Thursday, January 24, 2008

Sun acquiring MySQL: Something positive to blog about

The Acquisition of MySQL by Sun finally gives me something positive to blog about, I will try to explain in good detail why I think this is an excellent move by Sun.

Sun doesn't want to make most of its money from Hardware. They don't want to sell software either. What they want is to use their Hardware and Software products as marketing drivers for their services. The expenses in both categories pale in comparison with the amount of work and expenses needed to have hardware and software running smoothly in large configurations. To be able to effectively help companies to have smooth systems is something Sun is increasingly good at, therefore, their prospects look promising there and it is natural to be on the look for acquisitions to strengthen the company position.

People who has done Systems Administration probably would agree with me that Free/Open Source software gives you the feeling of being little bit "raw", or intimidating to people who is not master of a given package. This is where experience is valuable and companies such as Sun may provide assistance. On the other hand, once you are past what I call the "infantile mortality region" of the learning curve (when you do not have enough understanding of the software to make it useful for your purposes), Free/Open Source is much better than proprietary software: it does not imposes its policies on you. Proprietary software forces you to accept its policies, beginning with the "shrinkwrap" license on. For large organizations, that have non-mainstream requirements, frequently proprietary becomes increasingly frustrating, the "zealotry" of protecting licenses and stuff makes it much more complex to use third-party tools to assure corporate-wide compliance to policies, that is, proprietary software may become an obstacle for selling products and services that a compay such as Sun provide.

I think Java provides a good example to explain the principles that justify the MySQL acquisition. This wasn't obvious to me, it took me several years to understand how or why Sun may make money off things like Java, that are Free with capital F: In continuation with what I described above, Free software helps a vendor to create a free marketplace where products and services move fast [ free marketplace => market liquidity => efficiency ] where the best positioned vendor enjoys several advantages, and the best positioned vendor tends strongly to be the originator of the free software. I would say that while Sun invented Java, it didn't realize its full potential until recently, because, as I explain in "multiple personality complex" [ old enough to be the sixth article in this blog ], the original strategical role that Java was supposed to fulfill was ill-conceived: My interpretatin of "Code once, run anywhere" was old Sun's master plan to bite into the mindshare of software developers who only developed for x86 and where slighting Sun's hardware, but what really happened was that the few developers for Sun's boutique platforms used Java to free themselves of Sun's platforms [ I would say that Sun felt secure of its platforms 'cos Java would mean a significant increase in computing capabilities demand for the same software, therefore their reasoning could have been that software migrations to Java would be accompanied by hardware upgrades, since Sun commanded the high-ground, Java would highlight Sun's competitive strengths ]. I got to understand all the potential (for Sun) of Java not by looking at Sun, but becoming aware of how IBM had outcompeted Sun in the Java arena. From late 2001 up until late 2003 I worked heavily with Java, and I think this was a period of particular significance because it marked the sunset of IBM's Visual Age and the sunrise of Eclipse/Websphere while Sun was merely repackaging Forte as Netbeans. First, it came the observation that IBM had outcompeted Sun at Java, since the fact was very weird to me, it made me meditate long about how come IBM embraced so wholeheartedly a direct competitor's key product and turned it to its advantage. I had the advantage of a fairly good understanding of what IBM was looking for by embracing Linux, so, I was able to come to a conclusion by early 2006, soon after I wrote "multiple personality complex" [ like I said, it literally took me years! ]

It is nearly indisputable that IBM's transformation from a Hardware company into a Software services one was the success story of the 90's. Sun is also going that route, so, it is important to continue to discuss what IBM did. One of the key elements of that transformation was the development of a free marketplace where offering the services IBM could offer made sense, that's why IBM leveraged its participation in thus far compartmentalized markets such as PCs, Client/Server workstations and mainframes into a continuum of offerings. To accomplish this, IBM needed "glue" to hold things together when the customer wanted to shift gears up (or down) in scale, techniques focused on scalability. From old times, Virtualization, so that applications could be moved upward or downward without trauma. Settling on Linux for the Operating System wherever possible, because Linux runs well enough from embedded devices to mainframes, so that the Operating System won't become a problem, and Java for the applications themselves: "Code once, run anywhere". Since the PC market is a free enough marketplace already, it makes total sense for IBM to get rid of it if the corporate energines are better spent elsewhere, thus the selling of that business to Lenovo.

