Smackdown on Barry Burke, EMC
by Claus Mikkelsen on Apr 20, 2009
Barry, Barry, Barry…watch your mud…it’s about to sling back in your direction….
I’m getting a little ticked off over this banter of what company was first with what technology, and Barry Burke, the self-proclaimed storage anarchist, made some comments on our own Christopher Bertrand’s blog that need some serious corrections. I’m not sure where he gets his information, but he should be questioning his sources. The topic here is replication, internal and remote. Let’s take a trip through history to set the record straight…Barry claims that EMC invented it all which is not only false, but he is now hiding behind his proclaimed “legal rulings” to justify his claims, which are also bogus.
The first intelligent function introduced to storage systems was a mainframe capability called Concurrent Copy. CC was GA’d in early 1992 and was based on a patent filed in early 1987. I know this because I filed that first patent. It was based on a technology I (we) called “write intercept” that everyone now wants to refer to by its bovine name of “copy on write”. That’s right, the first CoW patent is 22 years old.
CC essentially introduced intelligent storage subsystems to the world. This CC intellectual property was licensed to Hitachi, Amdahl, and STK and a couple of other companies that no longer exist (in addition to the now non-existent Amdahl and STK!). EMC did not license CC, but did come out with a “compatible” version about 18 months later. What’s interesting is that in the very early EMC documentation on CC, it basically said “see the IBM manuals for details”. Who copied who, here? You were 18 months late to delivering intelligent storage function and you had to refer your customers to your competitor for information?
What came next, over the next 4 years, is a little less crisp and I’ll explain why. Both IBM and EMC were developing remote replication technologies. EMC was developing synchronous SRDF and IBM was developing synchronous PPRC as well as asynchronous XRC. The reason these times were muddy was that early distance replication technology was truly groundbreaking and was having to deal with a lot of new concepts like surviving “rolling disasters” and understanding the underlying reactions of DBMS’s in response to these failures. Both companies called these products “GA” but in fact all 3 products (SRDF from EMC and PPRC and XRC from IBM) were in controlled availability as we were researching and learning in the process. So I’ll give PPRC and SRDF a “toss up” in terms of novelty and availability (actually, no, PPRC had some very specific technology advantages that showed an understanding of data integrity issues that were absent in SRDF, but perhaps I’ll leave that for another blog), but XRC was a clear winner in asynchronous replication and a “first”.
Then came the “bombshell of 1996” when we all woke up one morning to learn that IBM was going to OEM the STK Iceberg. The threat here, to EMC, was that the Iceberg, because of its log structured file system, was able to deliver the first true “snapshot” capability and it was now in the domain of IBM, not just STK which, at the time, was still a relatively minor storage player. IBM now had 2 ways to clone volumes: the just discussed CC, and now Snapshot. IBM marketed Snapshot very heavily, integrated it into CC via the Virtual Concurrent Copy (VCC) interface. EMC’s response to this threat was TimeFinder. TimeFinder, by the way, was again NOT new since both IBM and Hitachi had the same capability since the early 1980’s called “Dual Copy” (now called RAID 1). But EMC did demonstrate that they, too, could clone a volume. Wow. Really. So far, not a not of “newness” here from EMC’s side, eh?
This era was mainly a mainframe IBM/EMC battle. HDS’ interest in this was to stay compatible with the IBM mainframe on both the processor side and the storage side. Thus, this period, especially with replication technology, was all mainframe. But things started changing in the late 1990’s when Hitachi, after licensing the XRC IP from IBM, decided to get aggressive in delivering technology not locked to the IBM mainframe. True Copy Asynchronous (TCA) in 1999, which supported both mainframe AND open systems, was the first step in this new strategy. Wanna bet when SRDF/A was delivered? Yes, 2003! That’s 4 years after HDS’ TCA!! Who is leading the technology here?
Now about Barry’s comments on EMC’s multihop solutions…this, I would argue, was to address the fact that EMC had no distance replication technology to counter TCA. Multihop was nothing but a bunch of complex scripting around SRDF and TimeFinder that, through iterations of suspend/resume, and as many as 7 volume copies (sometimes more), delivered remote replication in the absence of an asynchronous technology. It was basically “bucket brigade by Rube Goldberg”.
Now, Barry also makes reference to certain legal actions and who won what, but again, he is, again, completely uninformed. In addition to being involved in the design and architecture of all of the technology I have just quoted, I was ALSO involved in EVERY legal engagement between IBM and EMC, HDS and EMC, and even a very brief one between HDS and IBM. All of them ranging from defending IP, giving depositions, serving testimony, and actually negotiating the settlement in the last 2 between EMC and HDS, and in no case, at no moment, was there ever an acknowledgement, conclusion, or ruling, that EMC preceded either HDS or IBM in any replication technology. Never! I would love to disclose more detail on these events but they are still embargoed and THAT would get me into a ton of trouble, and in spite of that fact that many of my friends are attorneys, I never, ever, again want to get involved in another legal squabble between a bunch of storage companies.
