Part 17: Networking Technology for Virtual Environments

Transcription

1 Part 17: Networking Technology for Virtual Environments Virtuelle Realität Wintersemester 2006/07 Prof. Bernhard Jung Overview Introduction Data transfer Communication architectures Managing dynamic shared state Consistency-Throughput Tradeoff Centralized/Shared Repository Frequent State Regeneration (Blind Broadcast) Dead Reckoning Further Information S. Singhal & M. Zyda. Networked Virtual Environments Design and Implementation. Addison-Wesley,

2 Networked Virtual Environments... Virtual environments in which several users can interact with each other in real time Participants are represented through avatars Besides the avatars, autonomous agents may be present in VE Other participants' action can be viewed in real time Participants can communicate with each other (speech, gesture, chat, facial expression) Interaction with other participants and virtual objects grasp virtual objects and hand over to other participants draw on virtual whiteboards shoot (e.g. soccer balls)... Networked Virtual Environments Applications Games Military simulations SIMNET, DIS, HLA (High-Level Architecture) Concurrent engineering, design reviews in teams teleconferencing telemedicine... 2

3 Networking Primer: Fundamentals of data transfer Latency Delay of transfer: Amount of time required to transfer a bit of data from one point to another Reasons for network latency speed of light delays (8.25ms of delay per time zone) delays from the computers themselves delays from the network (LAN, telephony network, satellite, ) Bandwitdh: The rate at which the network can deliver data to the destination point Available bandwidth determined by wire and hardware e.g. modem 56 kbits per second, Ethernet 100 Mbps, fiber optical wire 10 Gps Reliability: Measure of how much data is lost by the network during the journey Two categories data loss - discarded by the network data corruption - content of data packets is changed When reliability needed send acknowledgement Networking Primer: Network protocol Network protocol Describes the set of rules that two applications use to communicate to each other Consists of three components: Packet format Describes what each type of packet looks like Tells the sender what to put in the packet Tells recipient how to parse the inbound packet Packet semantics Sender and recipient must agree on what the recipient can assume if it receives a particular packet What actions the recipient should take in response to the packet Error behavior Rules about how each endpoint should respond to various error scenarios Thousands of protocols in practical use: tcp/ip, ftp, http, rtp, simultaneous use of several protocols, e.g. www: http, tcp/ip, "protocol stack" 3

4 Networking Primer Network protocols TCP/IP TCP: transmission control protocol; IP: Internet protocol TCP layered on top of IP, referred to as TCP/IP most common protocol in use today point-to-point connection Main advantage: reliability guaranteed delivery of packets, in original order, check sum Each endpoint can regard a TCP/IP connection as a bi-directional stream of bytes between two endpoints Disadvantages overhead for packaging of data possible delays in delivery due to ordering guaranty Networking Primer Network protocols UDP User Data Protocol is a lightweight communication protocol point-to-point, connection-less transmission best-efforts delivery: not reliable, no ordering guaranty for package delivery little overhead for data packaging, instantaneous delivery UDP Broadcasting data are sent to all hosts in a network even if some of the hosts are not interested can only be used in a LAN environment IP Multicasting (using UDP) data are sent to some hosts in a network many routers are still not capable of handling multicast subscriptions 4

5 Communication Architectures: Client-Server (logical architecture) clients communicate with each other via a central server Server Client 1 Client 2 Client n communication via the server slower than direct communication server is bottleneck performance of complete system depends on server performance server may decide if delivery of certain data to a client makes sense, or may combine several messages to a client into one message Communication Architectures: Client-Server It might really look like this physically! LAN Client 1 Client 2 Client n Server physical architecture on a LAN 5

6 Communication Architectures: Client-Server or this, if we are using phone lines & modems e.g. simple video game architectures... Server phone line phone line phone line Client 1 Client 2 Client n physical architecture with phone lines Communication Architectures Client-Server with multiple Servers Server 3 Server 1 Server 2 Client 1,1 Client 1,2 Client 1,n Client 2,1 Client 2,2 Client 2,n overall system performance less dependent on performance of single server improved scalability over single-server systems typically very fast data networks between servers otherwise very different latencies of communication between clients on same / different server 6

