1 / 12

Receiver-driven Bandwidth Adaptation for Light-Weight sessions

Receiver-driven Bandwidth Adaptation for Light-Weight sessions Elan Amir; Steven McCanne and Randy Katz Presented By: Monu Mathur. Basic Concepts: What is Multicasting? -A single source can send data to multiple receivers at the same time.

odele
Download Presentation

Receiver-driven Bandwidth Adaptation for Light-Weight sessions

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Receiver-driven Bandwidth Adaptation for Light-Weight sessions Elan Amir; Steven McCanne and Randy Katz Presented By: Monu Mathur

  2. Basic Concepts: • What is Multicasting? • -A single source can send data to multiple receivers at the same time. • -Sender can send packets to a multicast group address. • What is Multicast Backbone or MBone? • -A virtual network embedded in the unicast internet that • implements IP multicasting. • What is the Light-weight Sessions Architecture? • -MBone tools multicast packets to the multicast group address. Receivers “tune-in”. • -Loosely coupled, real-time model.

  3. Issues Discussed: • All sources are not of equal interest to the receivers. • Current methods allocate bandwidth in a fixed and • inflexible manner. • Users have to configure, enable and disable their • transmissions manually. • Proposal: • Media sources should adapt their transmission rates • to meet collective preference of receivers. • Accommodate and avoid network congestion by • controlling the traffic injected into the network across • all senders. • Augment existing adaptation schemes with receiver • feedback.

  4. A number of schemes address the problem of adjusting the media • delivery rate at each source. However the issue of receiver feedback, that describes the receiver feedback throughout the session has not been addressed. • SCUBA • Scalable ConsensUs -based Bandwidth Allocation • New protocol that deals with trying to solve the issues discussed above. • A scalable, light-weight feedback protocol. • Reflects receiver interest back to the media services. • Employs receiver consensus to allocate session bandwidth • among resources. • Has 2 variants: • “Flat Delivery”:Receiver feedback constraints the rate chosen by sender- based adaptation algorithm.

  5. “Layered delivery”: Feedback is used to control the mapping of • the source signal layers to the network channels. • Advantage of Receiver based multicast: • The burden of adaptation is moved from the source • to the receiver. • Enhanced scalability of the system. • Working of the protocol • Problem addressed • by SCUBA: • Static bandwidth • allocation-Sources • send at a fixed rate • and bandwidth is divided equally

  6. The solution of the problem is given by SCUBA: Dynamic Bandwidth Allocation: Sources adjust their transmission rate dynamically and do the bandwidth allocation accordingly. Problem: How can we determine a single bandwidth partition that simultaneously satisfies all receivers? Solution: Participants reach “consensus” through voting and the bandwidth partition is done accordingly. SCUBA uses the “exit poll” principle: Source bandwidth partition is derived from a small unbiased sample of receiver population.

  7. Feedback Algorithm: • Each receiver associates a weight with every active source. • The weights define the relative priority of each source with respect • to a given receiver. • Receivers periodically advertise a source-weight report containing • sources and their corresponding weights with respect to that receiver. • Scaling mechanism is employed by announce/listen control protocols • including SRM and SAP. • Each receiver announces its source weight report by multicasting it. • Each source listens for the reports and combines them into an • aggregate metric called average source weight.

  8. Translation of average source weight to • sender’s rate control decision: • Case 1: Flat Delivery • Flat delivery adjustment decision maps the transmission weight • directly to the source transmission bandwidth. • The mechanism is arbitrary and can be customized depending • on the application or environment. • Allocation is done in such a way that each active source multicasts • it’s signal continuously even though at a very low rate.

  9. Case 2: Layered Delivery • Overcomes the limitation of sending the date at the rate of the • slowest link. • Each source generates a layered media stream striped across • multiple network channels. • Distinguishes more important • sources from less important • ones. • Given a source weight, • the source determines it’s • signal layer mapping. The • source then transmits each • of it’s signal layers defined • by the mapping.

  10. Scalability Issues • SCUBA makes decisions at each source on behalf of all receivers • in the network: Hence prone to scaling problems. • Sacrifices absolute consistency in favor of scalability through the • use of sampling. • The two factors that affect the scalability of SCUBA are: • - Control message size: Limit the number of sources for which • each receiver reports to a small constant. • - Expected source weight convergence time: Change in • receiver interest is not reflected back to the receiver • immediately. This latency is called average source weight • convergence time (ASCT).

  11. Solution to the problem is provided by keeping the time taken by • the receivers to reach the consensus (ASCT) small. • The time-scale of fluctuations in receiver interest is determined by • the users. Hence the average source weights must be computed • on the time scale of seconds. • To avoid the linear time growth of ASCT, sampling is used. • The convergence time of the algorithm is decoupled from the • session size allowing scaling to large sessions.

  12. Applications • SCUBA can support: • Video Conferencing • Media Gateway Control • Floor Control • Future work: • An alternative approach to Samplingin orderto enhance the scalability of the announce/listen receiver interest protocol.

More Related