Patent Application
20060149845
This invention pertains generally to the field of computer networks and more particularly to the area of requesting and managing high-quality communications for applications over shared networks.
Over the past several years, an increasing number of computer users in the United States have subscribed to high-speed (“Broadband”) Internet. As a result, network providers of these Broadband services are beginning to deploy advanced Internet services such as Voice over Internet Protocol (VoIP), Internet-based video-on-demand, on-line computer games, and business services. Because of the network demands from these services, there is a recognized potential for congestion resulting from oversubscription, thereby leading to churn and lost revenues.
This problem can be alleviated by managing the Internet traffic so that each subscriber obtains the quality of service (QoS) necessary to ensure these new services perform well. The Broadband cable industry has recognized the importance of maintaining subscriber satisfaction and, via its standards body, CableLabs, has specified a policy-based technology platform for guaranteeing QoS over the hybrid fiber-coax (HFC) network. This specification, called PacketCable Multimedia (PCMM), is intended to empower service providers to differentiate data flow to individual subscribers, thereby enabling a whole new class of “network aware” services.
The recent PCMM specification opens the door for multiple system operators (MSOs) such as cable companies to increase the overall value of their high-speed cable networks. Subscribers are now able to enjoy richer multimedia content in the home or office and benefit from packet-switched technologies such as VoIP and video telephone. By differentiating data flow to these subscribers on-demand, service providers can potentially maximize revenue from the content riding on their networks. PCMM further enables service providers to tap into the market for small and medium business telephone and data communication services. Until recently, this market was served only by dedicated lines capable of offering the service guarantees that can now be offered by Broadband cable.
This technology can be best exploited by making applications “network aware,” meaning that individual services and applications can dynamically signal their QoS requirements to the cable modem termination system (CMTS). Historically, there have been only two major approaches by which applications were made “network aware,” namely integrating “network awareness” into application software, and deep packet inspection hardware. The network infrastructure of broadband cable has not been capable of discriminating data flows based upon each application or content's QoS requirements, thus preventing applications from becoming truly “network aware”. Further, no dynamic processes have existed for managing QoS, thus network resources could not be re-allocated when the data flow requirements were no longer required by an application, and therefore the value of the network's data capacity was not maximized. Previous software vendors have tried unsuccessfully to capture policy-based QoS into their applications by embedding network traffic management; however these efforts failed due to a lack of support by the network infrastructure.
One unsuccessful method is an integrated application-oriented approach. This method requires every application developer to make their software “network aware” by including QoS features within their application software either on the subscriber's computer or the Web content server, as described, for example, in “Microsoft 2000 Server: The Microsoft QoS Components”, by Microsoft Corp. (November 1999). Thus, the traffic management software vendor and the MSO need to partner with numerous application developers and Web content providers. Web content providers or subscribers must upgrade the application software on their servers or computers, respectively, to enable “network awareness.” It is no surprise, therefore, that application developers have resisted integrating QoS into their applications. The wide range of applications and fragmentation in several segments of the software market (e.g., computer games) inhibit the deployment of a comprehensive service offering. Additionally, two or more similar applications in the same home or office LAN cannot be reliably identified separately, and thus not enough data flow is supplied to satisfy each user.
Another unsuccessful method uses deep packet inspection hardware to inspect every one of the billions of Internet packets traveling past it for a source and destination IP address, port number, and application type, such as that described by Narad, et al. in U.S. Pat. No. 6,157,955. The packet inspection hardware is generally located regionally at the MSO. In order to determine the application type, and consequently its QoS requirements, the circuits needs to evaluate an entire stream of data between each subscriber and the destination Web content provider. The major advantage of deep packet inspection is that it manages network traffic automatically, and with maximum transparency to both subscribers and content providers. Unfortunately, the packet inspector is highly intrusive in the network and sits directly in the data path making it a possible single point of failure. The unit must be deployed regionally and is subject to local power and space constraints. Hardware upgrades may be difficult and costly. Some applications may be difficult to decipher and the computation requirements may exceed currently available integrated circuit technology. Since the packet inspector must look at every packet as it traverses a decision tree, it is less efficient than other software solutions located closer to the user.
Other previously existing methods, such as those described by Jackowski, et al. in U.S. Pat. No. 6,141,686, merely serve to collect and aggregate application traffic data for retrieval and QoS management by a central policy server, but do not include mechanisms by which user-specific, application-specific customized QoS profiles can be stored and updated on a client machine. Thus, heavy loads are placed on the central policy server, which must process all QoS requests for all client machines, regardless of whether those QoS requests are legitimate. Further, such other previously existing methods are concerned with bandwidth, but not other quality metrics such as jitter or latency. Scheduling QoS for particular applications and users can therefore be problematic with such other methods.
In an embodiment of the invention, methods and systems are provided that embed “network awareness” into a system QoS proxy on the subscriber's computing device which dynamically signals the quality of service (including bandwidth, latency and jitter) necessary to ensure that networked applications run well over a shared network, such as a hybrid fiber-coax (HFC) network operated by a cable company. A solution can be rapidly deployed for almost any application or service, and at a lower cost than comparable approaches. It is versatile enough to manage the traffic on almost any network and for any application, since it embeds the core traffic management close to the user and computing device on which the applications are running. This more accurately relays the data flows necessary for each application, and also reduces the computing burden on the central office. Application-specific data flows are restricted exclusively to the application and its associated computing device. For example, one user on the home computing device network can participate in a managed, high-quality videoconference while another can transfer a music file using standard-quality “best-effort”. The solution can be extended to the home in support of CableLabs' CableHome 1.1 Specification CH-SP-CH1.1-I06-041216, December 2004, which is hereby incorporated by reference for all that it teaches without exclusion of any part thereof. The entire process is achieved with relative transparency to the user, so that the traffic management occurs automatically without the subscriber's interaction. The subscriber's only real awareness of this technology may be when the system QoS proxy is upgraded or the premium service tier is billed. Transparency is a benefit because it makes the system easy to use, and it forces applications to use the premium service.
