Caravela software uses graphics processing unit (GPU) power to boost performance, and to cope with the security issues of the grid. The design and implementation of Caravela software is the focus of this research paper.
The background section lists the essential features involved in using the GPU to address the security issues in grid systems. The paper proposes that incorporating the GPU as a computational resource is adequate to deal with security problems. The paper also proposes GPUs as a general computation resource, but from the given technical information, the authors’ Caravela system only uses the GPU if the user has its shader code available. This means that the system is solely designed for GPU applications, not for distributed computing.
In the conclusion, the authors present one example, including computation time and communication data and time; it would have been better if the total latency were also indicated. Having one type of example is not quite acceptable when deriving conclusions about a system’s performance benefits.
The authors depict the use of GPUs as computation resources that are accessed and explored by the users of the Caravela system. The distributed use of GPUs is a notable idea, and has remarkable motivation. However, current GPU technology and programming architecture still don’t support the use of GPUs for all types of applications.
In conclusion, the use of GPUs in a grid environment is interesting. Although the idea of the paper is appealing, there are many open issues, and mandatory technical issues are not discussed.