Coming back to Sun, now it is clear why Sun helps to develop its own market of services through not just Java, but Free/Open Source databases such as PostgresSQL and continued participation on proprietary databases alliances like IBM's DB2, Oracle and MS SQL Server, the whole point is that the database choice shouldn't be an obstacle. Just like the Operating System shouldn't: to continue to offer Solaris, the continued participation in directly competing Linux, and even Windows!. More recently, new offerings in Virtualization and now the acquisition of MySQL.

I would say that Sun's realization of these ideas was incremental, things like open sourcing Solaris could have happened much sooner, but I think they played a role in selecting Jonathan Schwartz as CEO, and by the time Schwartz was appointed the plan was very clear.

I read Dvorak's severely critical "The Sun-MySQL deal stinks" blog on the subject, and while I think he missed the whole point (I think beloved Dvorak wouldn't understand why Sun insists on Java either), he has provided me with the most convincing argument that Sun had a coup with the acquisition. I have to paraphrase and summarize his blog: The acquisition stinks because MySQL is far outside of the core competencies of Sun, so, the billion dollars will "simply vanish over time" together with the strength of MySQL. Dvorak suggests that since the great beneficiary of a dead MySQL is Oracle, Oracle would be using Sun through their long lasting partnership as its "stooge to do the job" of killing MySQL. One of the points that Dvorak wants to highlight is that Oracle is killing MySQL on the cheap by preventing a bidding war between interested parties: those interested in killing MySQL such as Oracle and Microsoft, and interested parties in its continued existence such as Google and Yahoo. I think that a monumental biding war for MySQL could indeed have happened, so, Dvorak's argument assures me that Sun did a great deal.

As to why the MySQL people sold themselves on the cheap, I am not sure, but it must has to do with two things: 1) Just like Microsoft is famous for acquiring companies just to kill their incipient competing products, MySQL's value for Microsoft and Oracle is "how much does it cost to kill it?", and "if we kill it, how much more could we exploit markets before an alternative fills the void MySQL would leave?", so, I guess companies with "ill intentions" to acquire MySQL have a tougher negotiation ahead for not so great benefits; and 2) Net customers such as Yahoo or Google may not have the willingness to develop the product or its market, so, the value to them is not so high. But Sun is particularly well positioned to make the most of MySQL on both senses: On behalf of its customers who want good and cheap databases, and on behalf of MySQL in terms of product development and corporate backing; this happened with IBM/Linux to the great benefit of the Industry and all the parties concerned, this acquisition will go the same route.


Scientia from AMDZone said...

It makes sense to me. Sun already has Star Office and the free version, Open Office. Open Office has Base, Calc, Impress, and Writer to compete with Access, Excel, PowerPoint, and Word.

We all know how popular the Apache webserver is. And, I suppose Xchange might be the equivalent email server. However, many applications require a medium level database server like Oracle or SQL Server. As far as I know, MySQL is the only free application in this class.

I can't think of why Dvorak feels so strongly that Sun can't continue to develop MySQL just as they do Open Office. As I recall, Novell was the one that purchased several applications and then ended up selling them to Corel for a fraction of the price. Novell has done far better with Suse Linux.

We know that Java has never replaced Windows on thin clients as Sun dreamed of doing. Oracle dreamed of this as well if memory serves. However, I think Star/Open Office is a great hedge against the Microsoft Works/Office monopoly and I don't see why MySQL can't continue to be the alternative medium scale database. With all due respect to Mr. Dvorak, who is better suited to handle MySQL?

Scientia from AMDZone said...

BTW, I have to say that although I disagree with you on a lot of points you do seem to have a reasonable point of view.

For example, I agree that RDRAM and FBDIMM have been bad because they make the memory more proprietary. This is why I thought micro-buffer would be such a good idea, because it would allow regular DIMMs to function as registered memory with serial communication like FBDIMM but with 2:1 fanout and therefore not as susceptible to FBDIMMs racetrack delay. However, with Bulldozer now delayed who knows when this might actually appear?

