|
|
René Copeland: New directions for CUG
"We [SGI] take CUG pretty seriously. We really feel it's important ..."
Note: Three weeks after René Copeland gave this interview, he was laid off in a downsizing that included more than 200 other SGI employees in the midwestern region of the U.S. and many additional employees worldwide.
 René Copeland, former SGI liaison to CUG |
Customer relations.
At the time of the Cray merger with SGI, Bo Ewald suggested that we devote somebody to sell the merger to our customer base. So starting at that time, my job was to get out and see as many customers as possible and sell them on the notion of the merger. At the same time, we were working with Customer Advisory Boards, bringing together customers of a similar background -- in this case, government and industry customers, customers in Japan and Europe -- to give them an opportunity to voice their opinion about the merger in a direct line to the organization. We'd spend a couple days with them and respond to their questions and concerns. We went through about a year of that -- it was part of customer contact concept. My relationship with CUG itself has taken a lot of different forms, but fundamentally it's been being a contact person for the CUG organization. My job has been to stay in touch with the CUG board to make sure the meetings come off smoothly -- and at the same time try to move CUG from the data center organization it's been traditionally to one that's perhaps more representative of the organization it was acquired by. We've been trying to get more SGI users involved, bring more relevance to meetings as they pertain to some of the products -- dealing with the Origin2000, for example.
I think CUG has got a good future. Number one, it's probably one of the most successful user groups high-performance computing has ever had. The industry over time has had small user groups -- Convex, Thinking Machines, IBM, and others. CUG has been one of the more successful. It's had 40 meetings, its history has been rich, and I think its future, likewise, is pretty bright -- although we do have to make some changes. Some of the key things we'll be looking at in the future -- and again, it's a customer organization, so this just reflects my thinking, it really has to be condoned by CUG board -- will be making the meeting more relevant for the entire Silicon Graphics install base. Right now we still have a group of members who are pretty much legacy Cray members. There are lot more large Origins, a lot more large Onyx users out there whom we could embrace and bring into the organization. For example, America Online is not a CUG member, but I think they should be; they probably have several hundred SGI servers. They may be the largest customer we have, outside the government. I may be wrong, it may be Disney or American Airlines. But whoever is that largest customer, or the largest three or five customers -- America Online, American Airlines, Disney -- we need to go out and get them into the organization. We need to get in front of them, we need to let them know that this group of users exists and we would like them to play a part in the organization -- not just to become members, which is relatively trivial, but to take a leadership role, to participate, to help define the agenda, to help run the meetings and set the direction. By getting new organizations such as the large SGI customers involved, I think we'll have an opportunity to change the composition of the meetings to be more reflective of the entire user base, not just the data centers.
We might see more talks on IRIX and fewer on UNICOS. We might see more talks on the reliability of RISC- and server-based hardware, fewer on T90 reliability. We might see more talks on the use of large parallel servers as opposed to single-processor vector machines. We might see Special Interest Groups and BOFs develop around customers who have more of an industrial, commercial focus. We might hear questions at CUG like, "What's your performance using Oracle as opposed to NASTRAN?" I don't think the other talks will go away, but I think this will make a nice blending of the two. And then in time -- well, the architecture will be merging in four years. That's a good question. Currently there are about 600 traditional Cray sites, and maybe as many unique SGI sites. So in terms of membership, probably somewhere between 1,000 and 1,200 is probably a reasonable population base. I would like to see membership somewhere in the 70-80% range -- that would be 700-800 members. In terms of actual meeting attendance, participation in the domestic meetings should be high -- I think we could get 75-80% of the members attending -- and I would guess probably 50-60% for the international meetings.
