2012 Ninth IEEE International Conference on e-Business Engineering
A Survey of P2P Virtual World Infrastructure
Bingqing Shen, Jingzhi Guo and Philip Chen
Department of Computer Science and Information, Faculty of Science and Technology, University of Macau, Macao
{yb17412, jzguo, philipchen}@umac.mo
huge number of simultaneous online players and uncertainty
of player distribution, virtual world infrastructure
architecture faces challenges again to address hotspot issue,
load balance issue, scalability issue, security issue and the
like. Traditional client-server structure [7] does not scale
well at least in the following three aspects:
1)
Limited number of players in each server prevents
players’ communication across different servers. Even
sharding technique allow more users to play a game
simultaneously, but two players in different shards
walk in two parallel worlds and cannot meet each other.
Hence the scalability is restricted.
2)
Single point of failure risk is unavoidable in
traditional server-client structure. Due to the isolation
nature of sharding, if one server is down even due to
maintenance or service patch release, players on that
server have to wait until the system recovery.
3)
Unbalanced computation resource restricts the
utilization of computation resource. This is a twofold
issue. On the server hand, due to the dynamic
distribution of players, sharding of servers is not the
optimal solution because some servers having fewer
players have redundant resources while others having
more players facing resource shortage. On the other
hand, client machines’ resources are never efficiently
utilized.
Since a single machine cannot meet the requirement of
contemporary virtual world development. Many researchers
turn to peer-to-peer (P2P) [9] architecture for solution.
Naturally P2P network structure has the merits in scalability,
resource efficient utilization and low cost of deployment,
operation, and maintenance. This survey investigates the
core P2P virtual world architecture issues and existing
methodologies. It starts with the introduction of client-server
structured virtual world and the innovation of them to
address the scalability issue. The second section discusses
the P2P virtual world architecture. It is further divided into
three sub-sections, the first part gives some background of
P2P technologies and how they gear into P2P virtual world
architecting; The second one lists the P2P virtual world
platform design requirements as the guidance of problem
solving; Then the third part concentrates on some critical
challenges in the P2P virtual world architecture. Section 3
introduces hybrid solutions combining both client-server and
P2P solution for the complementary of P2P architecture
unanswered questions. Finally, section 5 concludes the
survey.
Abstract—With the development of computer science and
virtual reality technology, virtual world evolves along the way
of computer game development, from arcade games, console
system games, LAN games, Internet connectivity games,
unstructured games, games with player generation of content,
worlds with designer-provided objectives, games with social
networks, and open virtual worlds [1]. Traditional serverclient structure does not scale well at least in the following
three aspects which are limited number of players in each
server, single point of failure risk, and unbalanced
computation resource. This survey investigates another
alternative, peer-to-peer (P2P) virtual world software
infrastructure, to address these traditional architectural issues.
Keywords - Peer-to-peer (P2P), Virtual World, Virtual
Environment, Area of Interest (AoI), Scalability
I. INTRODUCTION
With the development of computer science and virtual
reality technology, virtual world evolves along the way of
computer game development, from arcade games, console
system games, LAN games, Internet connectivity games,
unstructured games, games with player generation of content,
worlds with designer-provided objectives, games with social
networks, and open virtual worlds [1]. This evolution pushes
the improvement of both computing hardware and software
infrastructure, and industry innovation. In the early time, two
friends who want to play games together have to look for a
common place (either in a game pub or at a house of one of
them) to use a single game machine. The birth of Internet
changed the situation. With copper or fiber optic cable, two
remotely located machines are connected logically. Through
LAN, two or more people can share the common screen of
the same game world, like StarCraft [2], Counter Strike [3]
without walking out of their home. Web 2.0 brought social
network to us broadening the relationship among people and
making the whole world connected. Group and guild are new
social elements which add more diversities to traditional
Massive User Dungeon (MUD) games, such as World of
Warcraft [4] and Guild Wars [5]. Guild is more like an ingame community where members contribute to the group
which conversely will benefit each of them. Open virtual
world makes game design more flexible to allow people to
customize virtual products and create their own in-world
game collectively for fun. The representative of open virtual
worlds is Active Worlds [6] and Second Life [7].
Nowadays, millions of people spend significant time on
massively multiplayer online role-playing games
(MMORPG) or open virtual worlds each day. Due to the
978-0-7695-4809-8/12 $26.00 © 2012 IEEE
DOI 10.1109/ICEBE.2012.55
296
II. TRADITIONAL VIRTUAL WORLD STRUCTURE
Service Nodes are highly extendible to enrich a virtual world
by adding more sub-modules.
Traditional distributed virtual world was born from US
military simulation system along the history line of DIS [10],
SIMNET [10], HLA [11]. In traditional virtual world design,
to balance server loads and improve system scalability,
different system functions are partitioned into different
domains (sometimes called service). In such architecture,
one server or a cluster only responsible for one specific type
of service and then each kind of service is easy to be scaled
simply by adding more servers. Three typical structures are
introduced in the following content in this section.
