Researcher: Mahadev Satyanarayanan
Research Area: Mobility
Portable computers have been the driving technology of mobile computing since the early 1990's. Today the term "mobile computing" is almost synonymous with the use of laptops or handheld computers. Unfortunately, usability often suffers when a mobile device is optimized for size, weight and energy efficiency. On a handheld device with a small screen, tiny keyboard and limited computing power, it is a challenge to go beyond a limited repertoire of applications. We are exploring a new approach to mobile computing that takes advantage of the plummeting cost of non-mobile computing hardware, such as desktop computers and large wall-mounted LCD displays. In this approach, one finds and uses hardware transiently in any location. We refer to this concept as "infrastructure-based mobile computing." We are exploring two flavors of this concept. 1)The carry little approach leverages fixed infrastructure to augment the capabilities of a mobile device. 2) The carry-nothing approach relies exclusively on fixed infrastructure. Establishing trust in infrastructure is a prerequisite for real-world impact of these ideas. We propose to create a campus scale test-bed in which trust centric experiments can be carried out to understand usability and user behavior issues that relate to trust and security.
Carry Little Approach: This approach is related to techniques such as dynamically composable computing and cyber foraging. For this approach to work, the infrastructure must be provisioned with exactly the right software needed by the user. This is unlikely to be the case everywhere, especially on a global scale. To solve this problem, we are building a system called Kimberly that enables rapid software provisioning of fixed infrastructure for transient use by a mobile device. Kimberly uses virtual machine (VM) technology, but avoids the performance challenges of running VMs on resource-poor mobile devices.
Carry Nothing Approach: Internet Suspend/Resume (ISR), pioneered at Carnegie Mellon, is the classic example of this approach. ISR is implemented by layering a VM on distributed storage. We are working towards a new version, OpenISR, that embodies four new ideas:
1) VMM-agnosticism: To give the user the greatest freedom in the choice of resume site, OpenISR should allow resume to occur on any x86 machine with a VMM, even if that VMM is different from the one at the most recent suspend site.
2) Transient thin client mode: We are exploring an approach in which OpenISR resumes in thin client mode but transfers sufficient VM state in the background during an initial period of use to transparently and seamlessly switch to thick client modes.
3) Guest-aware migration: In all previous versions of ISR, the guest operating system has been treated as a black box, which does not allow the ISR to exploit the knowledge it possesses. We are relaxing this assumption in OpenISR to exploit many "guest-aware" optimizations.
4) Cross-parcel data-sharing: We are exploring usage models that simplify data sharing across VMs. Our approach is to use a distributed file system such as Coda in the guest operating system as the data-sharing mechanism.