|Published (Last):||14 May 2012|
|PDF File Size:||2.97 Mb|
|ePub File Size:||6.19 Mb|
|Price:||Free* [*Free Regsitration Required]|
VMware SRM 5. The Cogency Monitor An PowerVR Series5 Graphi A document in specification format considered largely complete, but lacking review by other VE and PE vendors. Drafts are susceptible to substantial change during the review process. A stable document, reviewed, tested and validated, suitable to enable cross-vendor interoperability. Threaded architectures One of the functions of the component called the Session Manager is to determine which resource managers are required for a particular session request and to communicate with the resource managers to allocate the resources required for the session.
This document describes architecture and signaling model for the negotiation of resources between the Session Manager and the resource managers responsible for allocating session based resources. Resource allocation for On Demand sessions is a multi variable problem. The optimal set of resources for an On Demand session request must take into account multiple interrelated factors. For example, when an On Demand session request is made from an On Demand Client for a particular asset, an Edge Resource Manager and On Demand Resource Manager may each have more than one resource that could be allocated for the request.
The list of Streaming Server ports and QAMs that may be used for this request may be further reduced when the connectivity and available bandwidth of the IP transport network is taken into account. The models used to implement resource allocation algorithms for the Next Generation On Demand architecture must take these inter-related dependencies into account.
There are a couple different models that can be used to implement resource allocation. One is called pessimistic resource allocation while the other is called optimistic resource allocation.
In addition to resource allocation models, there are a number of ways by which resource allocation functionality can be divided between the Session Manager and the resource managers involved in a session. One method is to make the Session Manager the final arbiter for the resource allocation for a particular session.
In this architecture, the Session Manager communicates with each resource manager. Another architectural model for resource allocation is one in which the resource managers themselves arbitrate the set of resources to be used for the session.
In this architecture, the Session Manager is responsible for determining which resource managers are required for a particular session as well as specifying rules for allocating resources based on service and business logic. Instead of communicating with each resource manager, the Session Manager lets the resource managers themselves negotiate the actual resources to be used for the session.
The architecture and interfaces for threaded SRM architecture and pessimistic resource allocation model are only described here as optional informative materials for future references. This phrase means that the item is an absolute prohibition of this specification. This phrase means that there may exist valid reasons in particular circumstances when the listed behavior is acceptable or even useful, but the full implications should be understood and the case carefully weighed before implementing any behavior described with this label.
To achieve this goal, it is best to design then architecture in such a way that functionality and the resulting processing requirements are divided as evenly as possible between components. The NGOD architecture spreads functionality across the components of the architecture by having the Session Manager implement the service and business logic associated with resource allocation while letting the Resource Managers themselves allocate the actual resources to be used for the session.
Scale with Large Networks As On Demand Services become more popular, the number of potential resources that may be used in a particular session may grow very large. While each resource manager may have many resources to choose from for a particular session, there may be a small number of resource combinations that meet the routing and resource requirements for the session. The architecture needs to support optimized resource allocation when many potential resources for a particular session exist.
The result is a distributed resource selection algorithm which pairs down the potential set of resources to a set which meets all the resource and routing requirements for the session. Minimal Session Setup Latency The architecture should be designed to result in minimal session setup latency. This results in a couple goals for the signaling architecture.
The signaling architecture should minimize the number of messages used to set up session and allow as much parallelism as possible between the components involved in setting up the session. The NGOD architecture supports a number of different resource allocation signaling models. Each of these models represents a different set of tradeoffs between optimal resource selection and session setup latency.
Models that use an optimistic approach to resource allocation have lower session setup latency but also exhibit a low probability of optimal resource allocation. For example, it may be more important to fulfill resource requests for sessions associated with a pay per view movie service than for sessions associated with a free On Demand Service. The architecture should be able to translate service and business logic into a set of rules to be incorporated in the resource selection process.
Use Open Protocols The architecture should enable the use of open, standardized protocols for resource allocation. The interface definitions associated with the architecture should map in a straight forward manner to the primitives and state machines of existing protocols if possible. Flexibility to incorporate new Resources, Resource Managers The architecture should make it possible to incorporate new classes of resources along with new resource managers without having to make major modifications to other components of the system.
This results in an architectural goal to use a single protocol for all resource signaling between the session manager and resource managers. Capability to Modify Resources for Active Sessions Some services may require the ability to add video assets or other resources to an existing session.
The architecture should allow the ability to modify or add resources to an existing session. This document does not include an example session signaling flow showing resources being added or modified in existing sessions. Future versions of the specifications will provide this information. Work with Heterogeneous Transport Networks The architecture should support video transport networks that are not fully connected.
Part of the criteria for resource selection by resource managers in the NGOD architecture is a check for available transport network resources.
This document includes the description of a component called the network resource monitor which can be used by resource managers to determine the availability of network resources between video components. The network resource monitor can also be used to provide information on partially connected networks. The methods used for allocating network resources such as link bandwidth should be able to scale to complex topologies and large numbers of active sessions.
