design Magazine has quoted Mr. David A. Patterson, Professor of Electrical Engineering and Computer Science at UCB as saying, “I wake up every morning shocked by the fact that the hardware industry expects us to solve one of the most difficult problems Computer Science has ever faced – how to write and process parallel programs correctly. We do not know what we are to teach our students. I ask for help from the private sector.”
Computer scientists recognize that the processing functionality must be decomposed into a set of small, separate, and independent processes that may be allocated and executed across any number of simultaneous processors. Decomposition to the lowest possible atomic level means that more processors may be assigned to completing the job and all processes must be of same processing duration to achieve optimal benefits of parallel processing (“load balancing”). Thus, the success of parallel processing depends upon the knowledge, experience, and skill of the person(s) performing the decomposition. Such persons are in great demand, if indeed they can be found. Such solutions encourage economic growth NOW – not 5 years from now.
N_gine LLC (pronounced “engine”) has discovered and patented a new method (or software architecture) for processing large volumes of daily financial transactions that customizes financial reporting to specific need, updates all relevant financial reporting in real-time, integrates disparate financial applications into fewer systems, extends solution life cycles, produces 5-10x productivity increased, and enables parallel processing.
N_gine’s initial tests proved 28,000 financial transactions in less than five minutes and 280,000 financial transactions in less than fifty minutes on a quad-processor machine costing less than $10,000. Each financial transaction updates Balance Sheet (including investment portfolios), Income Statement, Cash Flow, Capital Gains, Performance Measurement, Pending Income/Expense, Pending Corporate Actions, General Ledger, and Transaction Journal reports. N_gine’s present tests prove 44,000 financial transactions is less than five minutes, and future tests on blade servers are expected to increase productivity to well over 100,000 financial transactions in the same period. Each financial transaction performs 20-40 disk updates, which is several times the content of traditional serial financial transaction processing systems.
N_gine’s microkernel consisting of operating system, relational database, and N_gine per se fits into 2GB of memory. All business processes are separate and independent of all other financial processes, and all business processes approximate the same processing durations. Thus, memory thrashing, disk contention, and load balancing are achieved, eliminating many overnight processing jobs. Parameterization enables N_gine to transcend traditional vertical markets.
The result is totally new real-time financial services where users pay on a per-usage or “Software as a Service (SaaS)” basis. N_gine is ideally suited for hosted application machines attached to global networks. N_gine may be easily ported from quad-processors to blade servers to mid-range machines to high performance computing machines, in either Oracle 11g or IBM DB2.
The transition from serial to parallel processing will be pervasive and permanent. N_gine extends an invitation to attend a free conceptual presentation and live demonstration of imminent solutions, estimated costs, and returns on investment. All questions are invited.
Contact: Bill Hinkle, Founder & Manager, N_gine LLC; Telephone: (303) 995-7675