Protecting shared sensitive information is a key requirement for today's distributed applications. Our research uses virtualization technologies to create and maintain trusted data paths across distributed machines, for the services being run and their information exchanges. For trusted data paths, runtime protection methods control what data is visible to which distributed services operating on it, guided by online monitoring that determines the levels of trust inherent in the paths' machines, services, and service actions. This paper presents a key functional element of trusted data paths, which is the ProtectIT interception mechanism for controlling the data exchanges between the different virtual machines running trusted services. ProtectIT can be applied to any communication and/or I/O performed by virtual machines, and because ProtectIT does not require application, middleware, or operating system modifications, it can be used to construct trusted data paths without the knowledge or consent of such entities. Further, since ProtectIT operates in virtual machines isolated from those used by applications, it is not subject to the attacks faced by services exposed to the open Internet. ProtectIT's functionality consists of dynamic protection rules represented as data filters applied to virtual machines' communications. Examples presented in this paper include email services for which ProtectIT's filters control data visibility to mail servers and clients, and unsecured virtual machine communications morphed into secure ones via ProtectIT-based message interception.

No abstract available.

Growth in broadband connectivity is making media streaming applications increasingly popular. For scalability, media is streamed across sets of proxy servers embedded in overlay networks, where the quality of delivered content depends both on available network capacities across overlay nodes and the capabilities of proxy servers. This paper addresses proxy server performance for media streaming and for the delivery of live media content. Our approach to efficient content delivery is to develop a set of kernel-level data streaming abstractions, termed KStreams. Compared to user-level solutions, KStreams (1) offers improved performance for the multiple data forwarding models commonly used in data distribution networks, (2) reduces per stream overheads by eliminating unnecessary system calls and memory copying, and (3) offers improved levels of predictability for the Quality of Service (QoS) experienced by media streams due to its use of non-preemptable kernel-level threads and its ability to directly interact with the CPU scheduler and other kernel-level resource managers. (4) Once initiated, KStreams operates without further involvement of and asynchronously to applications, permitting them to carry out other tasks.

Monitoring the resources of distributed systems is essential to the successful deployment and execution of grid applications, particularly when such applications have well-defined QoS requirements. The dproc system-level monitoring mechanisms implemented for standard Linux kernels have several key components. First, utilizing the familiar /proc filesystem, dproc extends this interface with resource information collected from both local and remote hosts. Second, to predictably capture and distribute monitoring information, dproc uses a kernel-level group communication facility, termed KECho, which is based on events and event channels. Third and the focus of this paper is dproc's run-time customizability for resource monitoring, which includes the generation and deployment of monitoring functionality within remote operating system kernels. Using dproc, we show that: (a) data streams can be customized according to a client's resource availabilities (dynamic stream management); (b) by dynamically varying distributed monitoring (dynamic filtering of monitoring information), appropriate balance can be maintained between monitoring overheads and application quality; and (c) by performing monitoring at kernel-level, the information captured enables decision making that takes into account the multiple resources used by applications.