Part of my role here at EMC is to spend time with writers, bloggers and journalists on various topics. It's a global endeavor, often crossing language and cultural barriers.
Most of the time, it turns out pretty well. What ends up in print is a reasonable reflection of what I said - maybe not perfect, but close enough.
Once in a rare while, the result can get jumbled. When it does, I certainly share a decent amount of the responsibility: was I clear, did I speak slowly, did I make it simple enough?
A few days ago, a write-up of a recorded interview appeared in Storage Newsletter. I was dismayed that it had a bad case of that "lost in translation" problem.
I'm not quite sure how it happened. The questions were reasonable; I thought my answers reasonable -- but the result wasn't ideal. To make matters even more problematic, one of the industry bloggers pounced on one of the many misquotes as "evidence" of a broader set of issues.
Nothing goes unpunished on the social web.
So, in the interest of continual improvement, I thought I'd reprint the entire interview here, along with an improved transcription of what I said. The answers are the same, they're just a bit easier to read. I've also added some expansion points (clearly indicated) as needed.
(I’ve omitted the long bio the writer included in the interest of brevity)
StorageNewsLetter: EMC was pushing ILM years ago, not anymore. Now it's big data. Who invented these words?
It was a guy back in 2004 who wrote an article about big data and the new value from information.
No, I think it was academic. They talked about the world of big data -- it first came from universities, then they grew up and found echoes in Wall Street, casinos, things like that.
We didn't start to call it big data until 2010, and then there was a series of articles from McKinsey, Harvard, The Economist. All at once, everyone was talking about big data. So as we got in 2010, 2011 it was about first the cloud, and everybody wanted to talk about it.
[Look at "Big Data" Dynamic Factor Models for Macroeconomic Measurement and Forecasting, by Francis X. Diebold, University of Pennsylvania and NEBR, July 2000]
What does 'big data' mean?
One view is to look at it as a new way to make money from information: We're all familiar with BI, with queries, spreadsheets. Here's this new business model -- people using information to make money by building predictive models -- crystal balls if you will.
It's a very different style of IT. Most IT is about saving money and reducing costs. Big data and big data analytics is about making money.
Additional note: in addition to commercial pursuits, big data is also transforming research, public policy and societal welfare -- indeed the world we live in. More can be found here.
What's the difference between data warehouse and big data analytics?
We're all familiar with data warehousing and BI. Typically we're using relatively small amounts of information. It's very clean, very standardized and you're usually asking about history.
Big data analytics is many data sources, some of them clean, some of them dirty -- and correlating them. The goal is to build a predictive model about the future. People who are good at big data analytics rarely come from the BI world, they tend to have a background in the sciences.
Why did EMC never enter into tape?
We saw tape as a flat, declining market. For the last ten to fifteen years tape market has been declining. At EMC, we work with tape, we interoperate with tape but again our investment strategy is to invest in growth markets so -- for example -- we were very early in flash, in virtualization, in security, in big data analytics.
If I remember well, StorageTek went to Sun, that didn't work out, it then went to Oracle, that didn't work out. We tend to invest in growing -- not shrinking -- markets.
These past years, we didn't see EMC inventing any disruptive technology like scale-out NAS or de-dupe being your more successful technologies in term of growth, but you're just getting them through acquisition (Isilon, Data Domain, Avamar). Is your R&D only updating current hardware and software?
Right now we track something close to four hundred startup companies, so we're always interested in new technologies. There is no way that we as EMC can be smarter than four hundred start-up companies. We're very comfortable with seeing a good idea, buying it and then adding it to EMC: Data Domain, Isilon, XtremIO, VMware, RSA, etc.
And -- at the same time -- we are doing substantial amounts of our own engineering. I look at VPLEX as a disruptive technology, w.g. cache coherence at a distance. Another example is Atmos, now three or four years old, it was the first commercial cloud scale-out object technology in the industry. I consider that disruptive.
But I'll agree with you, more disruptive technologies come to us through acquisition than organically. (note: VMware would be a good example) It's true for lots of big companies. We have our own technology that we engineer, but when you can buy good ideas on the marketplace, you also buy good people. It's kind of a mix of both.
