Recently, advances in IT have engendered the spread of high-speed, low-cost Internet services. A ubiquitous-computing society has developed in which information can be accessed from anywhere—in companies, by municipalities, and at home. In the ubiquitous-computing society, Web systems are becoming the infrastructure that makes our life more convenient and improves the efficiency of businesses. Examples of the use of Web systems include Internet businesses, such as online shopping and banking, customer-service duties, such as issuance of resident cards by local governments, and employment management and travel expense management by companies.
As a result, if a Web system goes down or suffers from poor response, business and everyday life might be significantly affected. For example, if an online shopping system goes down, not only do orders cease to be accepted, but customers will no longer trust the service. If an internal system goes down, business stops, damaging the operations of the entire company. It is important to provide non-stop, convenient services to prevent a system from going down due to high levels of access or sudden increase in network traffic caused by, for example, attacks. Furthermore, if a Web system fails, the recovery time must be as short as possible to minimize the damage.
uCosminexus Application Server implements Web systems that meet the above requirements.
uCosminexus Application Server uses transaction technology developed in large-scale business systems to ensure stable response even in case of sudden changes in the level of Web server requests or under heavy loads, providing a convenient system.
uCosminexus Application Server ensures stable system operation by using a flow control function that tightly controls the number of concurrent processes in the Web application according to the contents to be processed. This function controls the number of requests by providing a queue for each unit of business logic (URL). This prevents the server from slowing down because of insufficient resources.
See larger picture in a new window
uCosminexus Application Server minimizes the possibility of server stopping, by using Hitachi's own Java VM memory-management method to prevent slowdowns due to a shortage of available memory.
uCosminexus Application Server uses a Java VM memory management function unique to Hitachi to suppress full garbage collection (Full GC), that might interrupt application processing and degrade system response. The Full GC-suppression function moves session information that gradually accumulates in the Java heap during processing to an area not subject to garbage collection. The user can also explicitly specify objects not subject to garbage collection. This improves the sustainability of the operations and brings into being a robust application server that runs 24 hours a day.
#: Java memory-management processing that releases all reusable memory areas in Java VM processes to increase the amount of free memory.
See larger picture in a new window
Reflecting the opinions of engineers struggling with troubleshooting, uCosminexus Application Server provides two functions: the trace function simplifies failure and performance analysis, and the class-based statistics output function allows users to easily determine the causes of memory leaks. These functions bring into being a system that can be restored rapidly after a failure.
The trace function of uCosminexus Application Server visualizes what has happened, where throughout the system, from the Web server to the database. The user can therefore easily determine the location of the failure. The trace function assigns a unique ID to each request from the client, and outputs the request to the log. In addition, the database-connection ID is also output to the log, allowing the user to understand the processing sequence up to and including database access.
Furthermore, the most recent version, version 9, allows tracing of user applications that use frameworks.
See larger picture in a new window
A function unique to Hitachi, that outputs statistics for each class, is provided as an enhanced troubleshooting function for Java VM. By using this function to output memory requirements for each class, the user can easily identify where the problem exists if a failure should occur.
The class-based statistics output function allows investigation of memory leaks in a production environment. Therefore, if a failure occurs, the user can soon start checking the cause of the failure without having to set up a reproduction test environment.