Second Life architecture [12] proposed from Linden Lab
is partitioned into agent domain, region domain, and viewer
domain. Viewer is a client software handling locations of
objects, getting velocities and other physics information, and
doing simple physics track of what is moving to where. An
Agent Domain [13] handles everything that has to do with
agents. It stores user profile data, user inventory, handles
login, etc. A Region domain [14] handles everything related
to regions, including land information, parcel information,
avatar positions, object information (metadata, map
information, region statistics), physics, and script execution.
The region and agent domain comprise three parts
respectively which are service, host and data store. Each part
can be hosted by one server or a cluster and linearly
extensible.
OpenSimulator [15] is an open-source project providing
Second-Life-Styled server side software. It separates the
service system into five groups [16], which are user service,
grid service, asset service, inventory service and message
service. User Service is responsible for authenticating the
user to the grid. It creates a session identifier for the client
which can be used to authenticate requests to the other
servers in the same grid. Grid Service maintains avatar or
entity geographic location in a region. Asset Service is a data
service storing virtual entity data in the virtual environment,
including attached sounds, textures, images, notecards,
scripts, etc. Inventory Service is another data service
responsible for maintaining the hierarchical structure of
reference to the entity of asset. Once an asset is created, the
inventory server keep track of where it is put. Messaging
Service is responsible for establishing text or voice
communication in the virtual environment.
Open Wonderland [17], initiate by Sun Microsystem Co.
is another framework for building 3D virtual worlds. Open
Wonderland framework consists three major modules: Client,
Web Server and Service Nodes [18]. Most of the visual logic
run on Client – the browser through which user can visually
navigate a virtual world. It contains functions of graphic
rendering, object collision detection and physics control. The
Web Server acts as a central management console, providing
web-based management of all services in the system,
regardless of what server they are running on. Core
Wonderland features operated on the Web server including
authentication, asset management, single sign-on (SSO)
service. Service Nodes provide world state management and
other value-added service such as voice communication,
video conference call and other in-world application support.
III. PEER-TO-PEER VIRTUAL WORLD ARCHITECTURE
P2P virtual world architecture is still a leading research
area. From the very beginning of 20 century, the wave of
P2P file sharing and social virtual world application pushes
the combination of them to overcome issues in traditional
system design. Based on sophisticated P2P underlying
communication theories and implementation, virtual world
architecting becomes a new branch of the P2P research
family. Currently issue of P2P virtual world design mainly
focuses on world state (both player incarnation state and
object state) management, including state partitioning, state
persistency and state consistency, since none of a peer has
the global information of the whole world. This section starts
introduce the P2P technology itself.
A. Peer-to-peer Technology Background
P2P technology has its wide range of research and
application in the last decade, including file distribution
(such as Gnutella [19], Napster [20], BitTorrent [21], etc.)
and video streaming (such as CoolStreaming [22], PPStream
[23], SopCast [24], etc.). Besides, some Web caching
systems also designed based on P2P manner [25] to scatter
server load to each peer and improve the overall
responsiveness. Given that of each peer node in a P2P
network has three traits: 1) unfixed role playing – a peer can
be a client, server, and router even at the same time; 2)
dynamics – a peer is free of joining, leaving the network and
changing of its computation capacity and 3) anonymous
identity – a peer’s trustworthiness and purpose are unknown,
a P2P system design have a strong consideration of tradeoff
in three aspects: 1) network structure and topology control, 2)
incentive management, 3) trust and security management.
Network structure and topology control research how
peers in a network are efficiently organized and the scientific
management of peer dynamics, including peer addition,
leaving and failure. Prevailing mainstreams of design based
on 1) unstructured topology or 2) structured topology.
Structured topology works with dynamic hash table (DHT)
to match resource identity with peer identity (or node
identity), which are centrally assigned and managed. Noted
structured topology includes Chord [26], CAN [27], Tapestry
[28] and Pastry [29]. Unstructured topology [30] does not
have such structure to organize resources. Instead, it relies on
flooding query to search a particular resource with the help
of a tracker. BitTorrent [21] is the famous example of
unstructured topology application. Resource in a structured
topology is easier to be located and found due to the avail of
global information. But it drops some extent of scalability
and autonomy, which becomes the advantages of
unstructured topology. Unstructured P2P topology has many
other nice characters, e.g. low network diameter and, more
importantly, robustness against peer dynamics and random
node failures. However, the lack of structure also makes it
hard to accurately locate data or peers. Thus many
researchers set out to design hybrid network structures to
combine the best of both worlds. [31] [32]
297
For incentive management, altruist and selfishness are a
group of contradiction on resource holding and sharing.
From a single peer view, it is always trying to grab as much
resource as possible from other people which providing as
few as possible to save its own computation resource (hard
disc, bandwidth, CPU cycle, etc.). Because according to
game theory, without knowing other’s information, one
always makes a suboptimal decision to protect himself. This
is where the incentive management comes from. Since
altruist from everyone is just a utopia in peer-to-peer
resource sharing scope, a sub-optimal solution is to maintain
a healthy and sustainable peer-to-peer network by adopting
incentive policies. The main idea is the more resource a peer
provides to others, the higher priority it should gain from
other peer’s support. Related approaches includes paymentbased strategy [33], reciprocity and reputation-based strategy
[34][35], penalty-based strategy [36], game theory-based
strategy [37], and auction-based strategy [38].
