Through TEPCO’s digitization initiatives, Unicage proved to be crucial in big data processing. Specifically, TEPCO aimed to webitize their operating fee system that supports the calculation of electricity bills through the use of Unicage. With a simple yet efficient architecture, Unicage has allowed TEPCO to process big data much faster than previous implementations, demonstrating the scalable nature of the Unicage system. TEPCO concludes, ‘The simple Unix philosophy behind Unicage has helped us grow and nurture employees with an engineering mindset and an eye for efficiency.’


TEPCO aimed to calculate electricity bills (which are usually collected from every household power meter on a monthly basis) by employees specializing in meter reading. The readings on the power meter is inputted into a handy software terminal available to each employee, which is then automatically saved in the company’s large database. Within the same night, the new inputted data is processed to recalculate and update the electricity bills. However, there needed to be an efficient and speedy system which aggregates inputted data into a structured database that supports employees to quickly compute various calculations, search, and even contact customers. More so, the previous system built solely on Java was too heavy and slow in the load of cooperative processing; as the data being processed grew, the Java program became unfit for the job, showing a lack of flexible scalability. Furthermore, a key aspect of the operating fee system was lost because it became impossible to compile with the terminal identification information as the ‘webitization’ of the system took place. However, it was still necessary to implement the equivalent functionality using the user ID.


Unicage was implemented in two systems for preliminary surveying and trial evaluation. The first was for the conversion process of cooperative files, and the second for the summarization of terminal usage status. The conversion process is a mechanism which converts character codes unique to the mainframe or files which include variable length format from host format to server format. An increase in the cooperation between the mainframe and the server system is expected, and the encountered performance issue when implemented in Java was apparent as the porting of the conversion function server is indispensable for suppressing the load of the mainframe.

Koji Kato: TEPCO 

“Since we were able to determine the specifications of the program during the design phase, we were able to shorten the implementation time of the mounting process thanks to the scalability of UNICAGE.”


Improved performance of batch processing in operational statistical processing: daily processing was reduced from 120 minutes to 10 seconds, and monthly processing was reduced from 240 minutes to 40 seconds

Logic reduced to 29 steps for daily aggregation, and 25 steps for monthly processing


The processing of 7.24million data entries (roughly 2GB) was reduced from 41 minutes to 43 seconds; a promising reduction in processing time of about 56.2 times

Employees with a simpler engineering mindset