Current Work Projects
HARE
This project focuses on re-evaluation and exploration of operating system and runtime services at scales of up to a million cores. It is a joint project with Bell-Labs, Sandia National Labs, and Vitanuova funded by the Department of Energy Office of Science.
External (non-IBM) Projects
v9fs: Plan 9 Resource Sharing for Linux
Contributed to and help maintain the implementation of the Plan 9 distributed resource protocol 9P under the Linux 2.6 operating system. The use of the 9P protocol along with the recent addition of private name spaces to the 2.6 kernel creates a foundation for seamless distributed computing using Linux. This code was recently merged into the mainline Linux kernel starting with version 2.6.14.
Website: http://v9fs.sf.net
Past Projects
Libra
This project re-evaluates the software stack in the light of para-virtualization technology and hypervisor support within next generation processors and operating systems. Libra is an infrastructure enabling the use of logical partitions (LPARs) for the execution of stand-alone applications along side traditional operating systems. The design goal is to provide an environment allowing normal users to execute, interact and manage these custom kernels in much the same way they would with typical applications. The development environment is a set of modular component libraries providing necessary system services, and a familiar debug environment provided by exposing partition memory and control interfaces to a "controller" partition.
Mambo: IBM's Full System Simulator for PowerPC
The IBM Full-System Simulator has been developed and refined in conjunction with several large-system design projects built upon the IBM Power Architecture. As an execution-driven, full-system simulator, the IBM Full-System Simulator has facilitated the experimentation and evaluation of a wide variety of system components for core IBM initiatives, such as the STI Cell and the IBM PERCS projects. This technology runs on AIX®, Linux®, and Mac OS X (see the requirements section for details). A version supporting the PowerPC 970 architecture has recently been released on AlphaWorks.
I specifically worked on Linux kernel support for running on top of Mambo including the now infamous Bogus Drivers which allowed for optimized disk, networking, and system performance (in tradeoff for simulation accuracy).
Blutopia Cluster Lifecycle Management
We have implemented a proof-of-concept prototype named Blutopia that embodies our ideas of a new software life cycle management paradigm driven chiefly by simplicity. Aiming at facilitating management, Blutopia allows administrators to deploy multiple machines fast and effortlessly in as little time as it takes to perform a system reboot. It is equally easy to manage and augment the deployed infrastructure by assigning new roles, upgrading existing roles, and rolling back software versions and/or configuration in the event of unexpected problems arising from an administrative task. Underlying our administrative paradigm is stackable file system technology providing the ability to efficiently manipulate layers of the file system, perform instantaneous snapshots, and institute copy-on-write layers to capture local modifications to an otherwise shared system or application image. We conducted a preliminary performance analysis of our mechanism and concluded that it degraded neither the performance nor scalability of the cluster nodes on which it was deployed.
Super Dense Server
With the right software, a cluster of low-power, low-performance server blades can equal the performance of a traditional multiprocessor server while using less space and power. The Super Dense Server is a prototype blade server system based on low-power Intel CPUs and the CompactPCI chassis standard. IBM Austin researchers developed the Super Dense Server hardware as well as software which solves crucial system management and performance problems inherent in blades. The Super Dense Server blades have no hard disks or consoles, yet they run normal Linux applications and can be completely remotely administered. The Super Dense Server team also developed Power-Aware Request Distribution, a technique to adapt to changing load by turning blades within the cluster on and off.
Website: http://www.research.ibm.com/arl/projects/sds.html
Past Lives
Plan 9 Research Operating System
Developed by the people who devised UNIX, i.e. Bell Labs, Plan 9 takes OS development back into the realm of research. Plan 9 is an attempt to work on the concept of operating system from the ground up, reworking the whole idea using modern concepts and technology.
Plan 9 is not UNIX. If you think of it as UNIX, you'll often be frustrated because something doesn't exist or works differently. If you think of it as Plan 9, however, you'll find that most of it works very smoothly, and that there are some truly neat ideas that make things much cleaner than you have seen before.
Website: http://plan9.bell-labs.com/plan9
Inferno Network Operating System
Inferno is a compact operating system designed for building distributed and networked systems on a wide variety of devices and platforms. With many advanced and unique features, Inferno puts an unrivalled set of tools into your hands.
Website: http://www.vitanuova.com/inferno