Trust management and security management work on
how to establish a trustworthy and secure environment for
peer-to-peer transaction by preventing technical attack and
fraud. Security and trustworthiness are two vital
characteristics in peer-to-peer marketplace building.
However, trust means to expose more user information and
drop some privacy requirement. Several trust models were
established including EigenTrust [39], Peer Trust [40],
FuzzyTrust [41], SuperTrust [42], PowerTrust [43] and
GossipTrust [44]. Azzedin and Maheswaran [45] suggested a
practical trust computation and management system that is
suitable for P2P computing environment and widely cited. In
the security domain, P2P systems are naturally vulnerable
from distributed denial-of-service (DDoS) attack. Yunhao
Liu et al. [46] propose a distributed and scalable scheme,
called DD-POLICE to detect malicious nodes to defend P2P
system from overlay flooding-based DDoS attack.
Brinkmeier et al. [47] proposed several heuristic techniques
to make P2P streaming systems more attack resilient from
DDoS attack. Other security issues include buffer map
cheating [48], Sybil attack [49], and P2P worm propagation
[50].
To glue these P2P related technologies, some framework
and platform design has been proposed for application
developers’ handy use. JXTA [77] prevailing in industries
was proposed by Sun Microsystems Inc. is a set of
specifications regulate standard API for P2P message
transport, peer resource management, group management
and security scheme. Bottom up, it has three layers [51]:
JXTA core layer, JXTA services layer and JXTA application
layer. Core layer provides primitive communication
mechanism and group management. Service layer
modularize functional components for application utilizing.
Figure 1 shows the JXTA layer structure. JXTA does not
regulate the topology structure implementation, while P2PS
Mozart [52], another peer-to-peer system platform, provides
a set of API to separate the implement of underlying network
topology and services above it. P2Ps Mozart’s algorithms
offer strong data lookup guarantees based on structured
topology Tango. Through API definition, it can also be
adapted to other structured topology such as Chord. Based
on the Mozart virtual machine, P2PS Mozart’s structure is
similar as JXTA. Other research projects like Chord project
[53] and FreePastry [54] are similar as P2PS Mozart, which
are just different from programming languages and the
network topology structure selection. These platforms all aim
at providing a solution to build a scalable, robust distributed
application.
Figure 1. JXTA Architecture
B. Requirement of Peer-to-Peer Virtual World Design
Many P2P virtual world architectures contain some
requirements of their designers for issues addressing. Doing
a comparison among them, J. Chen et al’s [55] requirement
list is relatively complete. It summarized a number of issues
involved in designing a networked virtual environment. They
are:
1)
Consistency: Object states and event orderings
received by each participant should be consistent;
2)
Responsiveness: The delay of user action and
observable consequence should be minimized despite
of network latency and heterogeneity of different peer
computing capability;
3)
Scalability: The number of peers participating the
game should be easily added without impacting the
network performance;
4)
Persistency: Contents within NVEs (e.g., virtual
objects and computer characters) may need to exist
persistently to provide a sense of realism;
5)
Reliability: System should be robust and resistant from
hardware or software failures;
6)
Security: Account theft and privacy exposure issues
should be addressed.
J. Chen et al’s [55] design based on the VON diagram
addressed first three of them. Other designers also raised
some requirements in their architectures. T. Hampel et al’s
[56] design focused on scalability, persistency, besides
adding the flexibility requirement. D. Ahmed et al [57-61]
work on consistency, responsiveness, reliability, persistency
and security. B. Knutsson, et al.’s [62] research addressed
the responsiveness, reliability and security issue. S. Shen did
a thorough investigation in the existing research of P2P game
design. He divided the research issue into several categories
which can be summarized to: 1) state management, 2)
overlay management, 3) content management, and 4)
security. The first three issues are related to the system
298
failover solution to address the server crash issue and how
resources should be handed over to other hosts when failure
occurs.
B. Bossche et al.’s [64] study of dynamic microcell
relocations proposes three schemes to optimize microcell
distribution and relocation, based on players’ walking
itineraries, border crossing rate and load distribution. These
schemes should be iteratively executed to reach the final
balance state.
Another approach is area-of-interest-based (AoI-based)
dynamic partition, which is totally different from the regionbased partition. Due to a player’s limited view of the
surrounding, he/she only needs to exchange state change
information with other players located nearby. Actually,
proximity is not accurate because interest area is not only
bounded with geographic distance. It will be discussed
specifically in the below section.
requirements of consistency, scalability and persistency.
Some of his work will be referenced in this survey.
C. State Management
State management is the critical part of peer-to-peer
virtual world design since it directly determines the system’s
consistency, scalability, responsiveness, and failover
attributes. This section of survey renders the state
management in three sub-sections which are state
partitioning, state persistency and state consistency. Area of
Interest (AoI) management as a supplementary section to the
above three aspects will be elaborated at last, because this
technique is widely employed to address issues in all the
three aspects.
1) State Partitioning
To overcome overload on central server system,
researchers and game designers suggest load of object states
should be shared by all participating peers. How to
effectively divide and manage different portion of world
state leads to different design structures. B. Knutsson et al.’s
architecture SimMub [62] uses a static partition approach.
