History (1995): Exclusive Interview With Exabyte CEO and COB Peter Behrendt

Computer Data Storage Newsletter: Where’s Mammoth?
Peter Behrendt: Mammoth is here. Mammoth is well. We ‘ve built several hundred units, we are shipping evaluation units to a small number of major OEMs, most of them are current customers, there’s at least one that has not done business with us before, somewhere in the video field. Last week, and this week and next week we’re shipping a fair number of units to ISVs so that they have an opportunity to test their software with Mammoth, not just test it but do the buffer management defining to make sure that their software gets the full benefit of the high speed. For the rest of this quarter we’ll ship a modest number of units to other OEMs and VARs, and we’ll start cranking up full production at the beginning of the quarter. The first announcement was about two years ago. Well, that was not an announcement of availability, that was giving a roadmap, a direction to customers so that as they plan their tape strategy, they knew what kind of 8mm product they should expect. If you’re saying that Mammoth is late, it is, but it’s not late 2 years.

When will you officially announce the product?
Announce is a vague concept. The existence of Mammoth as a product that will be coming was done years ago. We published a roadmap that said that in the 1995 time frame there will be a product that is 20GB and 3 MB/s and backward compatible with data compression in a half-height form factor. So that we’ve done.

But will you make an official announcement in the future?
I don’t know. We may be saying something like mass production of Mammoth has begun, but it’s not an event, it’s a process, right? Like writing a book. And we have started, we’ve built machines for quite some time, and we’ve shipped very small numbers to customers, and it’s a continuing process.

You’ve been sending OEM units. When will you be ready with final production units?
I don’t think that whatever products we’ll build beginning next quarter will be very different from these. We will be doing some additional testing, but the ASICs are all done, and there’s no change to that, the microcode is done, the tape path is done, the deck is done, the electronics are done, so we may do some tuning to improve yield, some things like that, but the product is finished.

When will you send the units for the first OEM orders?
When we get the first OEM order? I don’t mean to play games with you, but you know OEMs, they get a handful of these, they start testing, they want 10 more, they drop them from the roof, put them under water, they take anywhere from 2 to 6 months to qualify the product, and then they give you the order. And when we get the first order, we’ll make the first shipments.

When will you sell the units to the distribution channels?
In Europe, I would expect distributors to begin getting units in the first quarter. In USA, the industrial distributors will get units in the first quarter. At this point, we do not intend to offer Mammoth through the commercial US distribution market for a considerable period of time, probably not at all in 1996, so that we don’t have a hard deadline on that, but the plan is to offer this in priority to software developers, to OEMs, to high-end VARs, to library customers and to industrial distributors, like Anthem, for instance.

Why was Mammoth so late? Is it a technical problem?
Well, I think that if you look at what we were trying to do, which is to increase data rates 6x in one step, and capacity 3x in one step, you have to admit that’s a fairly significant jump. Can you give me another example of a tape drive or a disk drive that in one step increased capacity or data rates that much?

I can think of one, Magstar from IBM, 12x the capacity.
But how long has it been in development, and is it on time? How much more expensive is it than the last one? My point is that if you take something and you make it 50% larger or 50% faster, you have to change one thing, maybe. If you set yourself a goal which is that kind of a leap, you have to change the media, you have to change the head, you have to change the drum, you have to change the tape path, the electronics, the microcode. When you do all of these things and then insist they be right before you ship, you ‘re going to find some delays.

Could you be more precise, though? Which component was the most difficult to make?
They were all difficult. But I think the ASICs in there are very complex, I mean it’s a very large number of gates in some of these ASICs, something like 5 in there. Three of them worked the first time, 2 of them we had to do multiple passes, and we had to switch foundries, because one of them that we had originally gone with decided not to stay in that business. So for a while, the ASICs themselves were the primary delay. The ASICs are now perfect. We have no problem about those. But when the ASICs are late then the total system integration is delayed, because you can and we did do all sorts of simulations of the tape path, all sorts of simulations of the software, but until you have all of the pieces, you never know how well the thing works as a unit. So when the ASICs are late, then the systems testing is going to start late. We used the C language to do 90% of the code, so we’ve done a great deal of simulations, and the code looks to be in great shape.

Is the drive compatible with previous 8mm cartridges. Fully?
Yes. Read-compatible. To write, you have to use the new advanced metal-evaporated tape to get 20GB.

Will you have lower-speed versions?
Let me say that there’s a big gap between the product we ship today and Mammoth, and I wouldn’t be surprised if we find some way to fill that hole. There are 2 ways to do that, you can de-feature Mammoth, or you can increase the speed and capacity of the current product. I’m not going to comment on which of these approaches we’ll take, but I wouldn’t be surprised if we offer a product that falls in between those two.

Does the product have the specs you announced originally?
Yes, absolutely. In fact, it does things that we didn’t announce. If I can show you, here’s a Mammoth that has a display in it, not next to it, which everybody can do, but it has a small display actually housed within the form factor. And that’s something nobody knows, and something we hadn’t announced, so Mammoth actually delivers some capabilities that weren’t anticipated.