Future release the NGOD architecture distributes the function of network resource reservation using RSVP as a signaling protocol to reserve network resources.
RSVP has been deployed in complex topologies with large numbers of active sessions. Simple Component Implementations The architecture and resulting protocols should enable the components to be implemented using simple state machines. The architecture should also minimize the amount of state that needs to be maintained across components. The RTSP specification defines a client state machine as well as a server state machine. Both the RTSP client and server state machines consist of 4 states with well defined state transitions.
They are: 1. Service logic is used to determine which resource managers should be contacted for a particular session request. Rules associated with business logic can be translated into parameters of a resource request. For example, resource requests for a Pay per View service may have a higher priority associated with them than resource requests for a Free On Demand service.
This priority can be carried as a parameter within a resource request. Resource Manager Selection Before a set of resources can be allocated for an On Demand session, the Session Manager needs to select a set of resource managers that will be used to allocate the resources for the session.
Resource Allocation Once the classes of resources required for a particular session are known and a set of resource managers are chosen to allocate those resources, the actual resources to be used for a session need to be chosen. This logic provides a set of resource managers that must be contacted for resource allocation for the session. Once the resource managers to be used for a particular session are selected, the Session Manager and resource managers cooperate to implement a distributed resource allocation algorithm.
There are 2 different models that can be used to implement resource allocation in the Next Generation On Demand Architecture. A resource allocator performs the resource accounting function of resource management. A resource allocator keeps an inventory of available resources and determines whether or not a requested resource is available. The function of the resource allocator is always bundled in a resource manager component.
A resource selector is responsible for choosing a specific resource from a list of potential resources. The operation of different combinations of resource allocation models and architectures can be described by viewing how resource selectors and resource allocators are bundled into the Session Manager and the resource managers specified by the NGOD architecture.
This is illustrated in Figure 1. The Session Manager then informs each resource manager which resource it has selected for the session and the resource manager allocated the resource. When pessimistic resource allocation is implemented in the threaded architecture, both resource allocation and resource selection are implemented in the resource managers. This is illustrated in Figure 2. Since the resource managers implement the resource allocation and resource selection functions, the set of resources to be allocated for the session are negotiated between the resource managers themselves.
Each resource manager reserves a set of resources that may be used for the session and provides the list of those resources to the resource selector. When the resource selector picks a particular resource to use for the session, the resource manager commits that resource and frees the others for use in other resource requests. Optimistic resource allocation is based on the principle that Session Manager can rely on each Resource Manager to select and commit a single resource with a high degree of probability to satisfy the requirements of a session.
It is then up to the Session Manager to further validate all the selected resources for the session. If the Session Manager finds that a set of resources obtained from resource managers does not work, it must keep trying until it gets a compatible set of resources from the resource managers. Resource managers that receive a resource request from a Session Manager using an optimistic resource allocation algorithm implement a single phase resource allocation algorithm.
Upon receiving a resource request from the Session Manager the resource manager will commit a resource from its resource pool that matches the resource requirements specified in the resource request.
If the request is unsuccessful then the Session Manager needs to request another resource with different criteria from the resource manager. If the Session Manager does not have any alternate candidate criteria for a specific resource type, it will need to release other resources and iterate till it gets a compatible set.
Where the probability of success is very high, then Optimistic Allocation schemes can provide a faster allocation scheme than a pessimistic scheme. However, care must be taken if the probability drops since some algorithms can lead to thrashing.
It should be noted that an Optimistic scheme is not designed to produce the optimal result; rather it provides a valid solution. The Session Manager validates resource compatibility, but it does not select them. It is allowed that the Session Manager specifies a list of potential resources to each resource manager which each independently select a specific resource for the session.
This is illustrated in Figure 3 below. In the NGOD architecture, the Session Manager manages all interactions between session signaling and resource allocation signaling. This is illustrated in Figure 4. The Session Manager uses interface S2 to obtain authorization for the request and to obtain the list of assets and the required bandwidth for the session from the Purchase Server.
The Client sends two key pieces of data to the SM on session setup, the token representing the asset it wants to view and one or more QAM Names representing its location in the distribution network. The detailed signaling flows for SRM use cases are provided in a separate specification. After the Session Manager receives a list of assets and bandwidth from the Purchase Server, it implements the distributed resource management function using the resource managers specified in the NGOD architecture.
CableLabs Unifies Specs For Unicast Video
Sign In Register. The Edge. Private Networks. White Papers.
Concurrent Not in Comcast's VOD Future
But NGOD also will address the increasing complexities of the on-demand business: encoding and ingesting local content; advertising insertion; building open interfaces between hardware product; and scaling the current platform. For example, Comcast is in its second season of offering on-demand National Football League highlights and the first season of National Hockey League content, both requiring quick ingestion turnarounds and constant refreshment. More digital subscribers will place additional demands on the platform. A third element: Creating an architecture that allows for ad insertion into on-demand content, he said. The real-time ingestion of content and its quick propagation to edge servers is a growing issue for cable systems, as local content and sports highlights are added to the VOD platform.