Geographically neighboring players and objects are grouped
into regions. Each region assigned an ID will be run by a
host (a peer) and responsible for maintaining the region state
among all the regional players. But SimMub architecture did
not solve hotspot issue – If many players crowd into one
region (let’s say for an in-world event,), that host will
become overwhelmed with computing resources.
To address the hotspot issue, D. Ahmed et al. presented a
dynamic partition model that further divides each region into
smaller areas (called microcells) [57] [59] [60]. And each
server (each peer node playing the role of both server and
client) is assigned with one or several pieces of microcells.
Once a region becomes a hotpot, the management of some
microcells will be transferred to other less-crowded hosts for
load balance. As shown in figure 2, Server 1 hosts cells (0-0,
1-0, 1-1), Server 2 hosts cells (0-1, 1-2, 2-3) ... and Server n
hosts cells (4-1, 3-2, 4-3). When, let’s say, Server 1 is
overwhelmed with crowds, and Server 2 has some surplus
computing resource, cell (1-1) will be shifted from Server 1
to Server 2.
2) State Persistency
State persistency is important because if there is no such
mechanism, when a player shut down his / her machine, all
the data running will be lost and that will cause partial game
crash. However, due to the best of our knowledge, only a
few researches were invested on game state persistency
using P2P technique. In M. Kim, et al.’s paper “P2P Second
Life: Experimental Validation Using Kad” [66], state
persistency holds if no objects gets lost during the evolution
of the virtual world. Research by T. Limura et al. [67]
worked out a zoned federation P2P model based on the
region division approach in which peers are alternately
chosen to be the master of a region. The zoned federation
model decoupled the game data into transient data (such as
player location) and persistent data (such as player’s in-game
inventory). The copy of persistent state will be replicated to
each peer selected as a zone master. K. Zhang et al.’s [68]
design of Mammoth stores persistent data in a relational
database. In Mammoth, different objects has different
persistency requirement and they are stored differently.
Combining with Mammoth, the zonal federation approach
can offer a better persistency solution.
3) State Consistency
Game world state must be consistent among players of a
certain range (geographical range or AoI). For example, a
bullet fired by player A must be received by the target player
B and witnessed by nearby players C and D instantaneously.
B. Knustsson et al. [62] divided game state into 1) player
state and 2) object state based on different consistency
requirement.
Player state update is accessed in a single-writer,
multiple-reader pattern. In the paper “A hybrid P2P
communication architecture for zonal MMOGs”, D.T.
Ahmed et al.’s architecture [57][59][60] combines P2P and
multicast techniques to support player state exchange and
meantime minimize the range of data transmission. In the
hybrid model, each zone has one master node which is the
coordinator for intra- and inter-zone communication of that
zone. In the operation, all nodes within the same zone
register their information to the master node. To lower down
Figure 2. Microcell – Server Pool Mapping [59]
G. Chandana et al.’s [63] overlapping, zone-based
architecture is similar as D. Ahmed et al’s by dividing a
region into smaller pieces and adding redundant computing
resources to manage each of them. Chandana also proposed a
299
some burden from master nodes, D.T. Ahmed et al.’s model
uses dominating set and coloring technique to construct the
multicasting mesh among all the non-master nodes. Interzone communication occurs when a player approaching the
boundary of two zones. To prevent crossing zone messages
flooding to all the players in the foreign zone, area of interest
is again used to restrict the range of state message
broadcasting.
Object state update, however, is more complicated.
Kunstsson [62] pointed out because of the collaboration on
the same object from multiple players, the final state relies
on the result of the common actions. Kunstsson [62]
suggested assigning a coordinator in charge of collecting all
the information of actions on the object, computing the result
and sending it to all the receivers. Then the question
becomes how the coordinator system is established free from
single point of failure problem and system performance
impact.
To prevent plethoric inter-zone communication and keep
the continuous view when a player hanging around the
border, J. Oliveira et al. [69] suggest an architecture in which
part of the foreign zone scene can be embedded to the local
zone. In Oliveira’s architecture, foreign zone objects become
ghosts in the local zone and their states will be synchronized
among all the zones holding this part of scene. Inter-zone
communication will only occur when a player really jumps
(teleports) to the foreign zone.
not necessarily symmetric; 3) Domain of interest is not
always geographically divided though it is the truth in most
cases.
However, L. Katherine’s interest management theory is
based on client-server structure while peer-to-peer structure
double the complexity. A. Yu et. al summarized some
research efforts in this area in [72]. These efforts can be
classified into two categories in terms of P2P topology: DHT
and unstructured P2P approaches. DHT-based approaches
utilize network structure to facilitate the dissemination of
message by introducing multicast layer or coordinator. In
unstructured approaches, entities maintain neighbors’ states
more autonomously. Three proposals of unstructured
approaches were mainly discussed. They employ number of
hops in P2P network or geographical distance in the virtual
world to derive peer’s AoI. VON P2P network architecture is
the famous one of them. J. Chen [55] and S. Hu et al. [73]
give the details of how to apply Voronoi diagram in P2P area
of interest management. In J. Chen design, it defines a set of
concepts (illustrated in figure 3 (b)) regarding the AoI first,
including:
x
AoI Neighbor: the nodes whose positions are
within its AoI;
x
Boundary Neighbor: nodes whose regions
immediately surround the given node;
x
Enclosing Neighbor: AoI neighbors whose
enclosing neighbors may partially lie outside the
AoI.