Can we imagine an integration of VNXe and Isilon into a single platform?
If we did, we would end up with something similar to NetApp, and the market doesn't want that.
When we look at the VNX, it's very good at being one storage array - it has dual controllers -- I need some blocks, I need some files, maybe some snaps and similar. There is a great mid-market for that. These people don't get very big (note: capacity-wise).
When people start getting big they want scale-out. They don't want a product that's adapted to scale-out, they want a product that's built for scale-out. But you'll find common features in both.
For example the last Isilon release, known as Mavericks, you'll find some good snaps, good VMware integration, most of the stuff you'll find on the VNXe -- but for an Isilon I need a handful of nodes to get started.
Yes but these nodes could be VNX hardware.
Hardware is (becoming a) commodity, and I think we have to separate the hardware discussion from the software discussion. The hardware almost looks the same, Intel chips, etc.
When you look at what's NetApp doing with their last ONTAP 8.1.1, it's not a good scale-out system. We go head-to-head in competition, and when customers put the 8.1.1 next to Isilon, they want Isilon. Our belief is that when you get to things like scale-out or big data, you need an architecture that's built for it. It's very hard to add it on.
What will be the next storage killing application?
It's going to be big data analytics.
When will you have a dedicated all-flash array?
We announced our XtremIO acquisition this last year. We demonstrated it at VMworld and EMC World. It's an all-flash design. Native de-dupe, scale-out architecture, all flash. It fits very precise parts of the market.
Do you sell it?
No, right now we're in what's called 'early access'. Test it with the customers, make sure it works right. You'll probably see it in the first half of 2013 as a general product. (Note: this is now more like early 2H 2013 I am told) Remember: it's got a very specific profile: random IO/s, places where extreme performance is needed and tiering won't work -- a data set where everything has to run fast.
Will it be integrated as a cache with several tiers behind?
Nope, pure flash, no need for tiering. And because we have native de-dupe, the cost are pretty good. (Note: if someone wanted this, it would be either FAST as found in VMAX and VNX, the upcoming Project Thunder or perhaps VFcache).
Will it be possible to connect disk arrays to this platform?
No. For example, Isilon does scale-out for files. We need a scale-out block device for flash. That's where XtremIO fits. Now within flash we have tiers, MLC, SLC. So there's potential of tiering in the flash world but there's no disk arrays. (Note: VMAX supports the connection of external disk arrays, as do other EMC products such as VPLEX).
What's the percentage of your storage sales in virtualization environment only?
The challenge is that people are doing virtualization but they also have physical stuff. They bought it for VMware but they still have these older applications they put on the array. The percentage of our customers using virtualization is I guess over 80%. With Solaris and AIX virtualization it's probably at 98%. VMware, HyperV and Xen, -- yes – well over 80%.
You are not a big actor in the growing sector of HPC. Will you invest in this market?
We're starting, we are investing in this market, we see an opportunity to bring flash technology to HPC.
With the VNX or Symmetrix line?
Right now we're assembling things out of a VNX portfolio because it's closer to what HPC needs. But we have a lot of potential pieces, something like eight different boards, six different stacks, a lot of potential ingredients. Our current HPC solution is basically a flash-cache plus a VNX with a lot of flash in it. It's very good at check pointing.
And do you offer the Lustre platform?
We support it. We use mostly 10GbE but we're starting to do more work with Infiniband. Storage are becoming servers and I think in the next two years you'll see far more discussion about Infiniband not as a storage connect but as a memory to memory connect to storage.
Several competitors are doing better than EMC in backup software (CommVault, Symantec, Veeam, etc.) in term of growth, even if you invest in some companies (Dantz, Legato) in this sector in the past. What's the problem?
We see it as a backup market. Not as a backup software market. Let's take Data Domain: is that software or hardware? Well if you look at the hardware it's servers.
Yes but you don't have a product like CommVault Simpana.
Avamar is doing quite nicely, Networker hangs in there. We tend to look the market as the data protection market. If you want to cut it into software we do all right with our backup software products.
But what our customers tell us is that they want a backup solution. They want the services, the platform, the consulting, they want all of that. So we tend to look as the market a little bit differently.