In an embodiment of the invention, a method for establishing a high-quality network connection communications session between a software application running on a subscriber computer and a network service provider is provided, the method comprising capturing a network call from the software application running on the subscriber computer, identifying the software application associated with making the network call, verifying that the software application is authorized to request a high-quality network connection, obtaining a network quality profile corresponding to the software application, and causing to be transmitted to the network service provider a request for a high-quality network connection communications session according to the quality profile, whereby, after the network service provider has processed the request, communications between the software application and the network service provider are of a quality satisfying the requirements of the quality profile.
An embodiment of the invention provides a computer-readable medium including computer-executable instructions facilitating establishing a high-quality network connection communications session between a software application running on a subscriber computer and a network service provider, the computer-executable instructions performing the steps of capturing a network call from the software application running on the subscriber computer, identifying the software application associated with making the network call, verifying that the software application is authorized to request a high-quality network connection, obtaining a network quality profile corresponding to the software application, and causing to be transmitted to the network service provider a request for a high-quality network connection communications session according to the quality profile, whereby, after the network service provider has processed the request, communications between the software application and the network service provider are of a quality satisfying the requirements of the quality profile.
In another embodiment of the invention, a method is provided for establishing a high-quality network connection communications session between a software application running on a subscriber computer and a network service provider, the method comprising, receiving a request for a high-quality network connection on behalf of the subscriber computer, the request caused by a system quality proxy agent residing on either the subscriber computer or a content server contacted by the subscriber computer, the proxy agent having obtained a quality profile for the software application, and the request comprising information about the software application, authenticating the request, obtaining policy information corresponding to the request, and granting a high-quality network connection communications session to the subscriber computer for communications with the application according to the policy information.
Another embodiment of the invention provides a computer-readable medium including computer-executable instructions facilitating establishing a high-quality network connection communications session between a software application running on a subscriber computer and a network service provider, the computer-executable instructions performing the steps of receiving a request for a high-quality network connection on behalf of the subscriber computer, the request caused by a system quality agent residing on either the subscriber computer or a content server contacted by the subscriber computer, the proxy agent having obtained a quality profile for the software application, and the request comprising information about the software application, authenticating the request, obtaining policy information corresponding to the request, and granting a high-quality network connection communications session to the subscriber computer for communications with the application according to the policy information.
In still another embodiment of the invention, a system is provided for managing the quality of a network connection communications session provided to an application running on the computer of a subscriber of a network service provider, the system comprising a database containing information about the subscriber of the network service provider, a policy server storing network quality configuration settings for a variety of conditions, and an application manager, the application manager, in response to a request from a quality proxy agent running on the subscriber's computer, reading from the database and instructing the policy server to establish a network connection communications session with the application running on the subscriber computer at a quality according to an appropriate configuration setting, wherein the request from the quality proxy agent is made according to a quality profile for the application, the profile obtained by the quality proxy agent for the application.
In yet another embodiment of the invention, a method is provided for increasing the revenue of a network service provider, the network service provider providing network service to a plurality of customers, the method comprising, receiving from a requesting entity, on behalf of a computing device associated with a customer in the plurality of customers, a request for a high-quality network connection communications session for an application running on the computing device, wherein the request is in accordance with a quality profile for the application, the profile obtained by the requesting entity, causing network communications for the application to be of a higher quality than ordinary network communications, and billing the customer a fee for services including high-quality network connection communications sessions.
While the appended claims set forth the features of the present invention with particularity, the invention and its advantages are best understood from the following detailed description taken in conjunction with the accompanying drawings, of which:
The problem of managing quality of service (QoS) in shared networks is a growing problem facing the broadband industry. For example, in existing cable networks, there is a bottleneck for providing necessary QoS for VoIP in the upstream direction. Cable networks use a time-division-multiplexing (TDM) based protocol to assign transmission opportunities (known as mini-slots in cable modem terminology) to the cable modems. To ensure that QoS (jitter, latency, and bandwidth) meets the VoIP requirements for the duration of the call, the cable modem termination system (CMTS) (e.g., centrally located cable router) reserves the resources (mini-slots in the upstream and bandwidth in the downstream) for the call when it receives a QoS request from a session initiated protocol (SIP)-based softswitch (packet switching platform). When the call is finished it releases the resources.
Cable networks are usually engineered for 2000 users to share a ˜36 Mbps downstream channel and for 500 users to be sharing a ˜6-10 Mbps upstream channel. In cable networks there are usually 4-6 upstream channels per downstream channel.
The first problem in deploying QoS for VoIP is managing the QoS. The industry's recent standard, PacketCable Multimedia (PCMM), specifies the protocol for requesting and granting the QoS but does not specify how to manage the QoS. The PCMM standard is defined in CableLabs' “PacketCable Multimedia Specification PKT-SP-MM-102-040930”, September 2004, and “PacketCable Multimedia Architecture Framework Technical Report PKT-TR-MM-ARCH-V01-030627”, June 2003, which are hereby incorporated by reference for all that they teach without exclusion of any part thereof. Managing QoS requires more than just granting QoS because the QoS in the network is a finite resource. As QoS is granted for VoIP services, the best-effort data services will be affected. Therefore QoS management systems take this into account when making a decision on whether to grant the request or not. This is commonly referred to as “admission control.”