In the publication “VON: A Scalable Peer-to-Peer
Network for Virtual Environments” [55], J. Chen, et al.
bounded player’s AoI with voronoi diagram to decompose a
space. Figure 3 (a) illustrates a space decomposition based
on Voronoi diagram. Within boundaries of a region, all the
distances to the bold dot are shorter than to the bold dots
outside the region. Voronoi diagram spatially optimize the
range and shape of AoI within which messages can be
efficiently broadcasted to limited listeners. S. Hu et al. in [73]
elaborated the procedure of applying Voronoi diagram in a
P2P networked environment, discussing the player joining,
moving, leaving actions adapted in a Voronoi-based AoI
division.
D. AOI
Area of Interest is important, especially in large-scale
distributed virtual worlds which contains thousands of
entities interacting with each other, because it can restrict the
range of information exchange, thus resulting in high
responsiveness and scalability. D. Ahmed et al.’s publication
“Improving gaming experience in zonal MMOG” [60],
defined AoI “the visibility scope of player remains inside the
zone boundary”. K. Edward et al. [70] interpreted AoI “the
radius a user is able to view and directly interact with”. J.
Chen et al.’s architecture [55] also employed the AoI concept
as a circular-shaped area of which the host player is always
located in the center and within which other players should
be visible and exchange state information with the host
player. In S. Shen’s survey [65], AoI is defined “Due to
limitation of human perception, player will only interest with
a portion of the game world around his controlled-object.
The portion of the game world is called Area of Interest
(AoI).”
L. Katherine gave a framework of interest management
scheme presented in [71]. In [71], a set of related concepts
have been systematically defined. Interest management
studies entity’s data interests in terms of location and other
application-specific attributes. In interest management,
interest expression (IE) is a specification of the data on entity
needs to receive from others to maintain state consistency.
Entities’ Interest data is transmitted by interest manager (IM)
within a specified the domain of responsibility (DoR) or
domain of interest (DoI). L. Katherine made observation on
seven virtual environment simulation systems which implied
1) interest expressions is always self-referential; 2) interest is
Figure 3. Voronoi diagram [x]: a) dots indicate sites, and lines define
boundaries (edges) for regions; b) the large circle is the AOI boundary for
the center node. Squares (s) are enclosing neighbors; triangles (v) are
boundary neighbors; stars (5) are both enclosing and boundary neighbors;
circle (q) represents a regular AOI neighbor; crosses (×) represent
neighbors irrelevant (i.e., outside of AOI) to the center node. [55]
300
cell state update incorporating with server daemon threads
(central cell-daemons, or CCDs). Some of these clients
picked to play some of the server roles are called master cell
daemon (MCD) and each of them manages one of the cells.
And other qualified nodes in the same cell are the backups of
master nodes, called secondary cell daemon (SCD).
C. Bezerra et al. proposed a virtual world region partition
but on server side instead of client side based on the K-d tree
division of virtual world regions. In A. Marios et al.’s design
[76], server-side manages microcells dividing the world
regions while client side peers compute the AoI and maintain
object state in corporate with each grid server. C. Chris et al.
[75] expanded existing virtual world framework Homura and
Net Homura by adding peer-to-peer layers both on client and
server side. Many other massively multiplayer online roleplay games (MMORPG) introduce transfer the subsidiary
functions such as multimedia chatting data in peer-to-peer
manners to lower down server balance.
As aforementioned, L. Katherine pointed out entity
interest could be a multi-dimensional domain other than only
geographical location [71]. K. Edward, in the design of
VELVET [70], used a set of metrics to regulate the scope of
AoI. For instance, a set of metrics are selected MS = {M1,
M2, ... , Mn}. For example, Metric 1 (M1) can be
geographical distance; Metric 2 (M2) can be upload
bandwidth, and the like. Objects only meeting the all the
requirements will be visible inside the AoI. The image of the
world captured in each player’s AoI is called parallel virtual
world (PVW). To prevent frequently change of
communication list in one player’s AoI due to, for example,
objects’ movement or bandwidth change, VELVET regulates
an upper bound and lower bound value for each metric.
When an object is moving (using geographic distance as the
only metric) toward the direction leaving the AoI from inside,
it will only move out after it crosses the upper bound. When
an object is moving towards the direction entering the AoI
from outside, it will go inside only when it crosses the the
lower bound. To visualize these two movements, in figure 4,
each of them is represented with two rings with different
radius. The outer one represents the upper bound and the
inner ring is the lower one. In figure 8, an object is moving
along the border of AoI. The movement trajectory is
highlighted with a squiggly curve, and the crossings of the
border indicated with arrows and numbers. It can be
observed that with two borders (figure 4 (B)), the number of
crossing is reduced comparing with the single border case
(figure 4 (A)), and it becomes less when the distance of two
borders increases (figure 4 (C)).
V. CONCLUSION
This paper started reviewing the virtual world system
development history and issues of existing architectures due
to the intrinsic features of server-client technologies. Then it
introduced some approaches of existing efforts to overcome
some of the issues with three cutting-edge examples. Peer-topeer system, naturally with strong scalability, as another
alternative is mainly discussed and anticipated to be the
substitution of client-server architecture in future virtual
server design. The critical issue in P2P virtual world system
is to maintain world state consistency, persistency and
system load balance. They are addressed through world
resource proper division and management of each player’s
area of interest. The aim is to maximize each peer node’s
resource utilization while minimize the communication
overhead between them. Some hybrid models are also
investigated for reference to solving P2P hard nuts.
