Multiplex communications – Diagnostic testing – Determination of communication parameters
Reexamination Certificate
2000-11-07
2004-10-19
Hsu, Alpus H. (Department: 2665)
Multiplex communications
Diagnostic testing
Determination of communication parameters
C370S395210, C370S395520, C709S232000, C711S216000
Reexamination Certificate
active
06807156
ABSTRACT:
BACKGROUND
The present invention pertains to systems and methods for monitoring and determining the quality of service (QoS) in a network. More particularly, the present invention provides QoS metrics including internal and external packet loss, the detection of stalled periods, and path delay estimates.
Most current network monitoring and analysis methods can be categorized into two groups depending upon where the monitoring is performed. The first category involves monitoring the performance of the IP network on a network level, where an Internet Protocol (IP) is defined to be the method or protocol by which data is sent from one computer to another on the Internet. Network level monitoring is performed by public and enterprise networks. The second category, which involves monitoring the subscriber access performance, is characterized by Service Level Agreement (SLA) monitoring.
Network level monitoring is usually done by the network operator and typically includes simple statistics, e.g., event counters on router interfaces for the amount of incoming and outgoing packets, bytes and number of lost packets. One of the most important aims of network level monitoring is to identify badly performing network elements and network congestion. On the other hand, SLA monitoring is usually performed by the subscriber to test whether the SLA is being kept by the network service provider. SLA monitoring typically involves information about the amount of traffic passing the access link, the Grade of Service (GoS) of the access link, and Quality of Service (QoS) of the access link, e.g., frame errors, bit error rate, downtime. The access link may be thought of as a selectable connection linking a subscriber from one word, picture, or information object to another.
A recent trend among IP service providers is to offer “finer grained” services to subscribers. For example, service providers offer finer grained services having different levels of TCP/IP service. The offered service can be loosely defined, as the case of Differentiated Services Networks (DSN), which provide a protocol for specifying and controlling network traffic by class so that certain types of traffic get precedence. The different levels are differentiated by a combination of access data rate (either guaranteed or average), guaranteed maximum or average packet delay (e.g., less than 100 ms), guaranteed maximum packet loss in the network (e.g., less than 1%). At present, only the so-called “best-effort” service is generally offered, which guarantees none of the above. But if, for example, the provider wants to enable voice or video, (as in UMTS), then there will be a need for these “better than best-effort” services, otherwise the quality would be unacceptable.
As an alternative to DSN, the offered service may be very rigid, such as in networks offering voice over IP (VOIP) or other interactive real-time services in which data delays are not tolerable. Due to developments such as these, the monitoring of subscriber perceived QoS, or user satisfaction, is gaining increasing importance for IP service providers.
Conventional monitoring methods used by network providers are not able to monitor the satisfaction for individual subscribers because traditional methods perform tests on large traffic aggregates which do not allow to estimate QoS for individual applications, e.g., WWW, File Transfer Protocol (FTP), voice over IP, streaming video or audio applications. Hence, it is not possible to accurately estimate the packet delay, delay variation, and loss rate of individual IP telephony conversations based on router interface statistics. On the other hand, different applications require different levels and types of packet service quality. Therefore, it may not always be necessary to monitor an individual subscriber's satisfaction for some applications.
In conventional circuit switched networks a simple network level measurement (e.g., average number of occupied circuits within a circuit group, or Call Blocking Probability) could be used very efficiently to calculate and engineer the GoS for the subscribers in a cost efficient way. In an IP network such analytic methods do not exist. Currently, Internet service providers (ISPs) generally apply a simple engineering rule-of-thumb based on one or more aggregate network level QoS measurements. For example, one rule-of-thumb could be: if the load or packet loss on a given link exceeds a certain level (e.g., 70%) in the busy hour, then the subscriber perceived QoS has probably degraded below the acceptable level, and so the link speed should be increased.
Such a rule-of-thumb approach can work well, and be economic, for large capacity links and in the case of best-effort services. In networks however, where economic considerations limit the possibility of overprovisioning (e.g., IP based mobile access networks), or if higher than best effort services are offered (e.g., voice over IP, DiffServ), it becomes desirable to have a better method for estimating the subscriber perceived QoS.
A number of conventional approaches have been used to obtain coarse estimates of user perceived QoS. Some examples of conventional approaches include NeTrueQOS, Concord, standards and drafts by the IP Performance Monitoring Working Group of the Internet Engineering Task Force (IPPM WG of the IETF), XIWT active network performance measurement architecture, and Ericsson Internet Network Monitor (INM).
A widely applied active method is based on active ping delay measurements. This is done by sending special Internet Control Message Protocol (ICMP) ECHO REQUEST (ping) IP packets to a host. When the host receives the packet, it answers the sender by a response packet within a very short time. By measuring the time it takes to receive the answer, the sending host can estimate the round-trip delay of the path between the two hosts. An advantage of ping is that the implementation of this method is not costly, since ping is available in all IP hosts and routers. Only the monitoring device has to be installed in accordance with the ping method. A related Ericsson product, INM, uses GPS synchronized clocks at network elements. A benefit of INM is that one-way delay can be measured.
Active methods tend to be disadvantageous in that they add significant extra load to the network. The main problem is that active delay measurements require considerable time and resources. In order to have a low variance test, an active delay measurement method would typically send hundreds of test packets. This drawback is exacerbated due to the fact that operators tend to be most interested in delays during the busy hours, when adding considerable extra load should be avoided. During low load periods, the extra loading is not as much of a concern. However, there is little interest in the delay during periods of low load.
Another type of convention approach involves active methods based on user emulation. Such methods uses active tests (e.g., test file downloads between two hosts, as a real user would do) and measures the throughput, loss and delay. This method is advantageous in that it is more efficient to approximate user satisfaction as the method emulates a user and the user's applications. Thus, the QoS of different applications can be more accurately estimated. One example of an active method based on user emulation is Micromuse/Netcool, which can generate active tests for a number of important applications (e.g., Hypertext Transfer Protocol (HTTP), File Transfer Protocol (FTP), Lightweight Directory Access Protocol (LDAP), Remote Authentication Dial-In User Service (RADIUS), etc).
A disadvantage of active methods based on user emulation is that they require even more time compared to Ping. The continuous use of active user emulation would disadvantageously result in considerable additional load to the network. Moreover, the monitored services may not be the same as those service most frequently used by subscribers.
FIG. 1
depicts a conventional system of passive performance monitoring in which packets passing a probe ar
Farago Attila
Veres Andras
Hsu Alpus H.
Nguyen Toan D.
Telefonaktiebolaget LM Ericsson (publ)
LandOfFree
Scalable real-time quality of service monitoring and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Scalable real-time quality of service monitoring and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Scalable real-time quality of service monitoring and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3316143