Microsoft’s System Blueprint
By Jia Wu, China Internet Weekly
October 5, 2009 2:00 PM PT

"I received my advanced education at Fudan University and pursued my doctorate at Cornell University. I worked for six years at Microsoft Research Silicon Valley after graduation and transferred to the Microsoft Research Asia System Group in May last year. Don’t hesitate to ask me for any information you desire." No sooner than I had figured out my first question, Lidong Zhou had finished his highly concise and informative opening remark, leaving me temporarily speechless.

Zhou’s working style bears the trademark of Silicon Valley: calm, rigorous and focused. A researcher of computer systems, he is not a person of many words, but he still walked the extra mile to make the profound issues in his research roughly understandable to a layman like me.

In Silicon Valley It Begins

Silicon Valley has a certain type of charm for all researchers. Home to many of the world's leading high-tech companies, it is a training ground for IT talent in the United States. Nearly 1,000 fellows of the U.S. Academy of Sciences work in the Valley, and there seems to be no better place for researchers to grow and prosper. A young Lidong Zhou therefore chose to start his career in Silicon Valley.

The Microsoft Research office located in Silicon Valley was really something special. It was still in preparation when Zhou arrived for his job interview, but few could imagine that its handful of formal employees at that time included milestone-setting researchers lured from globally renowned institutes such as Xerox PARC and DEC SRC, flanked by senior academics of the Academy of Engineering. As one of the early birds, Zhou had chances to talk with each of the formal employees before the end of his job interview.

"They have had a big influence on me. In fact, their relentless quest for truth instead of fame, their sincere respect for and collaboration with each other, and their willingness to learn from peers set an example for everyone around them. It would be a real blessing for anyone to be able to explore the depths of truth and to experience and contribute to that type of teamwork. Individuals of very high ranking worked side by side, and they offered substantial assistance and opportunities to me so that I could grow from being a junior researcher to a person with the type of wisdom needed to be a project leader” , Zhou said with emotion.

Microsoft Research Silicon Valley also has a unique corporate culture in that the formation of teams is purely dynamic and spontaneous: any researcher with a good idea is allowed to initiate a new project, to improve their concepts through peer-to-peer communication and to attract team members with personal charisma and technical expertise on their own. Leading a team means greater responsibility, but it’s necessary to a researcher’s professional development.

Zhou recalls being encouraged to organize his own research team after he had completed a project led by a senior researcher. The senior researcher even volunteered to play a supporting role in the project. "I benefited a lot from this team leader experience, from which I learned to unite and motivate team members, and to cope with difficulties and setbacks. I felt powerless and encountered bottlenecks in the first few days, but everything became easier and easier as each member of the team began to exert his or her effort in the same direction."

Shared Vision

The “take it easy” style of the leaders at Microsoft Research Silicon Valley help members of their teams remain on track with their research, "but it is more vigorous and passionate here,” Zhou said, referring to Microsoft Research Asia, “as younger minds feel less fear. We, after all, have more of a pressing need for accomplishments." Having done too much “routine work,” Zhou was thirstier than anyone else for challenges and personal influence after spending time in Silicon Valley.

"From my talks with people at Microsoft Research Asia, I sensed their great desire for clever minds. They are never satisfied with the status quo, but look at developments using a five-year or even ten-year timeframe: which kinds of people are needed to bring Microsoft Research Asia to the next level?" Having traveled half way around the world, Zhou came to know that a shared vision is the most powerful asset a company can leverage to retain the brightest souls, and is the most powerful catalyst to personal influence.

"If I am able to apply something that I learned in the U.S. here at Microsoft Research Asia, and if I can help researchers at Microsoft Research Asia better understand the projects they are working on, young researchers here are going to do much better than we did and are going to have a much more powerful impact on the country’s future growth. This, of course, is a huge attraction for me," he said, gesturing with his hands to demonstrate how “huge” the impact could be.

According to Zhou, the Microsoft Research Asia System Group has been working on the development, optimization and debugging of large-scale distributed systems -- the underlying technologies for cloud computing. Such systems involve applications in many traditional areas, including multimedia data processing, visual computing and text data mining, and extend to a wide range of disciplines in science, healthcare and environmental protection. "Our project is designed to allow users to simply and effectively apply cloud computing to emerging problems or problems that we’ve never been able to properly solve before,” Zhou said.

Beyond Computers

As a matter of fact, “system” is a term of very broad meaning -- in addition to computers, devices as small as sensors and cell phones or as large as data centers for cloud computing have issues that vary in nature. A system can be very interesting simply because of the wide areas it covers. Microsoft Research Asia, for its part, is devoting more of its energy and resources to system development in the future. "I really believe it is a very good time now to get a job in system research because technical advances in hardware and the implementation of cloud computing will enable us to achieve so many things that we previously could have only dreamed of,” Zhou said. “I think this is a time comparable to the days shortly before the first PC was born as there are too many opportunities awaiting us. We should not limit our vision to existing computer architecture because computers, as they evolve, will soon lose the cases that we are so familiar with and will become invisible but ubiquitous in our life. The boundary between virtual and physical worlds will increasingly blur. Tremendous changes will take place in the interaction between human and computing resources, which will have a far-reaching impact on system design."

So, What Would a Future System Look Like?

Zhou once again referred to cloud computing. "If you own thousands of computers and you want to develop a brand new search engine, or put everyone’s photos and videos in the 'cloud,’ you will encounter a typical system problem," he paused for a second, "This involves several factors; how do you make all machines parallel in high and reliable performance? This is an issue of top priority for our research. In addition, data reliability and reduced energy consumption are also issues for consideration."

Emerging system platforms and applications bring new challenges to the doorstep of the System Group. Now that the research of other groups at Microsoft Research Asia relies on the system, there must be diversified models in place to drive cooperation between groups so as to ensure parallel progress in research projects.

"We try to understand the properties of applications under development in other groups, and then design appropriate system architectures to meet their needs. Our efforts in the development of computing technologies make research and innovation easier for members of other teams, so we are actually the enablers."

Speaking of his most influential project, Zhou referred to the System Group’s cooperation with the Search Team. "Together with the Search Team, we redesigned the underlying architecture of Bing," he laughed, "Unlike the development of Windows where new editions take several years to come out, search engines are updated or upgraded constantly for improved functionality. From a system perspective, we jointly studied and designed frameworks and platforms that support innovative search engines. With our expertise, we managed to bring the system performance to a higher level, which could have long-term impact on the evolution of search service."

Microsoft is looking closely at all possible directions of system development. Apart from maintaining its competitiveness in computer systems, the company is attempting to circumvent Google’s moves in the battlefield of Internet service. Seeking breakthroughs from the system perspective is where Microsoft is the best of the best.