Finally, P2P virtual world is different from P2P file
sharing and video streaming due to its high state consistency
and constantly real-time update requirements. From This
survey, we draw some research issue of virtual world
infrastructure, which includes but not limited:
x
The network topology best fitting the client dynamics
requirement in virtual world application, structured
topology, unstructured topology or even hybrid one.
x
The most efficient strategy for world state
dissemination, which should support maximizing the
responsiveness while minimizing the communication
overheads.
x
The best strategy for world content distribution to
prevent content lost in case of retreat or failure in
single peer.
x
Performance metrics to evaluate the architecture and
operation, including in AoI, the attributes based on
which an entity’s interest data will be filtered.
x
The tradeoff between pure P2P structure or hybrid
structure which mitigate but not eliminate single point
of failure risk at server side.
Figure 4. Visualization of VELVET’s AoI boundaries. [70]
IV. HYBRID VIRTUAL WORLD STRUCTURE
Due to the lack of central repository of persistent data,
there are some critiques on peer-to-peer-based virtual world
design. Comparing with traditional server-client structure, at
least three drawback of can be concluded from existing peerto-peer virtual world design: Lack of central administration
makes state consistent hard to maintain globally; Lack of
central organization increase the difficulty to globally
optimize the computation resource arrangement from
heterogeneous capacity and bandwidth of peers; High
dynamics of peer causes high communication overhead.
Therefore, we can found some researchers dedicating to mix
two worlds together and build a hybrid virtual world
architecture. Based on the traditional server-client structure,
K. Kim, et al. [74] introduce cell-daemon-ation concept to
reduce the overall server-side bandwidth consumption. In his
demo system HYMS, a virtual world is divided into several
cells. In each cell, some qualified client nodes manages in-
301
x
[29] A. Rowstron, P. Druschel, "Pastry: Scalable, Decentralized Object
Location, and Routing for Large-Scale Peer-to-Peer Systems," In
Middleware 2001. Springer: Berlin/ Heidelberg, 2001, pp. 329–350.
[30] Qin Lv, Pei Cao, Edith Cohen, Kai Li, Scott Shenker, "Search and
replication in unstructured peer-to-peer networks," Proceedings of the
16th international conference on Supercomputing, 2002, pp. 84-95.
[31] K. Ohnishi, "Autonomously Reconstructable Semi-Structured P2P
Networks for File Sharing," International Conference on Autonomic
and Autonomous Systems, 19-25 June 2007.
[32] B. Lagesse, "UBCA: Utility-Based Clustering Architecture for Peerto-Peer Systems," International Conference on Distributed Computing
Systems Workshops, 22-29 June 2007.
[33] D. Hausheer, "Token-based accounting and distributed pricing to
introduce market mechanisms in a peer-to-peer file sharing scenario,"
Third International Conference on Peer-to-Peer Computing, pp. 200201, 1-3 Sept. 2003.
[34] Michal Feldman, Kevin Lai, Ion Stoica, John Chuang, "Robust
incentive techniques for peer-to-peer networks," Proceedings of the
5th ACM conference on Electronic commerce, ACM New York, pp.
102-111 2004.
[35] Q.Sun, H. Garcia-Molina, "SLIC: a selfish link-based incentive
mechanism for unstructured peer-to-peer networks," Proceedings of
the 24th International Conference on Distributed Computing Systems,
2004, pp. 506-515.
[36] M. Feldman, C. Papadimitriou, J. Chuang, I. Stoica, "Free-riding and
whitewashing in peer-to-peer systems," IEEE Selected Areas in
Communications, pp. 1010-1019, May 2006.
[37] K. Ranganathan , M. Ripeanu , A. Sarin , I. Foster, "To Share or not
to Share: An Analysis of Incentives to Contribute in Collaborative
File Sharing Environments," In Workshop on Economics of Peer-toPeer Systems, 2003.
[38] R. Gupta, A.K. Somani, "Pricing strategy for incentivizing selfish
nodes to share resources in peer-to-peer (P2P) networks,"
Proceedings of the 12th IEEE International Conference, 2004.
[39] S. Kamvar, M. Schlosser, H. Garcia-Molina, "The Eigentrust
algorithm for reputation management in P2P networks," Proceedings
of the 12th international conference on World Wide Web, pp. 640 651, 2003.
[40] Li Xiong , "PeerTrust: supporting reputation-based trust for peer-topeer electronic communities," IEEE Transactions on Knowledge and
Data Engineering, pp. 843-857, July 2004.
[41] S. Song, K. Hwang, R. Zhou,Y.-K. Kwok, "Trusted P2P transactions
with fuzzy reputation aggregation," IEEE Internet Computing, pp. 2434, Nov.-Dec. 2005.
[42] T. Dimitriou, G. Karame, I. Christou, "SuperTrust – A Secure and
Efficient Framework for Handling Trust in Super Peer Networks," In
Proceedings PODC, 2007.
