vbns usage alloc priority queuing video performance

priority queuing video performance

(nsfcode )
category: application type: research
adminRick Wilder
2100 Reston Pkwy, Restion, VA 22021
703-715-7114
tpoc Chuck Song csong@mci.net
2100 Reston Pkwy, Restion, VA 22021 703-715-7082
description This is a demo of priority queueing effects on the performance of multimedia applications. We will use one workstation at the SC'95 show to display video scenes captured by a workstation with video camaras at Reston node of vBNS TestNet, using the NV/VAT software. There will be two windows on the diplay workstation showing the same video scenes, one with good performance because of priority queueing for its traffic, and one with poor performance due to no/lower priority. The two traffic streams, generated from the Reston workstation, will traverse the TestNet where they are subject to traffic congestion. DEC/Alpha workstation on the TestNet will generate background traffic causing congestion on the Netstar gigarouter at Restion MCI Lab. Different Rate Queues on the gigarouter's ATM card will be used to support the two different traffic streams. Under congestion, one stream with lower priority Rate Queue should experience longer quueing delay and possibly higher loss rate than the the stream using the high priority Rate Queue.

This demo is designed such that the high intensive backgound traffic will be limited to only the TestNet and the bandwidth demand on the production vBNS will be low, not exceeding 4 Mbps. Sharing with other applications of the production vBNS resources is not affected, provided that other applications do not cause congestion on the vBNS.

This demo is a first step using the vBNS as a platform for experimental research in the area of IP integrated services.

non vbns resources needed

machines:
  1. Workstations with video quipement installed and used at Reston, VA.
  2. SDSC: A workstation (likely a SGI) with NV and VAT software. This machine will be on loan from SDSC to this demo for the duration of the SC'95 show. It will be physically moved in the Netstar show booth and connected to a Netstar gigarouter via FDDI interface.
  3. Netstar booth: A Gigarouter at the Netstar booth during the SC'95 show. The router can be shared between this project and other Netstar demo uses.
facilities:
other:
addnl sites:
attachments at existing sites:

dedicated vbns resources


PVC endpoints and parameters
A. endpoints : 2 in testnet
1 in backbone MCI Lab at Reston, VA
1 from SDSC to Netstar booth on the CS95 show floor
B. QOS parameters for circuit No specific guaratee on the testnet is needed. 4 Mbps bandwidth on the vBNS production backbone is needed for an acceptable demo performance. However, no special effort is required to make this a guaranteed resource.
C. Total hours VC required. Needed for the whole duration of the SC'95 show.
shared vbns resources:
expected peak b/w 4 Mbps
protocols: tcp/ip
workstation time: A. DEC/Alpha Workstation on the testnet (Reston and PSC). Total amount of time, none-exclusive, is for the duration of the SC'95 show.
test net time: TestNet nodes to be used are the Reston and PSC sites. Heavy traffic will be generated between the two sites.
other:
biographical information:

commodity access issues:

source of financial support MCI with equipment on loan from SDSC at the SC'95

update 7 jan 96 notes: current priority queueing implementation is not based on CBQ, but some proprietary stuff. MCI provided netstar with a definition of the interface spec with which they can integrate with CBQ or MIT's code relatively easily into their on-card software (August 95). meeting with managment and they sound supportive of the effort.
return to menu return to menu
last updated Thu, ,,95, 11:00
questions or comments: kc@nlanr.net