The rules for when to grant QoS are binary when there is only one QoS service is at issue: either there is QoS (e.g., for VoIP), or there is “best-effort” data (no QoS). In the multiple QoS service scenario, the decision becomes how to best divide up the network resources available for QoS-based services. Each QoS service generally has its own unique QoS requirements (bandwidth, jitter, and latency) and value to the MSO. The value to the MSO is a function of the revenue stream less the costs to provide the service. The cost of the QoS is a function of the bandwidth, jitter, and latency required in each direction. The revenue stream is a function of the premium the MSO can charge for the service and the customer satisfaction.
Embodiments of the present invention effectively manage the QoS in a shared network on a per-application, per-user basis. This allows an MSO to apply business rules that take into account the value and cost of the QoS for each application and the subscriber requesting to use the service.
An exemplary architecture for managing communications quality on a per user, per application basis is shown in
Once a request is received and processed by the MSO server 118, a high-quality communications session is established between the CMTS 114 and the application running on the subscriber's computing device. In the example above, application 121 is given a high-quality communications session with the MSO 112, so communications between application 121 and the MSO's 112 CMTS 114 are given higher priority (i.e., to ensure that bandwidth, latency and jitter requirements are satisfied) than communications from other end users 106, 108, 110 sharing the shared network 102. Additionally, application 121 is given higher priority than other computing devices 130 on the same local network as the application's 121 computing device 122 (e.g., sharing the same cable modem/router 123). Application 121 is even given higher priority than other applications 132 running on the same computing device 122.
Requests for high-quality communications can be made in several ways. One technique, as used in an embodiment of the invention, includes a system QoS proxy 134 running preferably as software on the subscriber's computing device 122. The QoS proxy 134, previously installed on the subscriber's computing device 122, registers with the application manager 120 upon startup. During this registration, application-specific QoS profiles 136 are authorized and/or modified and/or downloaded to the computing device 122 and preferably stored locally on the computing device 122, for example, on a hard drive or in a local system memory store within or attached to the computing device 122. In some embodiments of the invention the QoS profiles 136 are further updated at periodic intervals by the application manager 120. For example, the MSO 112 can offer a premium “gaming tier” subscription to allow high-quality communications sessions when subscribers are playing popular games; the MSO 112 can send game-specific QoS profile updates to the computing device 122 to be stored with the other QoS profiles 136. The QoS proxy 134 intercepts network calls from applications running on the computing device 122 and classifies information about the source, destination, port number and application making the network call. The QoS proxy 134 determines whether and what level of QoS is required for the calling applications or services by comparing the classified information to the quality profiles 136, and then sends a request to the application manager 120 at the MSO 112. The application manager 120 authenticates the subscriber, the application, and the appropriate tier of service. The request is authorized based upon business policies established by the MSO 112. If authorized, a message is sent to the policy server 126 for distribution to the CMTS 114.
As an example of the use of such a system QoS proxy, consider a subscriber running an online game that requires low latency. If the subscriber has subscribed to a premium ‘gaming tier’ service of the MSO, then a high-quality (in this case, low-latency) communications session is made available for the duration of the game.
An additional technique, as used in an embodiment of the invention, is shown in
Turning attention to
The QoS agent 302 typically runs as a background process and matches the calling application against a list of authorized applications and QoS profiles 314. In an embodiment of the invention, the QoS agent 302 looks at process information associated with the application making the API call. From the process information, the QoS agent 302 obtains the command line string that was used to launch the application. In addition, the QoS agent 302 can obtain meta-information associated with the application. The QoS agent 302 uses the command line or meta information and does a pattern match (for example, using regular expressions) to see if the application is known by the QoS agent 302. If so, the QoS agent 302 checks if there is a QoS policy for the application. The list of authorized applications and QoS profiles 314 are preferably stored locally on the subscriber's computing device 301. If the calling application is authorized locally, the QoS agent 302 generates a message 316 and sends it to a remote application manager 317 for subscriber authentication and authorization and establishment of a high-quality communications session according to the quality profile 314. In this manner, the QoS agent 302 automatically signals the precise QoS requirements to the remote application manager 317 for admission control. The message 316 includes the data flow parameters specified by the PCMM specification. In one embodiment, the message 316 is a SIP/DIP message. Alternatively, the message 316 is an XML message over the SOAP protocol. In some embodiments, the message 316 includes instructions to “color” packet bits to mark a change in packet priorities for networks such as private SONET networks and Local Area Networks (LANs) through the use of TypeOfService/DiffServ (TOS/DS). The QoS agent 302 also preferably makes a call to a local QoS traffic control API to instruct the TCP/IP stack to color bits for this flow. When QoS is no longer required, the premium service is terminated. The QoS shim 303 either intercepts an API call by the application to close the connection, or is notified through a call-back from the operating system that the application has terminated. The QoS agent 302 sends a request to the application manager 317 to terminate the high-quality communications session, and communications resume at their default levels.
In this manner, the QoS agent 302 and shim 303 can rapidly manage the traffic requirements for virtually any application on any network without involving the application developer, and without deploying hardware. Some applications are thus made “network aware” that cannot be enabled by other methods since these applications either lack the inherent QoS signaling capability, or their application signatures cannot be easily inspected. For example, the QoS agent 302 can uniquely enable an application or family of applications for transfers for computer data or digital photo back-ups. The QoS agent can extend PCMM by enabling a virtual private network (VPN), which re-creates being on the corporate environment but in a telecommuter's home. Since the core intelligence of the system sits near the application running on the user's computing device 301, the system is capable of efficiently providing an optimal amount of data flow to the application and computing device 301.
