I won’t be the least surprised to see that headline sometime in the next couple of years. Let’s take a look at recent trends in tech to see why:
VMware
Prior to VMware, no one knew what a hypervisor was outside of banking folk who deal with IBM AS/400 and LPARs. The rise of Open Systems and the x86 processors they ran on had gradually taken over all things midrange, with a few proprietary holdouts (AIX, Sun Solaris, and HP-UX). Serious companies were running serious application on Windows servers with x86 inside. Even Solaris had an x86 version, and Apple were moving there after their re-Jobs driven phoenix impression.
x86 hit critical mass with an excess of power around the same time that VMware really hit its stride. Citrix’s earlier attempts had gone basically nowhere, but with VMware it was suddenly easy. Everyone needed more servers, but were running out of freon, floor tiles, and electrons. It was just simpler to virtualise a dozen physical servers running at 4% CPU, so we did. VMware’s stock price went kaboom, and a bunch of copycats emerged, none of whom are even close to stealing market share from VMware.
Blade Chassis
x86 CPUs begat small RU ‘pizza box’ servers. We piled them into racks, row after row, the noise of cooling fans deafening all who came near. Then some bright spark though “hey, why are we encasing a motherboard, CPU and memory in all this extra plastic that keeps the heat in, and then sticking them in a rack? What if we just joined a bunch of them together and then stuck them in a rack?”
So we did. And now we had vertically mounted server blades in an enclosure with shared backplanes and pass-thru network modules and a bunch of new fun and games to deal with.
HP and IBM, and to a lesser extent Dell, made shedloads of money selling us x86 PCs in a new, datacenter friendly form factor.
Cisco UCS
All this time, Cisco had contented themselves with owning the IP networking market that drove the entire planet’s internet. They introduced their MDS line of FibreChannel switches because taking Juniper’s marketshare in dribs and drabs must have been getting boring. They’d begun to address the serious inconvenience of Brocade having any marketshare at all, when suddenly someone in product development must have thought:
“Hey, these blade chassis things look an awful lot like 6509s. I bet we could build a 6509 that took x86 motherboards instead of networking linecards.”
So they did.
Only they did it really, really well. No personality, all in software. Integrated with all the Cisco networking gear you know and love. IP, FC, iSCSI, FCoE, FCoTR, you name it, we got it. Internet scale, Cisco quality, price competative. Why would you buy blade chassis one at a time and then buy network kit from us, when we can just give you the lot for a bundle price? Buy in bulk and save!
The stack wars were on, and Cisco had used full cannon to fire an impressive opening volley.
The Stack Wars
We’re now in the opening to middle stages of what I’ll call the stack wars. The battle has been fully joined for a winner-take-all fight for marketshare in owning the entire infrastructure stack in company datacenters. HP fought Dell for 3PAR to fix holes in their storage line, IBM weren’t getting anywhere with XIV so they created the V7000, Cisco are partnering with anyone and everyone. VMware/EMC haven’t (yet) bought a networking company, but they’ve set up the Acadia partnership with Cisco.
Each keiretsu is trying their hardest to convince companies to buy all their VMware infrastructure from them. It’s not enough to just do one piece of the puzzle and interoperate with a bunch of other people. The way of the future appears to be to have preferential business alliances with a set of complimentary infrastructure companies so that you can guarantee a small slice of a bigger pie. Because the ultimate winner is going to get most of the pie.
Enter Oracle
So what have Oracle been doing all this time?
Oracle bought Sun, who made a bunch of those pizza boxes we mentioned earlier, and also made the M8000. They had SPARC and x86 based server platforms, running their own O/S: Solaris. Sun had just bought StorageTek, so they had also had tape and VTL devices. They also just settled the ZFS/WAFL lawsuit with NetApp.
A few years before that, Oracle bought PeopleSoft. They own the RDBMS market, and have their own flavour of Linux: Unbreakable Linux.
With the exception of networking gear, Oracle own all the components to run your enterprise, including the software. IBM is the only other company with all that, and companies tend to run SAP on their gear, so it’s not a full stack. HP is maybe third because they have 3Com network gear that almost no one uses any more, partner with Brocade for FC, and just bought 3Par. HP also partner with SAP.
So I’d look for two things in coming months/years:
1) Oracle buy an IP networking company. IP would be best, since it means they can connect IP storage to Oracle over pNFS via Direct NFS. Plan B would be to use FC networking and ASM, but that’s not where the industry is heading, unless you count FCoE.
2) Oracle bundle it all together as a ‘Data Appliance’.
Oracle will thus own the entire infrastructure stack including the database and the applications business run on all that gear. And they do cloud.
I put it to you that Oracle are poised to become the new IBM, selling you the new mainframe known as cloud. Sorry, leasing.
What’s not to love?