Cambridge researchers are developing a lean-and-mean programming framework called Mirage that is designed specifically to support applications running on cloud infrastructure platforms such as Amazon Web Services and Google App Engine.
"If we want to rewrite the Internet from scratch, how much can we improve the state of the art?" said Anil Madhavapeddy, senior research fellow at Cambridge University's Wolfson College and member of the Mirage team, during a presentation at the USENIX conference in Boston on Tuesday.
Mirage's key principle "is to treat cloud virtual hardware as a compiler target, and convert high-level language source code directly into kernels that run on it," team members wrote in a paper distributed at the event.
Typical operating systems "try to solve everything--from desktops to network servers to fast-paced games," the researchers added. "Mirage focusses on the domain of I/O-intensive cloud servers, which lets us specialise the stack and reap the benefits."
Applications that use Mirage "exhibit significant performance speedups for I/O and memory handling versus the same code running under Linux/Xen," the paper states.
Traditional stacks "are very thick, containing extensive support for legacy systems and code built up over years," they wrote. "This makes them cumbersome to build and deploy, inefficient to run, and complex to administer securely."
"It's kind of starting to lean over a little bit," Madhavapeddy joked as he gestured toward a rendering of a modern application stack containing some seven layers from hardware to application logic.
The team is using OCaml to develop Mirage. The programming language has healthy interest in the academic world but is gaining a foothold in the commercial market as well, Madhavapeddy said.
Other Mirage team members come from the University of Nottingham and Citrix's research and development arm, according to the paper.