[43] A.G.P. Rahbar, O. Yang, "PowerTrust: A Robust and Scalable
Reputation System for Trusted Peer-to-Peer Computing," IEEE
Transactions on Parallel and Distributed Systems, pp. 460-473, 2007.
[44] R. Zhou, K. Hwang, "Gossip-based Reputation Aggregation for
Unstructured Peer-to-Peer Networks," IEEE International Parallel and
Distributed Processing Symposium, pp. 1-10, 2007.
[45] F. Azzedin, "Trust modeling for peer-to-peer based computing
systems," Proceedings of the international symposium on parallel and
distributed processing, 22-26 April 2003.
[46] Y. Liu, Xiaomei Liu ; C. Wang ; L. Xiao, "Defending P2Ps from
Overlay Flooding-based DDoS," International Conference on Parallel
Processing, 2007.
[47] M. Brinkmeier, G. Schafer,T. Strufe, "Optimally DoS Resistant P2P
Topologies for Live Multimedia Streaming," IEEE Transactions on
Parallel and Distributed Systems, pp. 831-844, June 2009.
[48] D. Li, J. Wu, Y, Cui, "Defending Against Buffer Map Cheating in
DONet-Like P2P Streaming," IEEE Transactions on Multimedia, pp.
535-542, 2009.
Security issue in P2P virtual world operation.
ACKNOWLEDGEMENT
This research is partially supported by the University of
Macau Research Grant No. MYRG156(Y2-L2)-FST11-GJZ.
REFERENCE
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
B. Messinger, E. Stroulia, K. Lyons, "A typology of virtual worlds:
historical overview and future directions," Journal of Virtual Worlds
Research - Typology of Virtual Worlds, 2008, pp. 1-18.
StarCraft. http://us.battle.net/sc2/
CounterStrike, http://en.wikipedia.org/wiki/Counter-Strike.
World of Warcraft. http://us.battle.net/wow/en.
Guild Wars. http://www.guildwars.com.
Active Worlds. http://www.activeworlds.com.
Second Life. http://secondlife.com.
Client-server model
http://en.wikipedia.org/wiki/Client%E2%80%93server_model
Peer-to-peer, http://en.wikipedia.org/wiki/Peer-to-peer
D. Fuilford, "Distributed interactive simulation: It'b{s Past, Present,
and Future." Proceedings of the 1996 Winter Simulation Conference,
1996.
J. Dahmann, M. Morse, "High level architecture for simulation: an
update," Proceedings of the 2nd International Workshop on
Distributed Interactive Simulation and Real-Time Applications, 1995.
Structural
Design
Overview.
http://wiki.secondlife.com/wiki/Structural_Design_Overview.
Second
Life
Agent
Domain.
http://wiki.secondlife.com/wiki/Agent_Domain.
Second
Life
Region
Domain.
http://wiki.secondlife.com/wiki/Region_Domain.
Open Simulator. http://www.opensimulator.org.
OpenSim:
Introduction
and
Definitions,
http://opensimulator.org/wiki/OpenSim:Introduction_and_Definitions.
Open Wonderland. http://openwonderland.org.
J. Kaplan, N. Yankelovich, "Open Wonderland: An extensible virtual
world architecture," IEEE, Internet Computing, pp. 38-45, Sept.-Oct.
2011.
Gnutella
Protocol
Development
(2009).
http://rfcgnutella.sourceforge.net.
Napster. http://free.napster.com.
BitTorrent. http://www.bittorrent.com.
CoolStreaming. http://webtv.coolstreaming.us
PPStream. http://www.ppstream.com
SopCast. http://www.sopcast.org/
A. Rowstron, P. Druschel, "Squirrel: A decentralized peer-to-peer
web cache," roceedings of the twenty-first annual symposium on
Principles of distributed computing, ACM New York, 2002, pp. 213222.
I. Stoica, R. Morris, D. Karger, M. Kaashoek, H. Balakrishnan,
"Chord: A scalable peer-to-peer lookup service for internet
applications," Proceedings of the 2001 conference on Applications,
technologies, architectures, and protocols for computer
communications, ACM New York, 2001, pp 149-160
S. Ratnasamy, P. Francis, M. Handley, R. Karp, S. Shenker, "A
scalable content-addressable network," Proceedings of the 2001
conference on Applications, technologies, architectures, and protocols
for computer communications, ACM New York, 2001, pp. 161-172.
B. Zhao, "Tapestry: a resilient global-scale overlay for service
deployment," IEEE Selected Areas in Communications, pp. 41-53,
Jan. 2004.
302
[49] J. Dinger, H. Hartenstein, "Defending the Sybil attack in P2P
networks: taxonomy, challenges, and a proposal for self-registration,"
The First International Conference on Availability, Reliability and
Security, 20-22 April 2006
[50] L. Xie, S. Zhu "A Feasibility Study on Defending Against Ultra-Fast
TopologicalWorms," Seventh IEEE International Conference on
Peer-to-Peer Computing, pp. 61-70, 2-5 Sept. 2007.
[51] L. Gong, "JXTA: A Network Programming Environment," IEEE
Internet Computing, June 2001.