I would tend to disagree with your conclusions about ATI though. You say it was a suicidal mistake. I can't really see that. Overall, I think you were expecting results too quickly.

For example, AMD really didn't gain much market share with Athlon MP. So, by that standard it was a complete flop, right? Yet, without doing Athlon MP first I don't think AMD could have launched Opteron first with K8. I'm certain that Athlon MP gave AMD some much needed credibility in servers. I think in a similar fashion AMD is gaining credibility in commercial markets as a full supplier of both processors and chipsets. This is something that Intel has had for many years but AMD has not. I think this crediblity is essential for AMD to continue growing its markets.

Secondly, AMD has been hammered by Centrino ever since its launch due to Centrino's better battery life. This wasn't exactly fair because Turion actually had lower power draw than Pentium M. But, AMD was stuck with a much worse 3rd party mobile chipset. This year, AMD finally gets a real mobile chipset with Puma but I don't think this would have happened without the ATI purchase. And, AMD badly needs more mobile market because of the higher ASPs.

You mentioned that K10 scores badly on most of the benchmarks. This is true but you are also talking about desktop benchmarks and desktop is the lowest ASP. You also have no idea which of these benchmarks might have been compiled with the Intel compiler and there is no doubt that it produces worse code for AMD chips.

I think the Integer IPC for C2D is higher but I doubt it is 20% higher. I've seen tests where Intel's 20% speed advantage was cut in half with proper compiling. However, I've never seen this show up in reviews.

Your ideas about the L3 to L2 cache ratio are interesting. I keep wondering if you are talking about apps that use only one core though. I'm also very leery of applications that make heavy use of Intel's L2. These look great in benchmarks with only one app running but bog down under real world conditions.

Eddie said...

Scientia, thanks a lot for your comments, you made me realize that I forgot to mention Open Office and several other important things.

Sun is mounting a generalized attack on the concept of "Microsoft Windows as the Platform", to undermine it from all fronts. It may eventually succeed because the attack is using F/OSS weaponry:

Java to undermine Windows as an O.S. and specific platform independence.

Solaris as a high-end computing O.S.

Open Office to cut the vicious cycle of MS Office that leads to Windows usage that gives it the economies of scale to make it so popular.

And all the rest. I assume that this is not expensive for Sun to pursue, another key positive thing of F/OSS. Now, an objection that I thought after writing this article, is that you don't "buy" F/OSS projects to make money off them, you just become a developer, the same way IBM doesn't "own" Java nor Linux. I have to go back to meditate about why an acquisition. A pity, the article is otherwise very good, but it is incomplete.

<< would tend to disagree with your conclusions about ATI though. You say it was a suicidal mistake. I can't really see that. Overall, I think you were expecting results too quickly. >>

When the acquistion took place, I knew AMD had to try something rather radical, that is the only thing I was expecting. But I wasn't informed about graphics nor ATI. I had the queasiness that nVidia graphics were superior, that despite my ignorance about graphics, I can see the killer instinct in nVidia, and my skepticism for large mergers. I went with the flow not being able to formulate a better opinion; but upon closer inspection, this business decision is indefensible. A summary? let's see: Why would you pay cash for ATI if the acquisition itself was going to need lots of cash to speed up crucial projects like Fusion?, why would you acquire ATI if both ATI, and most importantly, nVidia, were forced by market pressures to fully embrace AMD and were already doing it? how are you going to compensate the negative synergy of alienating nVidia and putting obstacles for ATI to remain strong in the Intel based segment?. The demonstration that this was a suicidal mistake is that AMD slid into the prostration it is in now due to processor uncompetitiveness, but with the added pressure of a mountain of debt that is screwing the schedule of Capital expenditures and R&D, that is, AMD is lagging behind because the ATI acquisition tied it. Another demonstration is how far in the future is Fusion. For that, a joint venture would've been faster. This is a consequence of the acquisition trauma and the lack of funding, make no mistake about it. And finally, the nearly instantaneous lost of significant ATI market share followed by further declines.

I am replying to your comment in the other article too with some other things.