In some embodiments of the invention, the system QoS agent 302 and shim 303 are automatically installed by or for the subscriber, for example, as part of an initial setup with the subscriber's network service provider. Alternatively, the subscriber installs the QoS agent 302 and shim 303. The network service provider can further update the quality profiles 314 on the subscriber's computing device 301 as necessary, according to, for example, applications included in subscription tiers subscribed to by the subscriber, or new, recently-subscribed-to tiers. The subscriber's quality profiles 314 are kept current through the use of a state update messaging protocol between the QoS agent 302 and the application manager 317.
In an alternative embodiment, the QoS shim 303 sits below the transport layer 312 as a driver, and supplies a new TCP/IP stack for network communications. Such an embodiment is useful if the presence of anti-spyware software is present on the subscriber's computing device 301.
As noted above, the subscriber's computing device 301 can run any of a number of operating systems for which a system QoS agent 302 can be used to automatically request high-quality network communications. In the Microsoft Windows operating system, as illustrated above, the QoS shim 303 is a Layered Socket Provider (LSP) shim sitting between the Winsock 2 API and the transport layer. The LSP shim is a custom provider that is registered with Winsock, so that all application socket calls are dispatched to the custom provider. In addition to intercepting network service calls and creating QoS requests for the central application manager, it can also call the Windows traffic controller to set the DiffServ bits. The traffic controller is very flexible and can set the TOS bits or various configurations
In the Linux operating system, system calls go through an array, sys_call_table[] which directs network requests. The array stores the pointers which direct the calls. The QoS agent 302 is invoked by a pointer re-directing the applicable OS call to the QoS agent 302, which makes the QoS request to the application manager. In more detail, to intercept a system call, original pointers are overwritten with pointers to new functions. The original pointers are saved, and later written back at cleanup. Linux architecture provides means via modules (dynamically loadable/unloadable components) to extend the kernel functionality. The Linux module can be dynamically loaded at any time, or may included in a configuration file, which contains a list of modules to load when system boots. Linux further has the native capability to mark/color the DS/TOS bits.
In a similar manner to the Linux case, the QoS agent 302 can be made to operate in a computing device running other variations of the Unix operating system, including BSD Unix, upon which the Macintosh OS X operating system is based. For the Sun Solaris operating system, all processes make system calls to the OS. The QoS agent 302 sits at the /proc layer, which monitors and intercepts the applicable networking calls and makes the QoS request to the application manager. By monitoring /proc/pid, once can attach to a particular process, and specified system calls can be captured at entry and exit. The invention is not limited to the Microsoft Windows, Linux and Unix operating systems, however; the invention can generally be embodied with any computing device operating system that allows the intercepting and monitoring of network calls.
In
Turning attention to
A method for using a content server-side QoS proxy to request high-quality communications on behalf of a subscriber's computing device is shown in
Turning to
In some embodiments, the server 706 hosting the AppMgr 702 is the same as the policy server 708, and are not separate servers. In such embodiments, the AppMgr 702 can perform policy functions typically performed by a policy server, so that no policy server 708 need be present. The AppMgr 702 communicates directly with the CMTS 710 to manage resource availability and/or make admission decisions. Incorporating the functionality of the policy server 708 into the AppMgr 702 is particularly beneficial for a small MSO 704 employing one or a handful of CMTSes, since the MSO 704 is alleviated of the need for specialized policy server. In some embodiments, a discovery mechanism is included in the AppMgr 702 to determine the appropriate CMTS for a subscriber.
In accordance with the PCMM specification, some embodiments of the invention use a Record Keeping Server (RKS) 714 as the interface and mediator between network activity and the MSO's 704 billing system 715. The RKS 714 creates billing records that it forwards to the billing system 715. A billing record is generated whenever the RKS 714 can match two PacketCable Event records from the policy server 708 and the CMTS 710. Whenever the policy server 708 receives a “Gate-Set-Ack” it generates an event message to the RKS 714. Likewise, the CMTS 710 generates a corresponding event message to the RKS 714 when it actually creates the gate. When the RKS 714 gets both event messages it accurately generates a billing event. The same sequence of events occurs when the gate is deleted. The MSO 704 can bill subscribers according to a number of criteria and subscription plans. For example, a subscriber can be billed per transaction, time reserved, time of day, day of the week, quality of the connection reserved (amount of bandwidth, restrictions on latency and jitter), amount of bits transferred, and other criteria. The MSO 704 alternatively or in addition can charge for an unlimited use of a high-quality connection based on combinations of these criteria. Numerous other possible billing alternatives are available, any number of which can be offered by the MSO 704.
A general method for use by an application manager to manage QoS, as used in an embodiment of the invention, is shown in
Returning to
By way of background, DOCSIS permits 65535 gates per CMTS 710 where a gate is logical entity representing a unidirectional data flow with QoS. For example, a voice call requires two gates: one upstream and one downstream. A video call requires 4 gates: two upstream (audio+video) and two downstream (audio+video). A cable modem 716 can theoretically support up to 2ˆ14 gates in the upstream direction. In practice a cable modem 716 supports approximately 16 gates in the upstream (65,535 gates/2 (upstream+downstream)/2000 modems per CMTS). In actuality, the number is even less due to the fact that a large number current cable modems use a version of a Broadcom DOCSIS chipset that only supports four gates. One service flow must be used as the primary service flow (default) and thus there are 3 that are available to be used by the QoS management system. This limitation of three upstream gates limits the number of simultaneous VoIP phone calls that can handled by a single cable modem to three, or to one voice call and one video call.
The AppMgr 702 therefore detects (dynamically or otherwise) and takes into account the number of gates a cable modem can support and minimizes the number of gates required to support QoS services. In one embodiment, the AppMgr 702 attempts to dynamically combine similar gate requests from the same cable modem 716. Additionally, the AppMgr 702 uses logic in requesting QoS, and is cognizant of the underlying QoS mechanisms the CMTS 710 employs in honoring the QoS request.
