Multiplex communications – Data flow congestion prevention or control – Flow control of data transmission through a network
Reexamination Certificate
1998-04-07
2002-10-01
Ngo, Ricky (Department: 2664)
Multiplex communications
Data flow congestion prevention or control
Flow control of data transmission through a network
C370S253000, C370S401000, C370S411000, C370S468000
Reexamination Certificate
active
06459682
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention generally relates to architectures for delivering networking products across multiple platforms, and in particular to architectures for delivering Internet Protocol (IP) networking products which are enabled to support service level agreements.
2. Background Description
The operators and users of enterprise networks prefer that their networks be predictable and provide consistent performance. Predictability and consistency are often more important than the raw capabilities of the network, i.e. a network that provides a consistent medium throughput is often considered more desirable than a network which provides very high throughput at some times, but performs poorly at other times. For many business applications, it is important that transactions be completed in a predictable manner while the time taken for the transactions to complete is relatively unimportant (provided it does not exceed a reasonable limit).
Prior art solutions, such as SNA, provide network predictability by preconfiguring the network. This does not work in an IP network, because IP is dynamic and connectionless, and therefore relatively unpredictable. The typical enterprise network environment consists of several campus area networks interconnected by a wide area backbone network. The campus networks usually deploy high-speed links, and perform reasonably well. Congestion tends to occur in the backbone network, which consists of relatively slower speed point-to-point links, and in some of the campus networks which house the servers.
An approach is needed which will provide predictability on an IP backbone network, and do so for backbones with varying degrees of capability. If the network provider can predict the performance of the network, then he can implement service level agreements. A service level agreement is a formal contract entered into by a service provider and its customers. The service provider contracts to transport packets of electronic data between customer premise networks (branch offices, data centers, server farms, etc.) across the provider's backbone network with certain assurances on the quality of the transport. This is known as the Service Level Agreement (SLA). The SLA specifies customer expectations of performance in terms of parameters such as availability (bound on downtime), delay, loss, priority and bandwidth for specific traffic characteristics. An SLA includes acceptable levels of performance, which may be expressed in terms of response time, throughput, availability (such as 95% or 99% or 99.9%), and expected time to repair.
SLAs vary greatly from one network to the next, and from one application to another running on the same network. They are normally based on some level of expected activity. For example, if a large airline wants to ensure that the lines at the ticket counter do not get overly long due to poor response time at the ticketing terminals, some estimate must be made of expected workload, so that the network administrator can be prepared with the necessary resources to meet that workload and still remain compliant with the performance terms of the SLA.
Managing an SLA is an important task because of the revenue implications of failure to support mission-critical business applications. The problem is exacerbated due to diversity of the traffic and due to poor and varying degree of service differentiation mechanisms within the backbone networks. Commercially significant traffic must be prioritized above workloads which do not have a critical time dependency for the success of the business. Many of these workloads in an IP environment are far more volatile than those which have traditionally been encountered in the prior art, e.g. in native SNA environments. In order to meet customer requirements in this environment, a service provider must provide a large excess capacity at correspondingly high charges.
This situation dramatizes the need for effective tools which can monitor the performance of the IP network or system delivering a service over the IP network. While SLA management tools already exist in the native SNA VTAM environment, these tools do not generally exist for IP backbones. Also, there is a need for effective controls which allow the service provider of an IP network to manipulate the priority of the various workloads to be managed.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide support for service-level agreements in a corporate itinerant or an ISP-controlled portion of the Internet.
It is also an object of the invention to provide tools which can monitor the performance of an IP network as measured against multiple SLA agreements.
It is a further object of the invention to provide effective controls which allow the service provider to manipulate the priority of the various workloads subject to SLA agreements.
Another object of the invention is to provide means for achieving network predictability which are adequate to implement a variety of SLA agreements over IP backbone networks halving a variety of capabilities.
It is yet another object of the invention to provide network traffic control tools enabling optimum allocation of network resources and minimizing the need to provide excess capacity in order to implement a variety of SLA agreements.
This invention discloses an architecture (SLA architecture) which organizes the key components, the specific function placements and communication mechanisms so as to enable efficient means of implementing new tools which greatly facilitate both development and enforcement of an SLA. Further, these advantages are even more significant when the backbone network such as current IP-based networks provide very little means for such service differentiation.
The key components of a service level agreement are availability and responsiveness. Availability is maintained by managing network connectivity in the presence of failures, and responsiveness by maintaining a satisfactory Level of network performance. In an IP network, availability is largely taken care of by the adaptive routing mechanism used by IP, but responsiveness needs to be managed. The schemes that make the network predictable provide mechanisms that can estimate the responsiveness of an IP network, and thereby assist in implementing service level agreements. The approach taken in accordance with the present invention to provide predictability in an IP network is to provide a quasi-static configuration which adapts to longer term fluctuations of traffic and relies upon the dynamism of IP to react properly to short term fluctuations and congestion.
Quasi-static adaptations may be viewed as dynamic in relation to longer time scales. By extending the adaptive time scales to relatively gross periods of hours, days and weeks, as appropriate, a quasi-static configuration enables the network to modify allocation of resources in such a manner as to lower the load on the network, in contrast to prior art techniques such as Response Reservation Protocol (RSVP) which allow necessary resources to be requested bit impose a higher signalling load on the network.
The invention involves controlling packet traffic in an IP network of originating, receiving and intermediate nodes to meet performance objectives established by service level agreements. To implement the invention, traffic statistics and performance data such as delay and loss rates relating to traffic flows are collected at intermediate nodes. A central server processes the collected data to determines rates for different priorities of traffic. A static directory node is used to look up inter-node connections and determine initial traffic classes corresponding to those connections. The rates are combined with the initial traffic classes to define codes for encoding the headers of packets to determine their network priority.
REFERENCES:
patent: 5539745 (1996-07-01), Chang et al.
patent: 5631905 (1997-05-01), Yano
patent: 5812527 (1998-09-01), Kline et al.
patent: 5850395 (1998-12-01), H
Ellesson Edward James
Guerin Roch Andre
Kamat Sanjay Damodar
Krishna Arvind
Rajan Rajendran
International Business Machines - Corporation
Ngo Ricky
Schecter Manny W.
Tran Phuc
Whitham Curtis & Christofferson, P.C.
LandOfFree
Architecture for supporting service level agreements in an... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Architecture for supporting service level agreements in an..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Architecture for supporting service level agreements in an... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2970138