A Multi-Path Routing Service for Immersive Environments

1
A Multi-Path Routing Service for Immersive
Environments

• Sherlia Shi, Lili Wang, Kenneth L. Calvert,
• and James Griffioen
• Laboratory for Advanced Networking
• University of Kentucky

Supported by NSF, DARPA, and Intel
2
The Metaverse

• Goal design, implement, deploy, use and evaluate
  networked, collaborative, visualization
  environments that act as portals into the
  Metaverse.
• Basic Problem Visualize and interact with 3D
  models, simulated objects, time-dependent data,
  image data, avatars, and video streams in
  collaboration with remote users across the
  Internet
• Research Areas computer vision, graphics,
  networking protocols and services

Neil Stephenson, Snowcrash, 1997.
3
An Example Metaverse App
A CFD Example - Visualize a 3D animated
simulation - Interact with the model -
Collaborate with colleagues or students at
remote metaverse portals
4
Metaverse Components
5
Metaverse Portal
6
Metaverse Portal
Immersive Display Environment
Based on inexpensive, commodity,
self-configuring, networked components
7
Metaverse Portal
Immersive Display Environment

• rendering engines

8
Metaverse Portal
Immersive Display Environment

• rendering engines
• projectors
• cameras

Driven by cheap, commodity, graphics and video
cards
9
Metaverse Portal
Immersive Display Environment

• rendering engines
• projectors
• cameras
• motion sensors

10
Metaverse Portal
Immersive Display Environment

• rendering engines
• projectors
• cameras
• motion sensors
• local area network

11
Metaverse Portal
Immersive Display Environment

• rendering engines
• projectors
• cameras
• motion sensors
• local area network
• control processors

12
Immersive, Multiprojector rendering(18
projectors)
Walkthrough 1
View 1
Walkthrough 2
View 2
13
Metaverse Data

• Data consists of multiple components
• Texture maps
• Polygonal meshes
• Volumetric Data
• User Control Data (position tracking, 3D mouse)
• Same data may be represented in arbitrary ways
• Resolutions potentially variable resolution
• Encodings
• Time Series Data (gt video-like)
• Generated Data

14
Metaverse Applications

• Goal Offer the most compelling immersive
  experience possible!
• Assumptions
• Massive amounts of complex data to be transmitted
• Cannot assume over-provisioning or QoS guarantees
• Users definition of compelling depends on
  several factors
• Application can design for these factors
• Application does not want to know details of
  transport
• Implications
• Not all data can/needs to be sent
• Data that is the most important to send depends
  on the user and the network.
• Communication channel must support cooperation
  between the network and the applications.
• Need to maximize bandwidth subject to (good net
  citizen) constraints

15
Transport API Requirements

• Feedback about transport capabilities.
• Single end-to-end communication channel need a
  simple session abstraction -- App is not
  interested in implementation details.
• Data importance App specifies what data to
  transmit to achieve compelling
• Sender / Receiver Cooperation sender specifies
  data, receiver defines quality
• Reliability data transmitted reliably

16
Feedback

• Feedback could describe any aspect of the
  service
• Feedback aspects may be network-dependent
• Feedback could be at many levels of detail.
• Apps dont want to deal with the above
  complexity!
• Apps have no desire to control the transport
  service (i.e., twist transport knobs).
  Understanding and mapping knobs to user
  satisfaction is too hard.
• Simply want to know what if answers! Let
  transport service figure out the mapping!