By way of more background, QoS for DOCSIS networks is described more fully by Sunkad in “Quality-of-Service: A DOCSIS PacketCable Perspective—Part II”, (http://www.cablelabs.com/news/newsletter/SPECS/MayJune2000/news.pgs/story5.html) which is hereby incorporated by reference for all that it teaches without exclusion of any part thereof. QoS is provided at layer 2 or the media-access-control (MAC). The MAC protocols in DOCSIS in the upstream and downstream are different, but in both cases, the DOCSIS network is a shared media. In the downstream, MAC appears very much like Ethernet since it is a one-to-many (CMTS to cable modems). In the upstream, DOCSIS uses a time-division-multiplexing scheme to assign transmission mini-slots. The assignment of the mini-slots is dependent upon the scheduling routine for the service type. DOCSIS uses four types of scheduling routines: best-effort, unsolicited grant, real-time polling, and non-real-time polling.
For best-effort type services, the cable modem 716 sends a request to the CMTS 710 for a future transmission opportunity. This request is sent in what is referred to as the “contention” region of a TDM frame. Transmissions in the contention region are susceptible to transmission collisions from other cable modems making similar requests. When the CMTS 710 receives the request, it acknowledges it to the modem 716 and schedules the request at some point in the future. Best-effort type transmissions on busy networks are thus susceptible to wide variances in the time between scheduled transmission requests due to having to contend with other modems for these opportunities. As a result, the inter-packet latency varies, which is referred to as jitter.
In contrast to best-effort, Unsolicited Grant Service (UGS) addresses the problems associated with best-effort for services that have tight tolerances on jitter and latency such as VoIP. UGS works by reserving a set of mini-slots in advance for the cable modem 716. The CMTS 710 then grants transmission opportunities to the cable modem 716 on a regular basis whether the cable modem has something to send or not. UGS provides a constant bit rate service (CBR).
In contrast to best-effort and UGS, real-time polling and non-real-time scheduling use a polling scheme. When a cable modem 716 requests a non or real-time polling service, the modem 716 includes a polling interval. The CMTS 710 then polls the modem 716 at this regular interval asking if it has anything to send. If the modem 716 has something to send, then the CMTS 710 schedules a transmission opportunity for the modem 716 in the near future. Polled scheduling is a compromise between a UGS and best-effort. Polling eliminates the possibility of collisions between modems requesting transmission opportunities while at the same time not wasting bandwidth (mini-slots) when modems do not have anything to send.
QoS in DOCSIS networks is managed at the MAC layer by the CMTS 710. When a QoS request is accepted by the CMTS 710 it sets a gate (i.e., a logical entity within the CMTS 710). Every gate has associated with it a set of QoS parameters (bandwidth, jitter, latency, packet classifier, and service type) that define the service flow. For downstream data the traffic, classification and policing is done on the CMTS 710 before the traffic is sent across the shared network 712 to the modem 716. The CMTS 710 classifies the data and puts the data on the respective service flow's (SFID) queue, and then uses a queuing algorithm such as weighted-fair-queuing (WFQ), classless-fair-queuing, or priority-based-queuing.
For the upstream, the CMTS 710 issues a management message to the cable modem 716 instructing it to create a SFID classifier and queue for the packets to correspond to the gate. Additionally, the CMTS 710 inserts the SFID into its upstream mini-slot scheduler to schedule transmission opportunities for the modem 716. Thus for the upstream the classification and policing is done at the cable modem 716 and the scheduling (i.e., traffic shaping) is done by the CMTS 710.
In some embodiments of the invention, the AppMgr 702 optimizes the use of gates by the cable modem 716 and CMTS 710. For example, in one exemplary case a gate can be created for downstream and one for upstream, where the upstream gate is configured to use real-time polling scheduling and the downstream gate is configured with a guaranteed minimum bandwidth in order to improve response time for an interactive session. Sufficient bandwidth is allocated to ensure there is no congestion between the two endpoints. In a second exemplary case, a bi-directional email session can be established with a gate for each direction. If the two cases are simultaneously active, the AppMgr 702 modifies an existing gate to be the union of the two service flows. This is achieved, for example, by increasing the bandwidth in each direction to represent the sum of the bandwidth for both services flows, while modifying the latency to meet the smaller of the two.
In a third exemplary case, a peer-to-peer session, such as a two-player game or videoconference, can be established with two gates for each peer's cable modem 716. In such a case, the AppMgr 702 preferably matches the gate requirements for the upstream and downstream flows in each direction. For example, if the upstream QoS is:
In accordance with the PCMM specification, each gate has associated with it one or more packet classifiers. An eight-tuple classifier, as used in some embodiments of the invention, contains the following fields: Protocol, Source IP address, Source port, Destination IP address, Destination port, Priority, and DSCP/TOS mask. Protocol is an IP protocol field (RFC 1700) value (IP, ICMP, etc.), where a value of 256 matches any IP protocol, and 257 matches both TCP and UDP. The source IP and destination IP addresses are the addresses seen by the CMTS as the respective originator and termination point of the IP flow. The source and destination ports are UDP or TCP ports. Priority indicates a search order for which to apply the classifiers in case of a classifier overlap (i.e., highest priority is applied first). Classifiers may include wild-card fields (indicated by zero fields). Downstream classification is performed by the CMTS 710, while upstream classification is performed by the cable modem 716. If a packet does not match any classifier, then the packet is forwarded to the primary or default service flow, which is normally best-effort.
