Oracle discovers the power of the two socket server - or: A pivot that wasn't one, TCO still rules

Last week Oracle provided an update on its engineered systems – mainly the launch of the new X-5 line. When Chairman and CTO Larry Ellison took the stage – he had a surprise in the bag that few counted on – that Oracle would now start to compete in the space of two socket servers.

So let’s look at the implications:

It is worth to understand why this was a surprise – as in the past Oracle had focused on large, high end and high price engineered systems. The Exa line of products was a compilation of only the best that Oracle had in store after the acquisition of Sun. In numerous earnings calls Oracle’s executive team defended the strategy, while Oracle’s overall hardware revenue and profitability was suffering from lower end and cheaper servers that Oracle by now has gotten rid of.
 

Oracle's new view of the datacenter - Core (2 socket systems) and Engineered Systems (Source - Webcast)

For the first time Oracle and Ellison acknowledged, that next to the high end Exa machines, there are other needs in data center purchases, specifically around the two socket server aka low cost server demand. These machines are popular as they can handle any kind of load an enterprise may have, are cheap to procure, easy to operate etc. In contrast to that buying an Exa machine is significant CAPEX for even the largest enterprises – a decision that needs to be vetted, often employees need to get trained to operate the machines, sometimes application have to be re-written to take full advantage of them etc. Throughout this Oracle maintained that its Exa machines have always offered the best performance for the lowest cost. Turns out to get the machine was a pretty high hurdle price tag wise for enterprises.

So not surprisingly Oracle now announced a new product line, a two socket server, but true to Oracle’s corporate DNA - at a lower TCO. It looked like Cisco’s servers are the prime target and market leader that Oracle wants to challenge. True to long time tradition Oracle was not shy to compare purchase costs of its new machines compared to the machines of competitors.
 

Oracle compared VCA X5 with Cisco UCS and claimed significant TCO advantages (Source - Webcast)

But what we really have seen for first time is Oracle flexing its muscle around its ability to own the complete stack. While main competitor target Cisco has to partner with EMC and more vendors to get their machine to the market, Oracle has all the necessary pieces available in house (Storage – SSD of course, Networking – Infiniband (even Ellison joked how often he said that, Processors (surprisingly little said about that), Oracle Linux, Database (of course) and more higher level software). Reshaping the design of an engineered system to a different load profile and price point is easy when you have all the pieces available in house. The value proposition remains the same – one integrated and tested system by Oracle – just at a very different usage and a much lower price point.

Not surprisingly Oracle maintains that ultimately the large Exa machines are still better performance at a lower price. But the entry point for the new two socket server line is lower, and if customers ultimately may pay more to Oracle when buying a lot of two socket machines over time, nothing that Oracle has to complain about. Think renting vs buying a house. Plenty of people cannot afford a house, so they rent, but over time pay more for rent than for owning a house.

The question really is – what took Oracle so long? We can only speculate on the reasons and motivation – but here are a few:

• Oracle was still operating under the ‘big server for RDBMS, small machines for web tier’ paradigm of the 200x years. 
• Oracle has listened to customers (and sales reps) that selling large machines only, will not give Oracle the growth it wants to see in data centers.
• Oracle did not see it share of wallet on server / data center spending really growing.
• Oracle had to finish commitments and work on the high end side of the Exa machines, e.g. finish the 4 series. 
• Oracle had to get software components ready for the two socket market, 12c comes to mind as a candidate. 
• Oracle is serious about cloud, but needs two socket servers for its own cloud build out and did not want to buy them from the competition – but buy and built by itself. Ellison stressing the same architecture and versions in a side note of his presentation made this pretty clear. 

We will probably never know – but above seems to be plausible reasons to me.

MyPOV

A good move by Oracle, which will expand its data center footprint. Oracle will have to overcome some positioning with customers, but if Oracle can deliver and prove the very attractive TCO numbers it shared at the launch event, then CIOs will quickly forget ‘old’ positioning and look at Oracle’s brand new two socket server line. And more competition for servers is a good thing for IT budgets that are under stress every year.

But the real question remains – what took Oracle so long?

----------

Also worth a look for the full picture

• Market Move - Oracle buys Datalogix - moves more into DaaS - read here
• Event Report - Oracle Openworld - Oracle's vision and remaining work become clear - they are both big - read here
• Constellation Research Video Takeaways of Oracle Openworld 2014 - watch here
• Is it all coming together for Oracle in 2014? Read here. 
• From the fences - Oracle AR Meeting takeaways - read here (this was the last analyst meeting in spring 2013)
• Takeaways from Oracle CloudWorld LA - read here (this was one of the first cloud world events overall, in January 2013)