7 Communication Architectures Peer-to-Peer Host Host Host Host hosts communicate directly with each other, i.e. no server better scalability than server-based architectures broadcasting or multicasting (better) host decides locally which other hosts to inform about state changes area of interest management, e.g. based on spatial partitioning Managing Dynamic Shared State What is dynamic shared state? The dynamic information that multiple hosts must maintain about the net-ve Accurate dynamic shared state is fundamental to creating realistic virtual environments. It is what makes a VE multi-user. e.g. multi-user soccer game: positions of players and ball consistent on all hosts Management is one of the most difficult challenges facing the net-ve designer. The trade off is between resources and realism. Overview Dynamic Shared State Consistency-Throughput Tradeoff Centralized/Shared Repository Frequent State Regeneration (Blind Broadcast) Dead Reckoning 7

8 Managing Dynamic Shared State Consistency-Throughput Tradeoff information is generated on one host and mirrored on another because of network latency, data can be outdated when it arrives at the receiving host without synchronization, mirrored information can only be trusted to some extent ( Joe is near (10,20) ) Managing Dynamic Shared State The Problem: Latency Player A Sends Update Here Update arrives after 100 ms Player A is here 100 ms later Player B 8

9 Managing Dynamic Shared State Consistency-Throughput Tradeoff consistent states can only be achieved through synchronization but only at the cost of slower update rates (Consistency-Throughput Tradeoff) Consistency-Throughput Tradeoff Reliable (Gets there) Real-time (On time) Scalable (Group size) It is impossible to allow dynamic shared state to change frequently and guarantee that all hosts simultaneously access identical versions of that state. We can have either a dynamic world or a consistent world, but not both. Design Implications For a highly dynamic shared state, hosts must transmit more frequent data updates. To guarantee consistent views of the shared state, hosts must employ reliable data delivery. Available network bandwidth must be split between these two constraints. 9

10 Tradeoff Spectrum System Characteristic Absolute Consistency High Update Rate View consistency Dynamic data support Network infrastructure requirements Number of participants supported Identical at all host Low: Limited by consistency protocol Low latency, high reliability, limited variability Low Determined by data received at each host High: Limited only by available bandwidth Heterogeneous network possible Potentially high Managing Shared States Techniques Shared Repositories Blind Broadcast Dead Reckoning More Consistent More Dynamic 10

11 Centralized / Shared Repositories Maintain shared state data in a centralized location. Protect shared states via a lock manager to ensure ordered writes. Three Techniques Shared File Directory Repository in Server Memory Distributed Repository (Virtual Repository) Centralized / Shared Repositories: Shared File Directory Absolute Consistency! Only one host can write data to the same file at a time. Must have locks. Slow! Does not support many users. User User User Update Read Synchronization Locks state state Centralized Data Store Update Read User User User 11

12 Centralized / Shared Repositories Server Memory Faster than Shared File because each host uses does not have to open and close each file remotely Server crash is catastrophic Maintaining constant connection may strain server resources Virtual Repository Tries to reduce bottleneck at server Hosts communicate directly to each other following a protocol of information sharing can even tailor who you talk to Better fault tolerance (server crash less catastrophic) depending on protocol creating the virtual files Eventual Consistency 12

13 Centralized / Shared Repositories Discussion Advantages Provides an easy programming model Guarantees information consistency No sense of data ownership is imposed; any host can update any piece of the shared state Disadvantages Data access and update operations require an unpredictable amount of time to complete Requires considerable communications overhead due to reliable data delivery Vulnerable to single point failure Push systems may send info where it is not needed. Limited number of users (else you overload the server or the network) One slow user can drag everyone down Frequent State Regeneration / Blind Broadcasts Owner of each state transmits the current value asynchronously and unreliably at regular intervals Clients cache the most recent update for each piece of the shared state Hopefully, frequent state update compensate for lost packets Broadcast is sent blind to everyone No assumptions made on what information the other hosts have Usually the entire entity state is sent No acknowledgements No assurances of delivery No ordering of updates Why use? Can t afford overhead of centralized repository May not have demanding consistency requirements 13