In some embodiments of the invention, the AppMgr 702 uses classifiers to manage QoS for subscribers and applications. Classifiers are associated with service-spec objects of subscriber policies stored in the subscriber database 712. A service-spec is a three-tuple binding of: application group, traffic profile (i.e., QoS parameters for the flow), and packet classifiers. By using service-specs, broadband service is defined or created by associating a group of applications that have similar QoS requirements to a common end-point for the traffic flow. Thus, when the system QoS proxy is installed on a subscriber computing device, the classifier maps to one or more host or network destinations that are within the MSO's 704 managed network. For the case when the system QoS proxy is installed on a web content server, the classifier is used to determine if the connecting client is from the MSO's 704 network. An example set of classifiers is given in Table 1, corresponding to the examples illustrated in
In
In the examples of
((packet protocol==classifier protocol)|(classifier protocol==0))
AND ((packet source IP==classifier source IP)|(classifier source IP==0))
AND (packet source port==classifier source port)|(classifier source port==0)).
As can be seen in Table 1, the potential for overlapping packet classifier exists for the downstream cases of case 901 and 902. If a classifier of (source=TCP//192.168.1.0:0 and destination=TCP//10.2.2.2:0) was used to indicate any TCP traffic from the 192.168.1.0 subnet to the 10.2.2.2 both service flows, then it would be ambiguous as to which service flow the CMTS 918 should classify the packets. The same would be true if the reverse classifier was used for the upstream. The cable modem 920 would not be able to distinguish between the two service flows.
As the simple example illustrates, the process of constructing classifiers can become complicated when trying to define a set of classifiers for a large set of service specifications. This can get further complicated due to the fact that many applications use dynamic port assignments and therefore cannot be easily characterized by their well-known ports. Therefore, the AppMgr 922 in some embodiments of the invention tries to ensure that the packet classifiers do not overlap, and that when they do that the MSO is informed so that it may correct the problem or provide priorities to the overlapping classifiers.
Turning attention to
Application names are specified for the service specifications using two pieces of information: the application genre and an expression describing the application name, such as app*.exe. For example, the application genre, “WebBrowser”, may be used for browser applications, or “IM” might be used for instant messaging applications. In some embodiments of the invention, system QoS proxies use one or more expressions to detect an application when invoked. For example, the regular expression “ie*.exe” detects Internet Explorer launches. The FoxFire web browser could also be included in the WebBrowser application, and the expression “foxfire*.exe” can be the trap. If there are other web browsers to be included in the WebBrowser application, additional regular expressions can be added.
Traffic characteristics for applications preferably include information about the sustained bandwidth, bandwidth bursts, latency, and jitter that are to be used by the upstream and downstream connections that support the application. In some embodiments of the invention, traffic characteristics are determined in any one of several ways, including, but not limited to: default application profiles provided by the system developer; a QoS discovery analysis software tool; or MSO experience and experimental trials. The QoS discovery tool analyzes network traffic and determines the optimal application-specific QoS traffic characteristics. Traffic characteristics for both the upstream and downstream information flows can be defined.
In some embodiments of the invention, application-specific QoS can be restricted to specified endpoints, defined in terms of either the source or destination IP address and port number or URL. The list that defines permitted or restricted endpoints is called the Access Control List (ACL). For example, in the WebBrowser application genre, the IP address is preferably not restricted, but instead, the port number is restricted; the permitted IP addresses is defined as a wildcard, and only the port number is specified. Packets from the application with any IP address destination receive QoS over the HFC network, so long as the packet's destination port number is equal to the permitted port number defined in the ACL. For another application, such as the game application Quake, a specific endpoint is preferably defined. The endpoint can be an MSO game server, such that game-specific traffic would receive QoS over the HFC network. As such, an MSO can exclusively offer preferential handling of content that it owns and operates, resulting in both increased revenues and subscriber satisfaction. The MSOs can control the QoS of content riding over its HFC network by employing ACLs that define the list of source or destination addresses that are allowed as well as the list of source or destination addresses that are prohibited.
Some embodiments of the invention allow for the definition of service tiers. A tier of service is a set of service specifications that are offered as a product to customers. For example, an MSO can have a tier of service for on-line interactive games 1012 which bundles together the most popular games. A business services tier 1014 bundles together applications that small and medium businesses are most likely to use, such as remote file transfers, instant messaging, and video conferencing. Once the service tiers are defined, users can subscribe to one or more tiers of service. In an embodiment of the invention, groups of users are defined and service tiers are assigned to the group. For example, a subnet or group of subnets can be assigned to a tier such as “small business.”
In some embodiments of the invention, quality profiles are stored locally on subscribers' computing devices. The local storage of quality profiles allows a first level of admission control to be performed at the subscriber machine, rather than burdening the central application manager and policy server with potentially unnecessary requests. The quality profiles are preferably updated at startup or on a periodic basis, ensuring that QoS requests for the subscriber are current with the subscriber's subscription and application requirements. Subscriber computing devices use the locally stored profiles to make only appropriate QoS requests, and to make those requests with detailed requirements, including, for example, bandwidth, latency, jitter, burst rates, etc., facilitating optimal scheduling of QoS by the network services provider.
Policies are preferably configured through a web portal interface to the central application manager. The MSO can input the authorized applications, traffic profiles, ACLs, service specifications. The policies are configured in the web portal and are then combined into service specifications. Tiers are configured using the service specifications. Subscriber information is either configured via a web interface, or can be retrieved from the MSO's subscriber database. Once the subscriber information is retrieved, a list of tiers can be configured for each subscriber. It is also possible to define groups of subscribers by defining subnets or groups of subnets. In order to allow easy machine-to-machine access to these management tools, such as the MSO's provisioning system, the configuration of QoS policies can also be defined as Web Services in a published web services description language specification. Complementary OSS functions can use these Web Services to integrate the application manager functionality into the MSO's overall management strategy.