17
New Service Abstraction

• Sender
• Specify important data
• Change definition of important as needed
• Receiver
• Specify a prioritized delivery plan
• Let service map the delivery plan onto the
  underlying network
• Provide predictions of how well the service will
  perform
• Goto 1

18
Prioritized Data

• Specified as a hierarchy of nodes
• Leaf nodes represent a semantically meaningful
  data unit, called an object. Objects are the
  smallest logical unit and are delivered as a
  whole
• Internal nodes represent groups of objects and
  define the percentage of bandwidth to give to
  each subtree.
• Each tree denotes a priority level
• Children nodes may require ordered delivery

19
Example Transmission Specification
Priority 0
Priority j
1
1
0.4
0.6
0.2
0.5
0.3
c
0.5
0.3
b
a
0.2
Unordered, No weights
Ordered
20
Metaverse Transport Service API

• int open_session(struct sockaddr svrAddr)
• int close_session(int sessId)
• int set_recv_handle(int sessId, int objId, void
  buf, int len, CallBackFunction call_back)
• MetaInfo schedule_objects(int sessId, MetaInfo
  objects, int modified)

21
A Multi-path Transport Service

• Goal
• Maximize throughput
• Find independent paths
• Distribute load across paths
• Minimizing network load
• Limit the hops and delay
• Share bandwidth across links fairly
• Need a sharing policy
• Implication
• Need more control Use Overlay/Grid
• Control routing
• It becomes a Shared Service

22
PlanetLab Example
Improving Throughput (v2)!

• Improving Throughput (v1)
• use TCP unfriendly protocols
• use multiple TCP flows

23
Transport Service Tasks

1. Discover network characteristics to find
   high-bandwidth communication channel(s) between
   source and destination
2. Select a set of routes subject to constraints
   limiting the load imposed on the network
3. Dynamically map object hierarchy to paths (i.e.,
   distributed load across the selected paths).

24
Residual Bandwidth Discovery

• Each overlay node passively monitors current
  bandwidth usage to other overlay nodes
• Avoids active probing (ping, iperf, traceroute,)
• Record two variables for an active virtual link
• Peak bandwidth
• Average measured bandwidth
• Difference is an estimate of residual bandwidth
• Information built up over time and shared among
  nodes

25
Path Selection

• Selection algorithm must choose multiple paths to
  optimize some objective function, subject to
  various constraints.
• How many paths should be selected?
• What is the maximum length of a path?
• Are some paths preferred over others?
• Are the paths disjoint?
• We evaluated three path selection algorithms

26
Random Path Selection

• Randomly select paths from the source to the
  destination.
• Start from source and randomly select a link that
  satisfies the constraints, otherwise backtrack
  and try another link
• Only guarantees that the virtual paths are
  disjoint, not the physical paths

27
Closest First Path Selection

• Closest First Multipath (CFM) routing uses a
  breadth-first approach to select paths with the
  shortest hop count.
• Start from the source and select all edges
• For each selected node, select all edges
• When destination is reached, add path to feasible
  set
• If more feasible paths than allowed, select from
  feasible set randomly.

28
Widest First Path Selection

• Widest First Multipath (WFM) routing selects the
  paths with the highest bandwidth
• Start with the source, following the widest path
• If the destination is reached, the widest path
  edges are removed, and the algorithm restarts
• If there are more feasible paths than allowed,
  paths are selected from the feasible set randomly

29
Mapping Data to Paths

• A variety of mappings possible
• Our approach sends all fragments of an object
  over the same path to minimize fragmentation
  effects
• Our goal is to minimize the finish time
  (adhering to the data prioritization plan), while
  satisfying the buffering constraints at the
  receiver.

30
Experimental Results
Random
Closest First
Widest First

• Alpha sharing policy ( of residual for local
  traffic)
• When alpha is 0 (no protection) and network load
  is high, all algorithms perform poorly
• Medium values of alpha produce good throughput at
  low load and converge to the shortest path at
  high load
• Random tends to pick paths that are unnecessarily
  long and have low bandwidth

31
5th Percentile Results
32
Correlated Paths

• Are hard to identify, because topology info is
  not readily available
• Traceroute is not sufficient
• Not clear how important it is to know if paths
  are correlated or not
• We developed two passive techniques to identify
  correlated links each has its own advantages.

33
Summary

• Metaverse-like applications require a new API
• We proposed an API that allows application and
  network to work together
• Proposed a passive monitoring approach to
  estimate available bandwidth
• Designed and evaluated three algorithms for path
  selection
• Developed an algorithm to load balance across
  paths based on the data priority spec.