14 Frequent State Regeneration / Blind Broadcasts Explicit Object Ownership With blind broadcasts, multiple hosts must not attempt to update an object at the same time Each host takes explicit ownership of one piece of the shared state (usually the user s avatar) Commonly used in online gaming (DOOM, Diablo) Request Ball Lock Lock Manager Request Ball Lock Grant Ball Lock Reject Ball Lock HOST A Update Ball Position HOST B Explicit Object Ownership: Proxy Update Updating state of objects owned by someone else Request Ball Lock Lock Manager HOST A Grant Ball Lock Update Ball Position Request Update Ball Position Update Ball Position HOST B 14

15 Explicit Object Ownership: Transferring Ownership Notify Lock Transfer Lock Manager HOST A Acknowledge Lock Transfer Update Ball Position Request Ball Ownership Grant Ball Ownership Update Ball Position HOST B Blind Broadcasts - Discussion Advantages Can support a larger number of users at a higher frame rate and faster response time Disadvantages Requires large amount of bandwidth Network latency impedes timely reception of updates and leads to incorrect decisions by remote hosts Network jitter impedes steady reception of updates leading to jerky visual behavior Assumes everyone broadcasting at the same rate if this is not the case then noticeable to users may be very noticeable between local users and distant destinations 15

16 Dead Reckoning Protocols Transmit state updates less frequently by using past updates to estimate the true shared state. Prediction: how the object s current state is computed based on previously received packets. Convergence: how the object s estimated state is corrected when another update is received. Current Position Updated Position Time (y) Time (x) Time (z) Convergence Time (y) Predicted Position Dead Reckoning Protocols Each host estimates entity locations based on past data No need for central server Sacrifices accuracy of shared state for more participants 16

17 Prediction Using Derivative Polynomials Zero Order (simplest) x(t + Δt) = x(t) really state regeneration-assumes the object doesn t move First Order (velocity) x(t + Δt) = x(t) + v x Δt Second Order (acceleration) x(t + Δt) = x(t) + v x Δt+ a x (Δt) 2 Higher Order Approximations greater bandwidth required greater computational complexity Object Specialized Prediction Object behavior may simplify prediction scheme e.g. a plane s orientation angle is determined solely by its forward velocity and acceleration. Land based objects need only two dimensions specified. Desired level of detail - often do not need to be precise with some aspects e.g. exact flicker of the flames of a burning vehicle not necessary; enough to say it is on fire. e.g. the same with smoke 17

18 Convergence Algorithms Tells us what to do to correct an inexact prediction: Updated Position Current Position Prediction Error Predicted Position Trade-off between computational complexity and perceived smoothness of displayed entities Convergence Algorithms Zero order or snap convergence: Advantage: Simple Disadvantage: Poorly models real world. Jumping entities may distract users. Updated Position Predicted Position 18

19 Convergence Algorithms Linear Convergence Advantage: Avoids jumping Disadvantage: Does not prevent sudden or unrealistic changes in speed or direction. Updated Track Convergence Path Predicted Track Convergence Algorithms Cubic Spline Advantage: Smoothest looking convergence Disadvantage: Computationally expensive Updated Track C+1 C Covergence Path T-1 T Predicted Track 19

20 Dead Reckoning - Discussion Advantages Reduces bandwidth requirements because updates are sent less frequently Potentially larger number of players Each host does independent calculations Disadvantages Not all hosts share the identical state about each entity Protocols are more complex to implement to develop, maintain and evaluate Must customize for object behavior to achieve best results Must have convergence to cover prediction errors Collision detection difficult to implement Poor convergence methods lead to jerky movements and distract from immersion Conclusions Shared state maintenance is governed by the Consistency-Throughput Tradeoff Three broad types of maintenance: Centralized/Shared repository Frequent State Regeneration(Blind Broadcast) Dead Reckoning The correct choice relies on balancing many issues including bandwidth, latency, data consistency, reproducibility, and computational complexity 20