Because policies change over time, the MSO can review the policy information and add, change, or delete individual elements via the web interface, in some embodiments of the invention. For example, if an MSO has been using a default traffic profile for a particular game, and has since discovered a more effective profile, it can easily edit the game's traffic profile. Similarly, it is possible to add service specifications to the on-line interactive gaming tier as new computer games become available. The web services enables the MSO to offer subscribers self-provisioning capabilities such as adding and deleting tiers, as well as listing the tiers for which the subscriber is assigned. And, of course, it is possible for the MSO to add subscribers to new services, remove subscribers, and change the definitions of groups of subscribers.
As described above, the subscriber policies and application information can easily be customized by the MSO, in some embodiments of the invention, using a simple Web portal or web services to the provisioning system. Such an interface enables the MSO to manage subscribers, tiers, applications and services, and QoS profiles. New application-specific QoS profiles can be easily designed using a “profile builder”. An exemplary web interface for managing subscribers, tiers, applications and services is shown in
In
Traffic flow specifications can be edited and created through a submenu as shown in
Traffic profiles are created and edited with a submenu such as the one shown in
Flow classifiers are edited and created through a submenu such as that shown in
The service specifications bind traffic profile, flow classifier, and application group, and are edited and created in a submenu as shown in
Additionally, some monitoring and management for QoS may be performed at the subscriber's computing device through the use of a web interface, according to some embodiments of the invention. For example, a subscriber can use the interface of
In some situations, the network IP address from which specific content is retrieved by the subscriber computing devices varies from one subscriber to the next, based upon various factors such as the subscriber's location and placement of content caches. That is, two subscribers connecting to a server using identical URLs may actually be routed to different caches of the server. Furthermore, the location from which a particular subscriber retrieves data may vary over time depending on factors such as network congestion, failures and reconfigurations. Thus, if configuration of classifiers for service flows is based solely on network addresses, specific configurations would need to be made for each user, and these configurations would have to be updated whenever the location of the content was changed. Some embodiments of the invention solve this problem by altering the IP address for an application's URL returned from the local name server. The system QoS proxy residing on the subscriber's computing device uses the URL (rather than the IP address) to configure the classifiers for a service flow. This process is performed at the time the content is requested, so that the most up-to-date location is used.
In order to update QoS profiles at regular intervals, the QoS proxy preferably communicates with the AppMgr with a Keep-Alive message to maintain state information. One element of the message is the current policy configuration for the subscriber. If the QoS proxy detects a QoS profile is not current due to an edit or change at the AppMgr, the QoS proxy requests an updated policy configuration.
Additional details are now provided with respect to the communications interface between a system QoS proxy and an AppMgr, as used in some embodiments of the invention. The interface is designed to use the Session Initiation Protocol (SIP) and the Secure Session Initiation Protocol (SIPS). The transaction between the client (i.e., the machine hosting the system QoS proxy) and the AppMgr is modeled after the RFC 3264 (An Offer/Answer Model with Session Description Protocol (SDP)), and RFC 3312 (Integration of Resource management and SIP) which are hereby incorporated by reference for all that they teach without exclusion of any part thereof. A basic QoS session establishment between the QoS proxy and the AppMgr is shown in
In some embodiments of the invention, all transactions between the client and the AppMgr preferably use SIP version 2 (SIP/2.0). The interface supports the use of various SDP parameters (descriptors), including many at the Session Level, and many at the Media Level.
At the Session Level, the interface supports a protocol version parameter (v=). As per RFC 2327, the protocol version parameter is used to represent the version of the interface definition. Currently this is set to “0” (zero), but will be incremented as future versions of this interface definition are released.
The interface also supports an Owner parameter (o=). As per RFC 2327, the parameter takes several arguments of the form: O=<username><session id> <version> <network type> <address type> <address>. For all transactions, <username> contains the user's login that was used during the registration. As per RFC 2327, the <username> does not contain any spaces. The AppMgr verifies that the tuple, <session id> <version> <network type> <address type> <address> is unique, in order to prevent the processing of duplicate QoS requests. In the cases of a duplicate request the AppMgr responds with a response code of 491. <session id> and <version> are 10-digit Network Time Protocol (NTP) timestamps as per RFC 2327. <network type> is “IN” to indicate Internet, as per RFC 2327. <address type> is “IP4” to indicate IP version 4 as per RFC 2327. <address> is the public IP address of the sender as per RFC 2327. The address is preferably sent in the dotted-decimal representation of the IP address. Thus, when the client is behind a NAT router, it must determine the IP address that is assigned to the NAT router and not the IP address that was assigned the client by DHCP server running on the NAT router. Some embodiments of the invention include a STUN server that provides the QoS proxy with an appropriate IP address for classification. A STUN server can be included at the AppMgr or at a third party site located elsewhere on the Internet. The QoS proxy configuration includes an entry for a URL to the STUN server.
The interface also supports a Session Name parameter (s=), which is “−” to indicate no session name as per RFC 2327.
The interface also supports a Connection Information parameter (c=). As per RFC 2327, the Connection Information parameter takes several arguments of the form c=<network type> <address type> <connection address> where <network type>=IN, <address type>=IP4, and <connection address>=dotted decimal formatted destination IP address for session.
The interface further supports a Times parameter (t=). As per RFC 2327, t=<start time> <stop time> where the <start time> and <stop time> are set to 0 to indicate permanent and unbounded.
At the media level, the interface preferably supports several parameters, as per RFC 2327. A Media Name parameter (m=) takes the form m=<media> <port> <transport> <fmt list>. The <media> argument may take the values of “audio”, “video”, “application”, “data”, and “control”. The media fill is always be set to “application”. The <port> argument is the port that the media will be sent. The <transport> argument specifies the transport protocol. Supported protocols include udp-17, tcp-6, authentication header (AH)-51 and Encapsulating Security Payload (esp)-50. The <fmt list> argument is set to 0.
