Peer-to-Peer Massively Multiplayer Online Games
NICTA has developed network technology to enable Massively Multiplayer Online
(MMO) Applications. Traditional MMO applications are based on a server based
architecture that has several drawbacks.
- High cost of deployment: As the number of users for an MMO application increase, additional servers have to be deployed to manage the increase in demand. This increases the cost of deployment as more and more servers are added to the network.
- Single point of failure: If a server goes down, the entire MMO application goes down. MMO applications therefore have a single point of failure and are heavily dependent on servers that are reliable with minimal downtime.
- Scalability: MMO service providers can dimension their resources (network and server) for a certain number of maximum users. If there is a sudden increase in the number of users, the servers are bound to crash causing disruption to the MMO service.
NICTA's technology is based on a distributed architecture that is server-less. It uses the concepts of peer-to-peer computing to distribute the data processing to the end users. The technology is therefore highly scalable, does not have a single point of failure and is very cost effective (does not involve expensive game servers). The technology comprises of a distributed network engine that interfaces with a game/graphics engine. The technology has been successfully tested on the PlanetLab network (network of more than 600 computers around the world).
Contact: Santosh Kulkani, Aaron Harwood, Egemen Tanin.
CAESAR
Recent research advances in Peer-to-Peer (P2P) computing have enabled the P2P
paradigm to be used for developing complex applications beyond file sharing and
data storage. These applications have drawn significant benefits, specifically
scalability and low cost, from the P2P paradigm. However, the current approach
for designing P2P applications introduce issues that prevent the development of
high quality complex P2P applications. These issues, namely tight coupling to
P2P protocols, limited logic sharing between peers and complicated recovery
processes, motivate us to introduce a service-oriented architecture for P2P
applications. This project aims to design and develop a middleware to support
the development of service-oriented P2P applications applying the principles of
abstraction, dynamic binding, loose coupling and information hiding. We have so
far developed the architecture and the APIs for the middleware, which is called
CAESAR. Currently we are building the CAESAR middleware as an extensible,
component based event driven framework.
Contact: Shanika Karunasekera.
e-Science Desktop Peer
The e-Science Desktop Peer project aims to provide a Peer-to-Peer based
software infrastructure which is tailored for use by communities of scientists.
Our infrastructure is designed to work equally well on commonly available
desktop machines as on clusters and other typical resources used by scientists.
We have researched and developed a Peer-to-Peer based Distributed Virtual
Machine for the purpose of deploying large scale parallel scientific
applications over the Internet.
One of the major challenges in our P2P based approach is that the peers are highly dynamic and they can become unavailable (e.g. due to user specified behavior or machine failure) at any time. This has a negative impact on applications as it interrupts the progress leading to performance penalties. We are examining different methods on handling such peer failures to address the issue.
Contact: Aaron Harwood, Shanika Karunasekera
Intrusion Detection
The increasing number of stealthy and coordinated attacks such as distributed denial of service (DDoS), worms, and coordinated scans pose a significant threat to network security. A major issue for detecting these attacks in a timely and scalable manner is that the evidence of these attacks can be distributed across multiple networks. Collaborative intrusion detection systems (CIDS) have therefore been proposed to address this coordinated defence challenge by correlating suspicious evidence from multiple networks. However, the design of a robust CIDS raises several research challenges, which can be summarized as follows:
- How to remove the need for a central controller in the CIDS.
- How to balance the trade-off between the detection rate and false alarm rate in the CIDS.
- How to improve the scalability of multi-dimensional alert correlation in CIDS.
- How to relieve the load hot spot among participants during worm outbreaks.
We proposed a peer-to-peer based decentralized CIDS to address the first challenge. We presented an optimizing scheme to balance the trade-off between the detection rate and false alarm rate based on an analysis of a large, real-world intrusion dataset. We also proposed a "correlate-and-filter" algorithm to analyse multi-dimensional alerts in a CIDS. We are currently working on a load balancing of CIDS to address the last challenge.
Contact: Chris Leckie, Shanika Karunasekera
VoIP
Voice over Internet Protocol (VoIP) has changed the way people communicate by enabling telephony services on the Internet. The main technical difference between conventional telephony and VoIP is that it uses a packet switched network, the Internet, compared to traditional telephony's circuit switched network. The introduction of VoIP has resulted in a number of VoIP specific protocols that are not interoperable and are not compatible with existing Internet protocols. Client development process for these solutions is harder as developers have to implement specific protocol stacks at the client end. Changes in these architectures require upgrading all the clients which involves significant work and cost.
In this project we concentrate on locating Internet hosts in the wired network. We use a landmark (machines whose physical location is known) based approach by generating a knowledgebase of surrounding nodes through trace route and ping information.
Contact: Shanika Karunasekera