Root-cause approach to problem diagnosis in data networks

Details Root-cause approach to problem diagnosis in data networks Root-cause approach to problem diagnosis in data networks

2000-12-27

2004-03-02

Le, Dieu-Minh (Department: 2184)

Error detection/correction and fault detection/recovery

Data processing system error or fault handling

Reliability and availability

C714S004110, C709S224000

Reexamination Certificate

active

06701459

ABSTRACT:

TECHNICAL FIELD
The invention relates generally to monitoring and managing complex data network environments, and is particularly suitable for analyzing and diagnosing problems in an E-business system.
BACKGROUND ART
Component integration and other design issues have received a significant amount of attention in network settings, especially those used for electronic business (E-business). In the global communications network referred to as the Internet, portal sites have been created for enabling business-to-business transactions, business-to-consumer transactions, and consumer-to-consumer transactions. Much of the effort has been in the area of integration, so that a single system includes the backend databases used in the ordering procedure, the order fulfillment capability, and the payment processing capability. Another area that has received considerable attention is load balancing to ensure that one component does not act as a “bottleneck” for activity.
FIG. 1
illustrates one possible embodiment of an E-business system. To ensure redundancy, the system uses multiple Internet Service Providers (ISPs)
10
,
12
, and
14
to connect to the Internet. An access router
16
manages the connectivity to the ISPs. At least one load balancer
18
is responsible for receiving user requests via the ISPs and directing the requests to one of the available web servers
20
,
22
and
24
used by the system. The web servers forward the incoming requests to the appropriate E-business applications. The E-business applications execute on middleware platforms commonly referred to as application servers
26
and
28
. A firewall
30
is used to provide security.
The application servers
26
and
28
enable a number of features from which different applications can benefit. These features include optimization of connections to database servers
32
,
34
and
36
, caching of results from database queries, and management of user sessions. Data that is indicative of user information, a catalog of goods, pricing information, and other relevant information for the E-business system is stored in the database servers and is available for access by the application components. To process payments for goods or services by users, the system maintains connections to at least one remote payment system
38
. Links to shipping agencies
40
are also provided, so as to enable the E-business system to forward the goods for shipping as soon as an order is satisfied.
Also shown in
FIG. 1
are a Domain Name Service (DNS) server
42
and a Wireless Application Protocol (WAP) server
44
, and Lightweight Directory Access Protocol (LDAP) server
45
. As is known in the art, the DNS server is accessed to provide users with the Internet Protocol (IP) address. The WAP server may be used for frontending applications accessed via wireless devices such as mobile phones and Personal Digital Assistants (PDAs), while the LDAP server is used for storing and retrieving information in a directory format.
As compared to the emphasis on design issues of the E-business system, monitoring and managing issues for such systems have received significantly less attention. Many systems are managed using ad-hoc methods and conventional server and network monitoring systems, which are not specifically designed for an E-business environment. As a result, the monitoring capabilities are limited.
Since the business applications of a system rely on application servers for their operation, the application servers
26
and
28
are in a strategic position to be able to collect a variety of statistics regarding the health of the E-business system. The application servers can collect and report statistics relating to the system's health. Some of the known application servers also maintain user profiles, so that dynamic content (e.g., advertisements) generated by the system can be tailored to the user's preferences, as determined by past activity. However, to effectively manage the system, monitoring merely at the application servers is not sufficient. All the other components of the system need to be monitored and an integrated view of the system should be available, so that problems encountered while running the system (e.g., a slowdown of a database server or a sudden malfunction of one of the application server processes) can be detected at the outset of the problem. This allows corrective action to be initiated and the system to be brought back to normal operation.
FIG. 2
illustrates monitoring components as used with the E-business system of FIG.
1
. The core components for monitoring include a manager
46
, internal agents
48
,
50
and
52
, and one or more external agents
54
. The manager of the monitoring system is a monitoring server that receives information from the agents. The manager can provide long-term storage for measurement results collected from the agents. Users can access the measurement results via a workstation
56
. For example, the workstation may be used to execute a web-based graphical user interface.
As is known in the art, the agents
48
,
50
,
52
and
54
are typically software components deployed at various points in the E-business system. In
FIG. 2
, the internal agents are contained within each of the web servers
20
,
22
and
24
, the application servers
26
and
28
, and the LDAP server
45
. By running pseudo-periodic tests on the system, the agents collect information about various aspects of the system. The test results are referred to as “measurements.” The measurements may provide information, such as the availability of a web server, the response time experienced by requests to the web server, the utilization of a specific disk partition on the server, and the utilization of the central processing unit of a host. Alternatively, tests can be executed from locations external to the servers and network components. Agents that make such tests are referred to as external agents. The external agent
54
is shown as executing on the same system as the manager
46
. As previously stated, the manager is a special monitoring server that is installed in the system for the purpose of monitoring the system. The external agent
54
on the server can invoke a number of tests. One such test can emulate a user accessing a particular website. Such a test can provide measurements of the availability of the website and the performance (e.g., in terms of response time) experienced by users of the website. Since this test does not rely upon any special instrumentation contained within the element being measured, the test is referred to as a “black-box test.”
Often, it is more efficient to build instrumentation into the E-business elements and services. For example, database servers
32
,
34
and
36
often support Simple Network Management Protocol (SNMP) interfaces, which allow information to be obtained about the availability and usage of the database server. An external agent, such as agent
54
, may execute a test that issues a series of SNMP queries to a particular database server to obtain information about the server's health. Since such a test relies on instrumentation built into the database server, tests of this type are referred to as “white-box tests.”
External agents
54
may not have sufficient capability to completely gauge the health of an E-business system and to diagnose problems when they occur. For example, it may not be possible to measure the central processing unit utilization levels of a web server from an external location. To accommodate such situations, the monitoring system can use the internal agents
48
,
50
and
52
.
In the presently available manager-agent architectures for network monitoring, each measurement is associated with a state. The term “state” is defined herein as being synonymous with “health.” The state of a measurement is computed by comparing the results of the measurement with pre-specified thresholds. When a measurement exceeds its threshold, the state of the measurement is changed to indicate that a problem has occurred and an alarm is generate

Affiliated with

Also associated with

Rating

Root-cause approach to problem diagnosis in data networks does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Root-cause approach to problem diagnosis in data networks, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Root-cause approach to problem diagnosis in data networks will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3265343