I think they can play a role with SGI similar to the role they played with Cray in the early years. The original Cray product was very fast hardware. As it became more pervasive, we weren't quite certain what had to be added to the software -- the compilers, the operating system -- to make it more usable for the specific customer buyers. We didn't know if the reliability was good enough -- and if it wasn't good enough, what could we do to make it better? We relied on CUG -- which was probably a lot smaller at the time but still contained our key users -- as the source of that information. Every six months the people who worked on hardware and software and maintained the equipment had to get together to face a fairly large and demanding group of customers and answer their questions. Sometimes the answer was "I don't know" or "We're working on it," but it wasn't a situation that could be avoided. Six months later the customers were going to be there saying, "The last time you said you'd work on it," or "You said were going to figure it out." And I think that rigor, that discipline, on a regular basis, of facing the customers -- not as an obligation or a punishment, but as an opportunity to connect with them again and see if what we were doing was right, if the direction we were taking made sense -- is one that helped us a lot in the early days in Cray. I think the role that CUG can play with SGI is a similar role. The SGI hardware and software is much more mature than the Cray stuff was at the inception, but it's fundamentally focused on a more general market. And CUG users tend to be toward the technical high end. So the role CUG can play would be to help the people working on products ensure that those products are adequate for the very high-end part of the market. I think the company does a good job in server-based integration, in small-to-medium-sized integration. But CUG can help them determine what makes sense in 128-256-1000-processor-sized configurations -- what kind of reliability and software we need, and what kind of performance expectations. CUG can be a good solid source of input into the organization. Our customer base is really the best in world. We have the largest organization of media and visualization sites, research labs, industrial companies -- and the supercomputing sites are part of that base, they're a huge, resource-rich group of people to draw from. If you can satisfy that group, I think you can very easily satisfy the general market. We met on CUG today, actually, we take it pretty seriously. We really feel it's important.
I joined Cray itself in 1980, but as salesman, I typically didn't go to the meetings. Around 1991 I went into marketing and started to work with CUG a little bit. In 1994-95, somewhere in there, I picked up full responsibility for CUG
Locationwise, I'd give it to Barcelona. But in terms of the most memorable CUG for me, the one I thought was most critical for the organization was probably the last one in Stuttgart. There are a couple reasons I'd give it to that one. One is, I thought the organizers at the university did a wonderful job of it. Walter [Wehinger] absolutely loved what he did, it worked well. Two is, it was in a location where SGI really does have intense competition in Germany -- there are places where the SX-5 is being sold in a big way. And despite the competition, we got one of largest turnouts that we've had in a long time in the international arena. To that I would add that we had a new CEO -- we lost some old executives and gained a couple of new ones. Some stalwarts and luminaries, icons if you will, were leaving, and there were new people. I thought the statements by the current CEO, and the willingness of the audience to listen and the corporate executives to participate, represented the true feelings of everyone. The message was, we're important to each other. We're not always going to be able to do everything that the other requires, but as we go forward, I think we have a pretty solid partnership. Each side had some needs that had to be satisfied. The CUG group needed to validate its existence; the senior executive group needed to let customers know they were still committed to CUG, although they might do things a little differently. I think both sides heard that. I think it was important to make a smooth transition. There was a qualitative feeling of excitement in air, despite the woes of the company, that things are going forward, that CUG has a future and will be supported -- and it happened at a time when everything was in doubt about CUG and what CUG was going to be about. But at the end of the day, Rick did say he would support the plans and try to make it to future CUGs. We also did a product announcement. So for me, Stuttgart was one of the more memorable CUGs. It took a little bit more work to coordinate the product announcement, but at the end it paid off, and I think everyone's kind of set for whatever comes next -- both CUG and SGI
Correct. I think you can even say there's no reason not to. CUG is a separate corporation. It's going to go on. It has its own budget, there's people drawn to it, there will be numerous people at any given point. I couldn't see a vendor not supporting a user group unless it was going out of business.
OK ... do you have any other message to the membership? To the members? If you are a CUG member, participate -- send your people, be on the committees, come to the SIGs, let us hear from you, vote, pick the officers, work on the program. If you're not a member, join. It's 500 bucks -- join, be a part of it. CUG is a small, very important part of our business. To a great degree, we're affected by what our customers say. I think it's probably safe to say that the very high end is probably more affected by customer input than any other part -- more than, say, the Alias/Wave Front, workstation, or server components. Everyone listens to the customer, but the high-end users have more impact, proportionately. Stand up, be counted, let your opinions be known.
|
