Death of the mainframe - again
- Summary:
- John Appleby takes on the behemoths as he predicts the end of the mainframe. We've heard this before. Is now the time? Judge for yourself.
Two years ago, I called the HP Superdome out as being as dead as a dodo. According to Gartner, HP is the largest server vendor by both revenue and shipments, but there are several reasons to think that it's time to put the final nails in the mainframe coffin.
First, HP Superdome revenues continue to decline – of the recently recorded $3.4bn sales, just $228m was Business Critical Systems (BCS), down 25% year on year. Second, HP are investing in Intel Xeon based NonStop systems – replacing the Itanium, and these systems will fall into the BCS container. HP has a history of building excellent business critical systems, so this is very interesting news.
While everyone else was sleeping, Huawei grew 200% year on year to take fourth place by shipments after HP, Dell and IBM. Note that this isn’t fourth place by revenue, because Huawei is commoditizing the x86 market. The top three should be very afraid, because while Huawei's sudden ascendancy might surprise, it hasn’t gained significant inroads into the EMEA and North America markets – yet. Their current sales and marketing is very much Asia-focused but you had better believe this will change.
IBM has reacted by selling off what it calls low-end, the x86 System X server business, to Lenovo. This means that IBM isn’t really an Intel partner any more, which means it is repositioning itself to compete with Intel head to head on enterprise systems. As part of this strategy, IBM are trying to position Intel as low-end. To quote IBM Leader Alfred Freudenberger:
“Are you ready to bet your business on x86?”
If you’re willing to look at what Amazon, Apple, Facebook and Google are doing, the answer is unequivocally “Yes”. This matters, because Google was the #5 server vendor in the world in 2012. I don’t have evidence to back it up, but I’d bet that Amazon, Apple, Facebook and Google are numbers 4,5,6 and 7 today (not necessarily in that order - but you get the idea.)
Oracle is hedging its bets for now, having acquired Sun Microsystems for mainframe systems, but also building its own Intel-based Exadata servers, that are designed for mission-critical workloads. Despite this, Oracle has server revenues that declined 16% in 2013.
IBM doesn’t split out revenues for its different server groups, so we don’t know what portion it has retained. IBM certainly argues that businesses that run mission critical applications require the stability of its mainframe systems. Historically, and despite much pontificating to the contrary, IBM has consistently driven up MIPS over the years.
Redefining the server market
My prediction is that the server market will radically shift in 2014. The bottom will fall out of the commodity server market, driven by competition from companies like Huawei.
This is a natural evolution of the tech market that we see played out time and again so in one sense at least, no-one should be surprised. When I think about the fact that Google and Facebook have both, in their own way, openly contributed specifications to this market then it can only be a matter of velocity before we can declare 'game over.'
Think about the newer companies coming into the market offering hardware and software solutions. Nutanix for example is taking on the storage incumbents like EMC with commodity hardware they source based upon Google specs, enhanced by hypervisors to produce solutions that are a fraction of the cost of assembling from brand providers.
Previously, the high-end server manufacturers created the perception that low-end was Intel, and high-end was RISC. Intel perpetuated this by producing low-end hardware.
This is about to shift because HP and Oracle in particular are creating very high-end Intel hardware. IBM will be a lone wolf selling only mainframe systems made in-house. Meanwhile, Oracle hedges its bets by also having a mainframe platform.
The death of the mainframe
The problem is this – the UNIX mainframe market is in terminal decline. IBM is doing well because it is eating up market share from the other vendors. But it will start to hurt in the coming years because once you achieve market dominance, there's nowhere to go other than down. In a market where everything is else is commoditizing, you get hit by the double whammy of price erosion.
And with technologies like the SAP HANA database running on HP NonStop, there will be no reason to purchase proprietary UNIX mainframe systems for net new business applications, even for mission critical scenarios.
To add to this, the elephant in the room is the cloud, and none of the cloud vendors that I know of are building on UNIX mainframes any more. Salesforce is an interesting outlier, because it will/is using the Oracle Exadata platform.
Viewed from those perspectives, my prediction is we will see a dramatic change in the hardware vendor landscape. It will not be kind to the incumbents.
Do you agree or think I am overthinking this element of the IT landscape? Discuss in comments.