Welcome to my personal webpage
Here you will find every information you may need to know about me. Have a nice browsing experience.
Some brief information about me
I'm a PhD student working at LIP6 in the field of massively distributed and scalable services.
Thesis Abstract:
Massively Multi-player Online Games (MMOGs) aim at gathering an infinite number
of players within the same virtual universe. Yet all existing MMOGs rely on centralized
client/server architectures which impose
a limit on the maximum number of players (avatars) and resources that can coexist
in any given virtual universe. This thesis aims at proposing solutions
to improve the scalability of MMOGs.
There are many variants of MMOGs, like role playing games (MMORPGs), first
person shooters (MMOFPSs), and battle arenas (MOBAs), each with a specific set of
design concerns: low latency, artificial intelligence (AI) for the control of characters,
expandable universe, so on and so forth.
To address the wide variety of their concerns, MMOGs rely on independent services
such as virtual world hosting, avatar storage, matchmaking, cheat detection, and game design.
This thesis explores two services that are crucial to all MMOG variants: matchmaking and
cheat detection. Both services are known bottlenecks, and yet current implementations
remain centralized.
Matchmaking allows players to form teams and to find opponents. Current service implementations
raise important issues regarding player experience. They often lead to mismatches where strong
players face weak ones, and response times can be exceedingly long. This thesis conducts an
extensive analysis over a large dataset of user traces in order to propose a set of
matchmaking solutions that scale with the number of clients. Interestingly, some of these solutions are
centralized and still deliver excellent service.
Current refereeing services for MMOGs also rely on centralized architectures, which limits both
the size of the virtual world and the number of players.
Conversely to matchmaking, cheat detection requires a high frequency of computations
per player: therefore a distributed solution seems best for scalability purposes.
This thesis shows that it is possible to design a peer to peer refereeing service on top of a reputation
system.
The resulting service remains highly efficient on a large scale, both in terms of performance and in terms of
cheat prevention.
Since refereeing is somewhat similar to failure detection, this thesis extends the proposed approach
to monitor failures. The resulting failure detection service scales with the number of monitored nodes and
tolerates
jitter.
Area of interest : Peer-to-peer, gaming, decentralized architectures, high performance computation
Thesis supervisor : Pierre Sens
Thesis advisor : Olivier Marin,Sébastien Monnet