So, Barry, please be careful with your facts, because they are not even close to the mark. We all get caught up in our marketing campaigns, slogans, and claims. In fact, a while ago at a conference, your EMC speaker even claimed EMC “invented” the term “rolling disaster”. Did you also invent sliced bread?
But anyway, I’m here to refute pretty much everything you claimed or implied in your comments to Christophe’s blog. You guys have excellent marketing and no one will doubt that. You also manage to get the most out of your (somewhat delayed) technology. But pretending that EMC has been this technical dynasty that is inventing and leading all the way is a bit much for me to stomach, I’ve been in this industry for 43 years, I’ve experienced a lot, and I know better.
Comments (7 )
Your history is correct by my participation as a Software Systems Consultant for Amdahl Canada from 1980 and later HDS to 1999.
Even in those days, EMC FUD was easy to refute with just their own announcement material.
It is always revealing to see what perts of a comment a competitor responds to. Here, your gleeful “smackdown” defends to the 9′s your personal perspective of who was first, which in the end matters not to anyone but you.
I’m still standing…
How about responding to the stuff that is actually important to customers?
To refresh your memory, I went on to say…
And “By the way” (as you say) – since TSM requires customers to stop their replication sessions for the duration of relocating a LUN, said customer is out of Compliance and has to notify the SEC before and after every move. SEC phone calls are never a good thing…fortunately for the Financial Services industry, V-Max Virtual LUN migration doesn’t suffer this limitation.
In fact, V-Max VLUN can concurrently relocate up to 128 times more LUNs than TSM (1024 vs. 8), moving them at least 2.5 times faster than TSM, and with near-zero impact on running application performance during the move whereas TSM can double or even triple response times.
V-Max didn’t COPY USP-V – it simply does similar things BETTER, FASTER and EASIER.
[...] 4) Your legal comments about REPLICATION were TOTALLY MISLEADING. I think you already read Claus’ blog. [...]
Claus’ is spot on. Prior to joining NetApp, I worked on the customer side and lived through the testing of nearly all of the products mentioned above – XRC, PPRC, SRDF, Timefinder. In fact, one of the reasons I appreciated the NetApp story prior to joining the company in 1999 was due to my appreciation for the Iceberg (snapshot) technology. As a customer, our team was really disappointed to see it basically fade away when the IBM/STK relationship fell apart.
Regardless, I would suggest that the point Claus is making, Barry, goes to credibility. When you’re factually inaccurate it casts a big shadow over your other product claims. There are some good things about the V-Max but, after the big marketing build up for the Liger, all we essentially got was a big Symm. Nothing wrong with a big symm but, there seems to be an attempt in your post to Chris’ blog to make a link between innovation at EMC and a new innovative Symm. EMC did some great work in popularizing intelligent storage. I just think you picked the wrong battle when it comes to replication technology. For those of us with a little snow on the roof, we have a sense for the history behind this topic and that’s where you lost us. Your basically, well, wrong with your recollection on replication innovation.
Look, NetApp didn’t invent Snapshots, but we certainly helped popularize it in open systems. There are patents that surround how we take snapshots but going into a sales meeting I don’t say we invented snapshots because someone in that room can probably remember Iceberg and whether they call attention to it or keep it to themselves, I would lose all credibility if I stood up and claimed, “NetApp invented snapshots.”
I stand humbly corrected.
[...] on is the “who introduced what, first” debate. I thought we had that issue resolved in my “smackdown” post earlier this year. I say that because the final comment on my blog was from Barry and it [...]
Apparently the folks that read and respond to these BLOGs are much younger than Claus and I…
As I recall Concurrent Copy was a technology solution developed to protect against the common (in those days) disk crashes (can you say 3380 and plenum) IBM customers faced. CC did not provide any of the “ease of use” features customers have come to expect from array based local replication technology and has fallen into disuse because of the lack of these capabilities.
What is gratifying and sad to see at the same time are the references to the Iceberg and later the IBM RVA (aka RAMAC Virtual Array.) Yes STK’s Snapshot offered great promise and was an advanced technology for the time. What most people fail to recognize, or worse yet admit, is the significant scalability limitations the Iceberg/RVA had. IMHO in today’s world of limited environmentals the Iceberg would not provide an acceptable tradeoff between capability and environmental costs. Another issue that Iceberg/RVA customers faced was the significant performance impact of the custom inline compression/decompression ASICs. Even with IBM’s massive effort to support the RVA, the Iceberg/RVA engineering was not sustainable. There were simply to may code limitations with no way forward at an acceptable engineering and materials cost.
There is a reason that companies are successful and IMHO BLOG’ing is not one of them… However you BLOG’er keep on slinging, while ignoring actual facts because it is much more entertaining this way.