What's your current acquisition strategy?
We have publicly announced that we're very interested in investing in next-generation security. In the storage world we're pretty good, there's no space in storage where we think we're underexposed. There's a couple little pieces that are interesting but if you look at our storage portfolio we got two, three flavors of blocks, two flavors of files, object, three management stacks, we've got data protection, etc.
You lost market share vs. NetApp for last known fiscal quarter (storage sales down 1.4% sequentially for EMC, up 2% for NetApp). What's your explanation?
(note – I have no idea where he got those numbers – here are the IDC numbers that most people use – which show the complete opposite – so I try and be diplomatic)
We're in a lot of different markets, high-end enterprise where they're not, consumers with Iomega ...
Yes but they are growing faster in storage sectors than you are.
On the product side, NetApp growth is flat. I think we're showing pretty good growth. But keep in mind we have many storage segments. Some are going quite fast like scale-out NAS, growing incredibly fast. One of the opportunities when you have a portfolio is that some things are hot, some things are less so.
What's your opinion on the more recent acquisition of Cache IQ by NetApp?
NetApp is under pressure to expand their market. If you go to Wall Street they'll ask "Where is the growth going to come from?" It's an interesting acquisition, as very few NetApp acquisitions have worked out well. But the question for me in on the NetApp side, less about the technology.
Why did you sell the majority of ownership of Iomega to Chinese Lenovo?
We're very interested in the Chinese market, and it makes sense to partner your way into the Chinese market. Lenovo has a strategy to grow from the bottom up into the storage business. We saw it as a two-fold. One: it's a great way to partner our way to the Chinese market. And second, most of the consumer electronics comes from China. It’s a strategic play but also a business Lenovo wants to invest in: getting better into the Iomega range.
Is EMC going to continue selling Iomega products?
Perhaps the higher-end. They also do consumer stuff, we're not so interested in that. But for example they do a very nice single controller.
What's your reaction if I tell you that the best quality of EMC CEO Joe Tucci is to recruit around him at high price excellent executives?
Joe has done a amazing job in building an executive team. I work with all these guys -- it's like a dream team. Do you want a cheap team or do you want a good team? And they work together as a team. I get to work with some of the best executives in the industry.
I don't think there is such a thing as cheap leadership -- we're in a very dynamic part of the industry, very competitive and changing very fast. This is not a stable, predictable industry. And we're not just storage. A lot of our executives came from other companies and we're glad to have them. Certain vendors out there are not doing so well.
What's your bet concerning the successor of Tucci as CEO?
Any way we win, there are so many good leaders there that I think our stockholders, our customers, our employees win for any of these executives choices. No bet from me. Do I have my own opinion? Yes. Does it matter? No. We've got six or seven great guys here, any one of them could be the next CEO and we'd be good. Dave Goulden, Pat Gelsinger, Howard Elias, Jeremy Burton, Paul Maritz, every one could be the next guy.
You were at EMC when Moshe Yanai, Mr. Symmetrix, was there? Why did he quit as he appears to be later the founder of remarkable start-ups in storage technology?
EMC was growing from a single product company to a multiple product company. Some people are comfortable with that, some people aren't. As soon as they were multiple horses, Moshe was not so happy. He didn't like the acquisitions but we had a situation where we, as a company, had to build a portfolio -- we couldn't be a single product company anymore. He always liked SRDF that was his baby so it's no surprise he's involved with Axxana.
(note: the interviewer brought up Axxana, I didn't)
What are more precisely your relation with Axxana?
(Note: I believe they might be in EMC Select). We have a program called EMC Select where we will sell on the same purchase order third-party products. We don't directly support it, we don't engineer it, but in creating a solution with EMC product and non-EMC product, like tape. It's a nice value add on top of SRDF with little extra protection.
Final Thoughts On This Experience
Part of the opportunity – and challenge – of being a public spokesperson is that you’re not always in control of what gets written. And, of course, there’s no “undo” on the web.
That being said, I hope you don’t mind me taking a few minutes to clean up the interview, and better share what the discussion was about, make it perhaps a bit more readable.
And I’m looking forward to my next interview with Storage Newsletter :)