A few days after my first visit to Oracle OpenWorld in over 15 years, I’ve taken a few days to digest the experience and wanted to summarize my thoughts and “analysis”. Attending with my wife, who still works for Oracle and is a director associated with Exadata POCs, was also a fun time – between her contacts within Oracle and my “contacts” in the outside ‘”Oracle underground” (as my wife refers to things like the OakTable network) I think we were able to get a nice, full picture of the conference.
I thought there were many interesting “themes” at the conference, from the official ones like Exalogic and the Fusion Apps to unofficial ones like how Exadata has been changing the existing dynamic around databases in the enterprise.
Exalogic was an enigma at the conference – as the opening keynote announcement, and repeated at the closing keynote, one would have thought that there would have been a lot of buzz about the box. Instead, I saw a fair amount of confusion as people debated the meaning of “cloud” vs. “box” vs. “appliance” vs. “platform”. Information about the configuration was scarce, and many people at the conference seemed to ignore it completely. From initial descriptions (Ellison’s claim, for example, that 2 of them could handle all of Facebook’s web traffic) it appears that Oracle has built a high-performance, high-density Java platform, that, coupled with several Exadata database platforms, could easily handle most, if not all, of an enterprises application and data hosting needs. It remains to be seen if organizations could actually run their entire application portfolio on a set of Exa-platforms – and non-Oracle software “support” seemed to be “lacking” on the Exalogic. (I visited the Informatica booth to ask them if PowerCenter could run on the Exalogic – their response: “What’s an Exalogic?”, was telling…)
I opined at the conference that Ellison was trying to replace the generic hosted “LAMP” stack with and “LJE” stack: Linux, Java, Exa. And in the main, I think it’s a good idea – the ability to provide a private shared hosting environment within an enterprise is attractive. (Full disclosure – I used to work for Network Solutions, so I have a fair appreciation of hosting environments). It was interesting to see the response to my “tweet” on the subject, as several people shot back how an LJE stack is so much more expensive than a LAMP stack. And for small applications, they’re right. However, contrast the size of the Facebook LAMP stack with how Ellison described a hypothetical Facebook LJE stack (2 Exalogic platforms, several Exadata platforms) and I’d guess that the price difference would be less than most people guess – not to mention a lot more energy and space efficient. As an example of the hosting paradigm, I enjoyed a presentation by Nicholas Tan (Commonwealth Bank of Austrailia) and Roland Slee (Oracle) in which they combined 300 databases into 1 database and started providing internal database hosting with dramatic cost and time savings.
In any event, the rest of the conference seemed to be more focused on Exadata, as more and more customers are coming to grips with it. Unfortunately, I noticed several “Exadata” presentations that had little Exadata-specific content – as Cary Millsap tweeted, it appeared to be a magic word that got presentations onto the agenda. The 2 new Exadata boxes were nice technology refreshes, and Alex Gorbachev has a good side-by-side comparison of them. On an another note, I wondered if Oracle was introducing Exadata models “too quickly” – I heard rumblings from attendees with v1 HP-based Exadata boxes about how they’re being left behind a bit.
As my first conference trying to fully utilize the power of Twitter (I know, I’m late to the party), I was happy with how it semi-allowed me to “attend” several sessions at once vicariously through other people. Greg Rahn and Gwen Shapira’s tweets kept me in the loop around other sessions all day. In particular, I was particularly happy to be following the analysts during their “private” sessions with Oracle – Merv Adrian and James Kobelius were very insightful and “non-snarky”. James’ tweets around the Exadata Intelligent Data Warehouse (whereby SAS PMML models could be run directly on Exadata platforms) were especially interesting to me.
In the end, I started to see somewhat of a sea-change occurring in more of an emphasis on applying the power of Exadata to problems rather than advice on how to technically configure them. And I think, Ellison agrees with that – one of his statements that customers shouldn’t be defining them selves on the basis of how unique their configurations are, but rather on what they are doing with them. Of course, that kind of statement starts to imply a reduction in configuration choices – run whatever you want as long as its on an Exadata or is running Oracle Linux or Oracle Solaris (x86) — (I went to a session on Database Flash Cache in which Oracle all but admitted that the feature was limited to operating systems under Oracle’s “control”). And Ellison’s pointed comments about Red Hat Linux didn’t go unnoticed by the crowd either. In any event, choice in the hardware space has been narrowing for some time, as non-Intel and non-IBM architectures continue to decline in the market (with the exception of the ARM-based “systems”). (BTW, speculation around Oracle and ARM has been entertaining). Ellison’s rather clear desire to provide pre-built solution platforms is also a subtle comparison to IBM and its large services arm – it will be fascinating to watch how this plays out in the market.
This narrowing of choice is continuing into the DBA space, as I believe that range of ways a production DBA can affect the performance of applications continues to diminish – and not in a bad way, but rather in the way that the database defaults are getting better and better and reduce the configuration options that DBAs need to manage. From Greg Rahn’s comments about how Exadata defaults seem to handle most reasonable workloads without difficultly, to Alex Gorbachev’s recommendation to use OMF to Tom Kyte’s presentation on how smart the optimizer is getting (how it eliminates tables from queries when it determines that they are not necessary) it’s becoming clear to me that the real ability to affect performance is shifting from the production DBA to what I term database engineering and/or development.
Watch Cary Millsap’s interview with Justin Kestelyn and you’ll see what I mean – I think it’s imperative that DBA’s who are interested and talented at performance work to become more involved in software development. Either by becoming dedicated development DBAs or forging ahead into database engineering. I had a good, quick talk with Gwen Shapira about this at the blogger’s meetup. And I was also struck by information from the Real World Performance team sessions in which they showed how application architectural choices affected design – Greg Rahn spoke about how necessary it was to start thinking in terms of set processing in order to harness the power of multiple compute and storage nodes; and in an understated demonstration, Andrew Holdsworth showed how fewer connections to the database can result in more scalability. These all speak to development and application choices in my mind.
Finally, I had a good time trying to attend sessions at the Hilton for the Oracle Develop conference. Cary and I talked about the wisdom of developing plug-ins for SQL Developer vs. Eclipse and I was pleasantly surprised to see Quest announce their Toad Plug-In for Eclipse at the conference. With the demise of SQL*Plus for Windows and the need to integrate more database development tools into the developer experience, these discussions (not to mention the reduction in price of SQL Developer’s Data Modeling tool to free!) really hit home for me – now only if we could get better SCM integration – perhaps Eclipse with the Quest Plug-in, Method-R MR Trace Plug-in and Mercurial Plug-in will do it for me…
(I like Mercurial because it emphasizes changes and change-sets – which I think is a better match to database refactoring than Subversion – but that’s a topic I’m still researching).