CNS Core:Small:Revisiting Process Isolation with Compound Processes

University Of Illinois At Chicago, Chicago IL

Investigators

Abstract

Modern computer software commonly consists of groups of programs executing together. This may be as simple as a web server and a database, or as complex as thousands of virtual machines, each running dozens of programs, all working together toward the same goal. Such groups of related programs are often started by a single user, who may even think of them as a single unit or "stack". Today, this common model of a stack of software, working toward a single goal on behalf of a single user is not reflected in the way the software actually executes. Instead, each individual layer of the stack is painstakingly guarded against incursion by other layers, through a mechanism called process isolation. What is more, many modern stacks consist of groups of virtual machines (VMs), where an additional layer of safeguards is imposed between VMs, often resulting in high overheads and low efficiency. This project introduces a new concept, a "compound process" which better reflects the realities of modern software. A compound process may host several "guest" programs in a single trust domain, eliminating expensive and often redundant safeguards between components of a single software stack, to yield substantial performance benefits. The project will proceed along three main thrusts: (a) a basic compound process loader and support libraries, allowing the user to run arbitrary combinations of programs (guests) as threads in a single compound process, (b) runtime supports for individual guests sharing a single thread, allowing further efficiency improvements, and (c) automatically running all of the processes of an entire a virtual machine as guests within a single compound process. By improving the efficiency of execution of popularly used software stacks, this project aims to reduce the equipment needs and increase the energy efficiency of the software that runs the modern world. Beyond direct research impact, experience from the project will both enrich the computer systems curriculum and provide research opportunities to both undergraduates students at UIC, a Minority (MSI), Asian American and Pacific Islander (AANAPISI) and Hispanic (HSI) Serving Institution. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.