Oracle: Mine is bigger and, um, more integrated
Ellison flaunts Sun unit
Lucky Sun shops
Luckily for Sun shops, Oracle doesn't have its own Unix already, a situation that Compaq's (well, Digital's) customers faced when Hewlett-Packard ate Compaq a decade ago. The PA-RISC and Alpha chips made by HP and Compaq were already on their way out, to be replaced by Itanium, and Tru64 Unix's days were numbered because HP had its own HP-UX. (HP had planned to graft Tru64 Clusters onto HP-UX, but high costs and a bad recession forced it to simply use Veritas tools and forget that promise.)
Oracle will be able to give customers the alternative of Solaris and its clone of Red Hat's Enterprise Linux, and on x64 iron, it will be able to knock together a commercial-grade Xen hypervisor stack from Sun's xVM and the recently acquired Virtual Iron tools. (Oracle has to decide what to do with VirtualBox, and selling it off to Novell or Citrix might be an interesting idea). Sparc customers will be relegated to dynamic hardware domains on Fujitsu iron and Logical Domain (LDom) virtual machine partitioning on Sparc T boxes. The Solaris on Sparc and x64 variants will also have Solaris container virtual private servers, and Oracle has to give lip service to VMware's ESX Server and Microsoft's Hyper-V too, now that it is a Windows server supplier.
The Java middleware stack from Sun is very likely to be mothballed, with the exception of identity management features, which will be Borged into the Oracle middleware mess. Java will probably continue on its current trajectory, as Oracle has to be careful to keep all of Sun's Java partners and the millions of Java programmers in the world happy.
Sun's storage business, which is a constant source of hope and despair, has some interesting possibilities. Chris Mellor, El Reg's intrepid storage reporter, says that Ellison should get out Oracle's checkbook and buy Pillar Data to provide itself with a decent midrange storage lineup.
Ellison, of course, has funded Pillar through Tako Ventures, his personal venture capital company. You can check out Pillar's latest announcements from this week here. It is hard to say what Oracle should do with the Sun open storage products, which mix Sun servers, lots of disks, Solaris, and the Zettabyte File System. Clearly, Oracle will splash flash all across the Sun hardware line like some kind of magic performance-enhancing drug.
It is hard to say what Oracle should do with Sun's InfiniBand switches, beyond using them in products such as the Exadata parallel database cluster announced last September by Oracle and Sun or the Constellation HPC setups that Sun has been peddling for the past several years. But one thing it should probably do, if it wants to be a credible integrated systems provider, is get 10 Gigabit Ethernet switches into its product line immediately and support the Fibre Channel over Ethernet (FCoE) and Fibre Channel over InfiniBand (FCoIB) protocols to get server and storage traffic running over the same backbones as Cisco Systems is doing with its "California" Unified Computing Systems.
One ironic possibility is for Oracle to acquire Arista Networks, the upstart 10GE switch maker, which has Andy Bechtolsheim, and on-again, off-again chief technology officer at Sun, as its founder and CTO. Those "Magnum" InfiniBand switches and the related x64 servers and open storage arrays currently peddled by Sun came from another Bechtolsheim company, called Kealia, that Sun bought in 2004 shortly after Jonathan Schwartz took the helm.
It is hard to say what Oracle will do with MySQL, but its natural place is in Web infrastructure systems and Oracle would be a fool not to leverage its popularity in some kind of integrated system for hyperscale Web application deployments. One of the reasons why the EU let the Oracle-Sun deal go through is that Oracle made commitments to continue to enhance and support MySQL, and there is every reason to believe the company will make good on its promises.
Lawyers are always looking for a reason to take on a big foe with deep pockets, and they can talk governments into funding antitrust lawsuits that can drag on for years. Oracle doesn't want any of that and will most likely behave itself as long as MySQL remains popular.
Whatever Oracle plans, we will all find out a little something on January 27, when the company hosts a marathon five-hour Webcast explaining its hardware and software plans. Expect lots of subtle comparisons to - and snide comments about - the integrated systems being peddled by the Cisco/VMware/EMC triumvirate, by the Hewlett-Packard-Microsoft tag team, and by IBM. ®