In the era of big data, the demand for secure data storage is rapidly increasing. To accelerate the complex encryption computation, both specific instructions and hardware accelerators are adopted in a large number of scenarios. However, the hardware accelerators are not so effective especially for small volume data due to the induced invocation costs, while the AES-NI (Intel®advanced encryption standard new instructions) is not so energy efficiency for big data. To satisfy the diversity performance/energy requirements for intensive data encryptions, a collaborative solution is proposed in this paper. We proposed a feasible hardware-software co-design methodology based on the stack file system eCryptfs, with quick assist technology, which is named adaptive crypto acceleration for secure data storage (ACA-SDS). ACA-SDS is able to choose the optimal encryption solution dynamically according to file operation modes and request characters. Adjustable parameters, such as α, β, and M are provided in our scheme to provide a better adaptability and tradeoff choices for encryption computation. Our evaluation shows that ACA-SDS can get 15%-25% performance improvement for big-data blocks compared with only software or hardware accelerations. Furthermore, our methodology provides a wide range of practical design concepts for the further research in this field.
Full Text Finder