What kind of cache do you have in this drive?
4MB! Yes, sir. It’s a hell of a machine. To sustain the high transfer rate. I mean, with data compression we run at 6MB/s, and we want to keep it streaming. Our simulations have shown that 4MB/s is the optimum

What is the price of this drive?
We don’t sell to end-users, but we think Mammoth will probably have a street price around $6,000.

How much did you invest in the Mammoth project?
Too much. Which means I’m not going to tell you that. From our point of view, everything about Mammoth is new, and it shows. It is the first helical design without a capstan, so there is no point where two rollers squeeze the tape and grind dirt into it, and it gives much better tape tension control. So while Mammoth is announced as a product from our point of view, the investment we made is also in a technology platform, and our intention is to use this technology base that we have to provide additional products, some maybe smaller, some larger.

Have you calculated what the delay has cost Exabyte?
There are two ways of looking at that. If you think about there being a sort of window of opportunity, I don’t think it’s an issue, because I think that the need for a product like this is just beginning. There are not many systems that can sustain a 6MB/s data rate, so I think the fact that it’s late does not preclude some future opportunities. It’s different from the disk business, because if you miss a certain design point for a disk by 2 months, say, you’ll never recover on that particular gen. Tape isn’t like that. So for the window of opportunity, I don’t think there’s a very great cost. On the other hand, it’s allowed DLT to be the only product for some period of time that can address a reasonably high data rate, and that’s given them an opportunity they wouldn’t have had if Mammoth had come out earlier, so there’s a cost involved there. But I think the opportunity for Mammoth is really beginning, not passed. Most of the people who are buying DLT now are very anxiously awaiting Mammoth. My guess is that in many cases, this is not going to replace DLT, but will be added. We see a tremendous interest from all sorts of OEMs and resellers in Mammoth. In a sense, I’m very pleased to see that DLT is doing well. If DLT were not selling product, I would be very, very worried, because in a way, DL T is a free market research project for us, they’ve very clearly proven that there are significant numbers of people who want a high-speed 20GB tape drive. We think that we have the better solution, but the fact that they’ve sold a good number of those is a very positive sign.

If you compare the market when you announced your first 8mm drive from the current market, the competition is completely different.
The cheap 005-3 is going to come next year. I don’t know how cheap it’ll be, but it’ll come.

The Tandberg 13GB is there, Magstar is going to come, the competition is much greater than before. How do you place Mammoth among the 4 products, 005-3, DLT, Magstar and Tandberg?
Let me ask a question, the 13GB is how late, a year and a half? The point I was trying to make before is that complicated products tend to take longer than people think? Arcurate [a Conner tape technology that are still waiting to see-Ed.] is what, 2 years late? Tandberg 13GB a year and a half late. 005-3 will probably be a year late, in my judgement, because it’s certainly not here. I can’t speak about Magstar. But high-end tape drives tend to be quite complicated, just as an aside. I think that if people have a hard commitment to DDS, 005-3 is the logical choice for them. If people have a hard commitment and good experience with 8mm, I don ‘t believe that they’re going to switch to 005-3, they ‘re going to switch to Mammoth. So that’s just the next logical step in both technologies. I don’t expect a lot of switching going on. If you look at the absolute comparison, this thing is 20GB vs. 12GB, it’s 3x faster, it’s in a larger form-factor. But we haven’t found that to be much of an issue. You pushed me really hard a few years ago about the 3.5-inch form factor and I told you I didn’t think it was going to be a big issue, and it hasn’t been. I would guess that way over 70% of all tape drives that can fit a 3.5-inch form factor go out in rails to fit a 5.25-inch hole. So I don’t think the form factor will be a big issue. Mammoth will probably cost more than 005-3, so if people want this kind of capability, I think it’s the best product. I can’t compare it to Magstar. I think the price is totally different, the backward compatibility is totally different, the physical size is different, the power requirement is different.

What’s your target In terms of sales next year?
I expect that the majority of our 8mm drives next year will not be Mammoth, because this is a high-end product, this has nothing to do with supply and demand. Well, supply. We will probably be supply-constrained, number one, for part of the year, but a high-end product will typically sell less than the lower-priced version. And Mammoth is not a replacement product. It’s an additional product, and we will continue to offer our current 8mm product with the Sony mechanism in it, and we have also introduced the B700. That’s the new 8mm product that’s beginning to become quite successful, and I think in 1996 will sell in large numbers.

In which quarter do you actually see Mammoth outselling your current 8mm lines, in terms of sales?
I don’t know that it ever will, in units.

And in revenues?
I don’t know. We’re going to sell our customers what they want to buy. And if they want to buy a lot more Mammoths than current 8mm, I’ll be delighted, and if they buy a lot of these and many of those, I’ll be delighted.

This article is an abstract of news published on the former paper version of Computer Data Storage Newsletter on issue 95, published on December 1995.

Note: Exabyte was acquired by Tandberg Data in 2006 for $28 million.