[52] V. Mesaros, B. Carton, P. Roy, "P2PS: Peer-to-Peer Development
Platform for Mozart," Springer-Verlag Berlin Heidelberg, pp. 125136, 2005 .
[53] Chord Project. http://www.pdos.lcs.mit.edu/chord.
[54] FreePastry. http://freepastry.rice.edu.
[55] J. Chen, T. Chen, "VON: A Scalable Peer-to-Peer Network for
Virtual Environments," IEEE Network. 2006.
[56] T. Hampel, T. Bopp, R. Hinn, "A peer-to-peer architecture for
massive multiplayer online games," Proceedings of 5th ACM
SIGCOMM workshop on Network and system support for games,
2006.
[57] D. Ahmed, S. Shirmohammadi, J. Oliveira, "A hybrid P2P
communication architecture for zonal MMOGs," Proceedings of the
15th international conference on Multimedia, 2007.
[58] D. Ahmed, "Fault Tolerance Procedure For P2P Online GamesS,"
10th International Conference on Information Sciences Signal
Processing and their Applications (ISSPA), 2010.
[59] D. Ahmed, S. Shirmohammadi, "A Microcell Oriented Load
Balancing Model for Collaborative Virtual Environments," IEEE
Conference on Virtual Environments, Human-Computer Interfaces
and Measurement Systems, 2008.
[60] D. Ahmed, S. Shirmohammadi, J. Oliveira, "Improving gaming
experience in zonal MMOGs," Proceedings of the 15th international
conference on Multimedia, 2007.
[61] D. Ahmed, S. Shirmohammadi, "A Framework For Provisioning
Overlay Network Based Multimedia," IEEE International Conference
on Multimedia and Expo. 2007
[62] B. Knutsson, H. Lu, W. Xu, B. Hopkins, "Peer-to-Peer Support for
Massively Multiplayer Games," INFOCOM 2004, Twenty-third
AnnualJoint Conference of the IEEE Computer and Communications
Societies, 7-11 March 2004.
[63] [63] G. Chandana, R. Wiegand, G. Brian, D. Troy, "An Architecture
Supporting Large Scale MMOGs," Proceedings of the 3rd
International ICST Conference on Simulation Tools and Techniques,
2010.
[64] B. Bossche, B. Vleeschauwer, F. Turck, B. Dhoedt, P. Demeester,
"Design of distributed microcell-based MMOG hosting platforms:
impact study of dynamic relocations," Cluster Computing Volume 14,
pp. 145-163, 2011.
[65] S. Shen, "Survey of P2P Game," Parallel and Distributed Systems
Group, Delft University of Technology, 2010.
[66] M. Varvello, C. Diout, E. Biersack, "P2P Second Life: Experimental
Validation Using Kad," IEEE NFOCOM, 2009, pp. 1161-1169.
[67] T. Iimura, H. Hazeyama, and Y. Kadobayashi, "Zoned federation of
game servers: a peer-to-peer approach to scalable multi-player online
games," NetGames, ACM, 2004, pp. 116–120.
[68] K. Zhang, B. Kemme, and A. Denault, "Persistence in massively
multiplayer online games," in NetGames. ACM Press, 2008.
[69] J. Oliveira, S. Yu, N. Georganas, "Synchronized World Embedding in
Virtual Environments," IEEE Computer Graphics and Applications,
July-Aug 2004.
[70] K. Edward, A. Room, N. Georganas, "VELVET: An Adaptive Hybrid
Architecture for VEry Large Virtual EnvironmenTs," Presence:
Teleoperators and Virtual Environments, Vol. 12, 2003, pp. 555-580.
[71] K. Morse, L. Bic, M. Dillencourt, "Interest Management in LargeScale Distributed Simulations," Presence: Teleoperators and Virtual
Environments, pp. 52-68, February 2000.
[72] A. Yu, S. Vuong, "MOPAR : A Mobile Peer-to-Peer Overlay
Architecture for Interest Management of Massively Multiplayer
Online Games," Proceedings of the international workshop on
Network and operating systems support for digital audio and video,
2005, pp. 99-104.
[73] S. Hu, G Liao, "Scalable Peer-to-Peer Networked Virtual
Environment," Proceedings of 3rd ACM SIGCOMM workshop on
Network and system support for games, 2004, pp. 129-133.
[74] K. Kim, I. Yeom, J. Lee, "HYMS: A Hybrid MMOG Server
Architecture," EICE TRANSACTIONS on Information and Systems,
pp. 2706-2713, December 2004.
[75] C. Carter, A. Rhalibi, M. Merabti, A. Bendiab, "Hybrid Client-Server,
Peer-To-Peer Framework For MMOG," 2010 IEEE International
Conference on Multimedia and Expo (ICME), pp. 1558-1563, July
2010.
[76] M. Assiotis, V. Tzanov, "A Distributed Architecture for MMORPG,"
Proceedings of 5th ACM SIGCOMM workshop on Network and
system support for games, 2006.
[77] JXTA Technology. http://java.sun.com/othertech/jxta/index.jsp.
[78] C. Bezerra, J. Comba, C. Geyer, "A Fine Granularity Load Balancing
Technique for MMOG Servers Using a KD-Tree to Partition the
Space," 2009 VIII Brazilian Symposium on Games and Digital
Entertainment, 2009, pp. 17-26.
303