Error detection/correction and fault detection/recovery – Data processing system error or fault handling – Reliability and availability
Reexamination Certificate
1998-08-19
2002-06-11
Maung, Zarni (Department: 2154)
Error detection/correction and fault detection/recovery
Data processing system error or fault handling
Reliability and availability
Reexamination Certificate
active
06405327
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention is generally related to general purpose, stored program, digital computers and more particularly relates to an automatic, resource efficient means for monitoring the performance of various portions of a computer system.
2. Description of the Prior Art
The term “performance monitoring” refers to the process of monitoring the performance of various system components within a computer system while the computer system is operating under normal operating conditions. Performance monitoring is a key factor in the operation and maintenance of many of today's complex computer systems.
In the past several decades, the demand on computer systems has steadily increased. Today's software packages require much more processing power and storage capacity than those produced just a few years ago. In addition, many more people are using computers to do tasks that were traditionally done using other means.
Many factors may reduce the optimal performance of a computer system. First, there may be a bottleneck at the input/output (I/O) interface causing the central processing unit (CPU) to be idling a substantial portion of time waiting for data. Even a simple relief on a key constraint in performance can greatly increase the throughput on a system. In the case of I/O, there may not be sufficient channels or disks to provide adequate response time. Second, there may be an insufficient number of processor cycles available to accomplish the program workload awaiting execution. In an interactive or high volume transaction system, such as an airline reservation system or a bank transaction system with thousands of terminals and ATM stations, work is held up at the human interface. This is known as latent workload and can result in unacceptable levels of service. Finally, there may not be enough internal memory within the computer system to store all of the computer programs and data that are to be simultaneously executed and used by the data processing system. This can result in paging. Paging occurs when internal memory limitations require the resulting data from the computer program to be loaded and unloaded from an external storage device each time a process becomes active. Paging data in and out of external storage can greatly increase the time required to complete a given process resulting in unacceptable levels of performance.
The above examples are given only to illustrate the necessity for performance monitoring techniques within a computer system and are not intended as an exhaustive list. It is recognized that many other performance inhibitors exist in modern computer systems and that many of them may be detected by using performance monitoring techniques. However, the basic metrics used in determining performance and levels of service in modern computer systems are input/output service time, processor utilization, and memory utilization.
Performance monitoring of today's computer systems is typically provided by using off the shelf software packages. Examples of such off-the-shelf performance monitoring software packages include the Viewpoint program available from Datametrics, the ALICE module of the SYSTAR products, and Online Activity Monitor (OSAM)/CMF Baseline available from TeamQuest. These software packages are executed on a particular computer or computer network and generate performance data based on a number of preselected factors. One such method is discussed in “Getting Started in 1100/2200 Performance Monitoring”, by George Gray, UNISPHERE Magazine, November 1993. All of the performance monitoring packages listed above use the Software Instrumentation Package (SIP), available from Unisys Corporation, for data collection.
These off the shelf software packages may prove to be useful for some users but they are not an ideal solution for others. Problems that exist with these software packages include: (1) only the performance parameters selected by the software developer are available to the user; (2) the software packages are typically only available for standard computer systems and therefore cannot be used during the development stage of a computer system or on less known computer systems without independent development of the performance monitoring software; (3) the software packages are typically run concurrently with and on the same CPU as the user software and therefore may slow down systems performance (in some cases, as much as 5-10%) while the performance monitoring software is executed; and (4) only hardware that is accessible by the software package, like CPU activity and I/O requests, can be monitored by these software packages.
Problems (1) and (2) listed above may be minimized by having the user write a customized performance monitoring software package for the user's system. However, this requires a significant investment in resources to develop such a program. Problems (3) and (4) listed above cannot typically be eliminated by having the user write a customized software package for several reasons. First, only the nodes within the computer system that are accessible to the performance monitoring software can be monitored. This limitation is a result of having the performance monitoring strategy determined after the computer hardware is designed. Many nodes within a computer system are neither controllable nor observable via software. Second, the performance monitor software is run on the same CPU as the user programs and therefore may decrease overall system performance. Since the performance monitoring software may effect the performance of the system in which the software is attempting to measure, the overall accuracy of the results obtained by the performance monitoring software packages may be limited.
Performance monitoring is often a highly technical process requiring an analyst with many years of experience to examine the performance monitoring results. In a typical scenario, a user first suspects he/she has a problem. The user may then run one or more programs over a period of time to collect utilization statistics, archive and compile the results, and give the data to an analyst, who will do a detailed inspection of the user data, and issue a diagnosis and recommendation. This process can take weeks to months to complete.
Another approach to performance monitoring is to use an external monitor. The external monitor is attached to the system, and often requires a first specialist to attach the device and a second specialist to interpret the results. This approach is both expensive and time consuming.
In many instances, users are unaware that their computer systems are either at or approaching performance limits, and thus performance monitoring is never initiated. When faced with a throughput problem, some customers will simply purchase a new computer system, unaware that the addition of a simple hardware upgrade to their existing computer system would provide better performance at a fraction of the cost of a new computer system.
SUMMARY OF THE INVENTION
The present invention overcomes many of the disadvantages associated with the prior art by providing an automated, real time performance monitoring facility for a computer system which runs periodically as a background process. This invention preferably uses built in performance data collection sites already present in the hardware of the computer system, microcode and/or operating system software. At a user selectable period of time, a sampling of key performance factors is taken from the performance data collection sites. The performance monitor then analyzes the sampled results by comparing the results against two or more performance threshold levels for each performance criteria. In an illustrative embodiment, two performance threshold levels (early warning and actual) are established for the performance criterion of processor utilization. When processor utilization reaches the 90 percent performance threshold, for example, an early warning performance limiter is detected. If the processor utilization reaches
Hansen Lee B.
Kunz Bruce T.
Sipple Ralph E.
Johnson Charles A.
Maung Zarni
Nawrocki, Rooney & Sivertson PA
Starr Mark T.
Unisys Corporation
LandOfFree
Apparatus for and method of automatic monitoring of computer... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus for and method of automatic monitoring of computer..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus for and method of automatic monitoring of computer... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2941246