The interface further preferably supports additional SDP parameters as defined by RFC 3312, which are used when reserving network resources. In some embodiments of the invention, the interactions between the QoS proxy and the AppMgr use a subset of the parameters, as listed below. No further additional parameters are preferably used. A Desired Status parameter (a=des:) is included in an INVITE request, and includes the desired status for the QoS. The Desired Status parameter takes the form: “a=des:” <precondition type> <strength-tag> <status-type> <direction-tag> (e.g., “a=des:qos mandatory local send”). The Precondition Type argument only accepts the value “qos” according to RFC 3312, but some embodiments of the invention extend the range of accepted values. The Strength Tag argument for all transactions is set to “mandatory”. The Status Type argument for all transactions is set to “local” to indicate the segment from the cable modem to the CMTS. The Direction Type argument is set by the client will set to indicate the direction(s) required to the best of its ability: Send=Upstream direction; Recv=Downstream direction; and Sendrecv=Upstream and Downstream.
Some embodiments of the invention define and support additional SDP parameters to allow the AppMgr to determine the QoS requirements for the session. These attributes are formatted as per RFC 2327 attributes, a=<attribute>:value. These additional parameters include an Application Name parameter, where the parameter takes the form “a=application:”
<application name>. The application name parameter is used by the client to indicate the name of the application running on the client machine that triggered the QoS request. An additional parameter is a Configuration Version parameter taking the form “a=configVersion:”<version ID>. The client includes the version value in the request to allow the AppMgr to keep the client's configuration synchronized with any changes made by the administrators. Another additional parameter is a Conversation ID parameter of the form
“a=conversationId:”<conversationID>. The client includes in the request the conversationID from the configuration data it received when it logged in to the AppMgr. Still another additional parameter is a Traffic Profile parameter of the form “a=traffic-profile:”<traffic-profile-name>. The values for traffic profile are consistent with the profile received in the client's configuration file at the time of the registration. Yet another additional parameter is a set of Traffic Classifier parameters, including: a Source IP parameter (“a=source-ip:”<dotted decimal IP address>); a Source Port parameter (“a=source-port:”<port #>); a Destination IP parameter (“a=destination-ip:”<dotted decimal IP address>); a Destination Port parameter (“a=destination-port:”<port #>); and a Protocol parameter (“a=protocol:”<IP Protocol number>). As defined by RFC 2327, this is the transport protocol. Supported protocols include: udp-17; tcp-6; authentication header (AH)-51; and Encapsulating Security Payload (esp)-50.
Turning attention to
In response to the POST 2304, if the subscriber is authorized, the AppMgr 2306 responds with a response 2308 with the client's configuration information as the payload. The response 2308 is preferably formatted to conform to the XML schema in Appendix A. Alternatively, a web services SOAP/XML message is sent instead of a POST 2304, using the same underlying parameters.
In
s=−
For the cases when the AppMgr is not able to honor the request, the AppMgr preferably responds with an error code as per RFC 3261. Such an example of an error is shown in
In some embodiments of the invention, the AppMgr uses response codes as shown in Table 3.
An XML schema for a Client Configuration Payload is attached as Appendix A.
The present invention is not limited to those embodiments described above. For example, in one embodiment, a high-quality communications session is established between two subscriber computing devices in a peer-to-peer configuration. In this scenario, embodiments of the invention invoke “pipe matching” techniques to ensure that each subscriber's QoS requirements are compatible. In another embodiment, the system QoS proxy uses an automatic detection protocol such as Universal Plug n' Play (UPnP) to allow QoS requests to be made on behalf of devices connected locally to the subscriber computing device running the QoS proxy. For example, a networked digital video recorder is detected by the subscriber computing device's QoS proxy via UPnP, and a QoS request is sent on its behalf.
Additionally, some embodiments of the invention are used within a local wireless network operating according to an IEEE 802.11e standard. In such embodiments, TOS/DS bits are marked in packets sent to and from the QoS-requesting application to indicate a higher priority that those packets should receive within the local wireless network. The wireless router honors the priorities set of these bits. Some embodiments of the invention can thus be used even over dedicated, non-shared networks such as DSL, to ensure subscribers receive desired QoS relative to other users on the local wireless network, or relative to other applications running on the subscriber's computing device.
In another embodiment of the invention, a communications session over a WIMAX or “broadband wireless” medium is made high-quality through the use of bit-marking or coloring. The system quality proxy on the subscriber computing device marks DS/TOS bits in the upstream direction, while a remote content or application server (e.g., operated by the network services provider) marks DS/TOS bits in the downstream direction. The subscriber's wireless modem and the service provider's wireless router place packets in queues according to the priority set by the DS/TOS bits. The AppMgr can be used for authentication and admission control, such that the system quality proxy on the subscriber's computing device intercepts a network call by an application and makes a request for QoS according to locally stored policies. If the user and application are authenticated, the AppMgr instructs the quality proxy to mark its TOS/DS bits, and instructs the content or application server to similarly mark its TOS/DS bits for the subscriber. Bit coloring is also used in this manner in some embodiments of the invention over networks other than WIMAX or broadband wireless that are policy based and honor bit coloring protocols.
In view of the many possible embodiments to which the principles of the present invention may be applied, it should be recognized that the embodiments described herein with respect to the drawing figures are meant to be illustrative only and should not be taken as limiting the scope of the invention. Additionally, those of skill in the art will recognize that the illustrated embodiments can be modified in arrangement and detail without departing from the spirit of the invention. Therefore, the invention as described herein contemplates all such embodiments as may come within the scope of the following claims and equivalents thereof.
