by Uwe Harms
San Diego, CALIF. — DeTeCSM (Deutsche Telekom Computer Service Management) is the IT provider for the Telekom Group with operations (data centers), techniques (infrastructure) and services (call center, advisory services). It provides all the services outside the group too – they are marketed through Deutsche Telekom’s sales channels. The company employs a staff of 6200 and operates about 28 700 MIPS CPU, 110 TByte mainframe disk memory (MVS), 3 200 Unix servers, 177 000 PC, 135 000 printers and 30 000 terminals. The 6 Service and Computer Centers (SCC) are located northern, west, south-west, south, east and in the middle of Germany.
As IBM had installed their 100th IBM RS/6000 SP supercomputer, Uwe Harms got the chance to interview Professor Dr. Alex Rˆder, Chairman of the Managing Board of DeTeCSM.
Uwe Harms: “As you are German Telekom’s IT provider, you surely got all their powerful servers and mainframes and overtook all their computer centers, what happened then?”
Alex Rˆder: “We now own all the mainframes and the servers. The consolidation process started in 1992 with about 108 computer centers, at nearly each Telekom location a computer center. The consolidation finished in 1996 with these 6 SCCs. We now offer 28000 mainframe MIPS, all in CMOS technology. This step was followed by the servers, where we checked whether it is appropriate to centralize it or when it is locally used, keep it local.”
U.H.: “You are not concentrated on Deutsche Telekom but look for other customers. Can you give some examples?”
Alex Rˆder: “About 15% of our revenues come from sisters or outside, about 180 Million Euro (about 155 Million US$) out of 1.19 Billion Euro (about 1 Billion US$) in 1999. One interesting project was Lufthansa Passage GmbH. We delivered, operate and do maintenance for their desktops. We installed 4000 desktops and integrated about 130 applications. Additionally we installed ticket printers and realize the help desk. Another project, just finished, was the German Federal Agency of Labour Exchange. In the offices of the Labour Exchange we installed 83000 PCs, 1200 NT servers and 21000 printers. Friday after work, the old equipment was disassemmbled, the new installed and tested. Monday in the morning the new machines were up and running.”
U.H.: “What about Deutsche Post AG – are they outsourcing their equipment too?”
Alex Rˆder: “Yes, we just overtook the field service, have employees for the PCs and software available, and their corporate network. In the next step, end of this year, the mainframes and servers are coming.
U.H.: “You recently installed your 100th IBM RS/6000 SP server, how many nodes do they have?”
Alex Rˆder: “In total we have 1100 nodes with 2 to 4 CPUs and 2 to 4 GB RAM.”
U.H.: “That leads to an estimation of approximately 3300 CPUs. In the actual Top500 list DeTeCSM has seven IBM SP with an aggregated peak performance of 1.2 TFlop/s using 1600 processors. Thus I estimate a peak performance of the 100 systems of more than 2.5 TFlop/s. This is about 20% of the peak performance of the fastest computer of the world, ASCI White. This IBM RS/6000 SP has 8192 processors and delivers a peak performance of 12 TFlop/s. What are you doing with your Teraflop/s ?”
Alex Rˆder: “The IBM SP plays an important role in the server consolidation task, about 1000 servers have been consolidated. We use the SP’s for datawarehousing, as application server in the SAP R/3 and inquiry services as well as in a Tivoli environment.
U.H.: “Can you give us some details?”
Alex Rˆder: “I start with Tivoli. In the project Enterprise Process Management Framework (EMPF) we use the next generation of system management. The Tivoli building blocks run on 11 RS/6000 SP with 180 nodes to manage all the mainframes, servers and clients in our huge network from a central point. By the way, only two people manage the 100 SP systems. An other project is the SAP HR (Human Resources), 3 IBM SP with 70 nodes are used as application server with a backend database on an IBM S/390 mainframe. Here 10 000 named users, more than 3200 concurrent user, in 220 locations access secure R/3 for example with smartcard for identification. The database has a size of 500 GB plus the mirror. The inquiry service uses 7 SP with 100 nodes.”
U.H.: “IBM is your only supplier?”
Alex Rˆder: “Wo do not believe in a single vendor strategy, as competition reduces the costs and improves the support. As you can see in the Top500 list, we have big systems from other vendors too.”
U.H.: “Why did you chose IBM as a main supplier – 100 IBM SP systems?”
Alex Rˆder: “The IBM machines have the high performance, we need in commercial applications. Now there is a single point of administration. As I mentioned earlier, two operators manage all the SP-systems from one location. Other issues are availability, reliability, and scalability. Soon we can extend our machines.”
U.H.: “You mentioned commercial applications, do you have any engineering, e.g. computer aided engineering, or scientific application?”
Alex Rˆder: “No, we concentrate on the commercial field.”
U.H.: “What about Linux, does DeTeCSM has a Linux strategy?”
Alex Rˆder: “This operating system is a candidate. Actually we run tests with native Linux on IBM S/390. One and a half year ago, we installed a SAP application on Linux and ran it. But today we made no decision, which Linux. An other issue is the user. Our experienced personnel could download and install Linux on their PCs, but we do not allow this for the reason of easier maintainability.
U.H.:”Thank you Dr. Rˆder for your time and this interview. It surprised me, which supercomputing power, measured in TeraFlop/s, is installed within DeTeCSM.”