In this post, I will look at T-Mobile with respect to the iPhone and AT&T. T-Mobile USA is the US operating unit of T-Mobile International AG & Co., the mobile communications segment of Deutsche Telekom.
Deutsche Telekom AG & Co. K.G. (NYSE: DT) is one of the leading telecommunications and information technology service providers in the world with net revenue of €61.3 billion in 2006. Its business is organized into three strategic business areas: Broadband/Fixed Network, Mobile Communications, and Business Customers. T-Mobile International is one of the world’s leading companies in mobile communications with about 112 million customers worldwide. It has a strong presence in Germany, the United States, the United Kingdom, the Netherlands, Austria, the Czech Republic, Hungary, Slovakia, Croatia, Macedonia, and Montenegro. >>>
So, you thought Sports was all about watching games and cheering your favorite team? Think again. It is big business on the Internet and is all set to win the game with live streaming, match highlights, scores, bulletin boards, fan blogs, social networks and fantasy sports services.
According to MediaPost, more than 35% of Internet users depend upon the Internet as their primary source for sports information between 9 a.m. and 4 p.m. Around 52% of the users go online for sports-related content at least occasionally, while 24% are frequent visitors. Nearly 52% of sports site users seek out scores, while 43% read news stories online. >>>
SM: The market you are pursuing is embedded processors, so you do not really have demand for the fatter operating systems, such as Windows Vista. AA: Right, that is not where we are, but if for some reason that became important to a customer it would be done. Each processor core is full featured, so they can run any operating system. We took off the shelf Linux and made our chip work with it because we did not want to build an operating system from scratch.
SM: You mentioned cache issues earlier, did you make any advances in cache structures? AA: That is the area of our third major innovation, which is a distributed coherent cache. Of all the challenges of multicore design, cache issues are one of the biggest. When you have multicore architecture, you can create a big L3 cache; in essence combine L1 and L2, and create a big giant cache.
The problem is big giant structures are low performance and high power. >>>
SM: In general, was the workforce at 3Com more aligned with your vision than to Krause’s? EB: It was a split workforce. We had some computer experts, and we had some networking experts. What I ended up doing was to choose one – we could not do both.
We built upon our roots at Bridge, this is what we would have done if we had not been interrupted and distracted by the 3Com transaction. In 1990, after validating that this was the direction I wanted to take, we ended up restructuring the company. We let workstation and server divisions go. I cut an agreement with Steve Balmer at Microsoft and we wrote a big check to get out of the contract. Bill Gates, Steve Balmer and I were doing keynotes at the ETRE conference, so I cornered them and we had a long drawn-out discussion and negotiated our way out.
Of course we had lost time. >>>
If you are one of my regular readers, you know that I am a huge fan of good movies, and am happy and willing to be swept off my feet to fantasy places like Hogwarts School of Witchcraft and Wizardry.
This weekend, I experienced another such world, that of Remy the rat, who aspires to be a chef. When fate places Remy in the sewers of Paris, he has an opportunity to emerge in the kitchen of a restaurant made famous by his culinary hero, Auguste Gusteau. Gusteau’s famous motto: Anyone Can Cook. Well, Remy does. >>>
SM: This is a radical redesign, and it seems it should be advantageous in other areas as well, right? AA: True. The other beauty of the mesh is that not only does it solve the performance problem, it also really addresses the power problem. A bus is a big centralized structure, and any big centralized structure will have to have high power. Big wires take a lot more energy than small wires. In mesh architecture, since each switch talks to a neighboring switch, the wires are shorter and the architecture is fully distributed making it much more energy efficient.
SM: The energy issue is tackled at the interconnect level? AA: That is a big part of it. For example, in bus based multicore setups, it is not unusual to see 30% of the power gets consumed in the buses. You also have big centralized structures like caches. By moving away from big structures, you get further power efficiency that comes from the tiled approach. Rather than having any single big structure, you make various small structures. >>>
Eric takes over as CEO of 3Com, and as a first item of business, makes some hardline choices.
SM: What catalyzed the CEO change at 3Com? EB: The board realized something needed to be done. In 1989, my two partners at Bridge, Bill Caraco and Judy Estren, left out of frustration. They realized this was going nowhere as a result of the merger. We were no longer calling the strategy. We were part of the management team but it was very frustrating trying to make progress. I decided to stay a few more months and things kept deteriorating.
One thing I have not mentioned is at the time Bill Krause had tied the company up in a complex agreement in Microsoft to build the next generation operating system. Microsoft was going to build the successor to DOS which was the OS2 operating system. IBM and Microsoft were writing OS2, 3Com was going to standardize on OS2 and build computers, servers and workstations to run OS2. With OS2 we were going to compete against Novell and drive them and their Netware Operating system out of business, and we would build computers and workstations just like Sun Microsystems. The difference between us and Sun was that instead of building for scientists and geeks we were going to build for the business community, and by the way we would commit to doing it with x millions of OS2 licenses. It was a terrible agreement. >>>
Anant identified five significant areas where innovation had to occur for multicore processors to really take off. He addressed each of these areas. Here we discuss the interconnect bottleneck issues in further depth.
SM: So you are doing some set of pre-routing on a switch. AA: Exactly. Now that you have a switch on each tile, what you do is take these tiles and you lay them out on a chip in a checkerboard fashion.
SM: What is the basis of the routing algorithm? AA: Switches talk to each other and route packets around, which was part of the research. It is called dimensional routing, which is how you route packets from one point to the other. You route in one direction first, and then you change to the next direction. There are other routing algorithms, and they differ in efficiency and ease of implementation, but the basic idea is that they can shift packets around.
SM: All of this is invisible to the programmer? AA: It is completely invisible. As far as the programmer is concerned they could think of it as a big switch which is a magical bus that instantly transports packets around. >>>