Steve Mills is the head of IBM’s Software business and also in in charge of STG. He is SVP and Group Executive, Software and Systems; and is a legend in IBM and software circles. As we say in Beantown – he’s wicked technical and is a “super-geek” which is a compliment in this industry.
Twenty years ago, Mills realized that fighting Microsoft head on was a losing battle. He initiated an aggressive strategy to transform IBM’s software business by embracing open source (in part as a means of commoditizing Microsoft’s value proposition) and investing in high value software (e.g. systems management, collaboration, analytics, etc.). The result has been that IBM is one of the world’s most important software companies and software integration has enabled IBM services and hardware groups to differentiate in the marketplace.
Over the past several years, IBM has been outspoken about not participating directly in commodity consumer markets and rather driving value through services and integration. Mills has been a key driver of that strategy and because of these accomplishments, prior to the appointment of Ginni Rommety, was often mentioned as a possible successor to IBM CEO Sam Palmisano.
In short – Mills is the man. When he talks analysts pay close attention.
Bottom line on this session – Not the #1 Mills’ presentation I’ve ever seen, mainly because his message was not that new to me. It was however credible. Mills emphasized that IBM’s strategy is to create value through integration and drive improvements in IT economics. He cited the mainframe as a strong example along with some other proof points inside IBM. His other strong point, which was useful, is that despite the fact that information technology is all about automating human tasks, labor (both internal and outsourced) remains, by far, the biggest area of expenditure around IT. This is IBM’s main focus – i.e. to attack labor costs.
Mills’ Q&A is always fun. Last year he said “SPARC is dead.” This year he took a few more shots at Oracle – essentially claiming that what Oracle is doing with Exadata and Exalogic is not that earth-shattering– in fact Oracle’s performance is largely enabled by SSD. The rest of the innovations that Oracle touts such as compression and indexing are things that IBM has been doing for years. Nonetheless, Mills’ comments underscore what CEO Sam Palmisano admitted last year – i.e. that Oracle is IBM’s biggest competitor. IBM keeps a close eye on Oracle especially because as Palmisano indicated, Oracle spends on core R&D.
32M systems installed and 1.2 Zetabytes of stored data. Clients talk about how 70-80% of their IT budgets are pre-committed – i.e. focused on keeping the lights on.
Mills showed a slide on the Producer Price Index from 1996 to 2010 and showed a 23X improvement in 15 years for systems and a similar trend for storage.
Mills showed a great IDC slide (I think developed by Michelle Bailey) that shows TCO shifting from CAPEX to server management, admin and power/cooling. Cost of acquisition is a minor element of TCO – data from 1996 – 2012 shows that CAPEX has declined from about 66% to around 33%. Total spending on servers is approaching $250B (all costs) with installed physical servers approaching 46M by 2012. Interestingly, it appears that the proportion of spend that goes to power and cooling has moderated relative to some earlier forecasts that I saw from IDC, Google, the EPA and others. All data presented is worldwide.
Mills spent a lot of time touting the benefits of the mainframe and how the mainframe admin teams in client bases have not grown but rather have remained flat. He cited the hybridization benefits of the mainframe as a key reason. My POV: He didn’t mention the fact that most CIOs aren’t allocating increasing budgets toward the mainframe. But the point remains – the mainframe is alive and well. As I’ve said many times, everyone wants a mainframe – that’s what VMware is building in software – they just don’t want to shell out the initial CAPEX costs.
“Mainframe is the biggest scale out system on the planet today – bar none…It’s all about consistency and sharing and that’s what brings down cost.” Debates about architecture represent a religious debate that is not of material value to businesses any more.
Pre-integrated capabilities offer the promise of potential time-to-value. Appliances are a good example (e.g. Riverbed, F5, others). It’s all in the box which provides convenience of acquisition but may not be optimized to lower cost. Other examples actually do deliver high degrees of integration and optimization – Mills used Netezza as an example that’s easy to understand – pre-package versus roll-your-own.
All of this discussion about integration is about economics. “We’re not going back to the 1960′s” (vis a vis pricing power). Mills is putting forth a “reasonable” scenario of integration that balances simplicity, time-to-value, lower cost with not too much lock-in.
The IT industry spends more on labor than any other component. So the key issue remains how to bring labor costs down – buying, setting up, deploying, managing, maintaining, etc. The value on the asset side is increasing at a much faster rate than the value on the labor side – so IBM will focus on improving the labor equation. It’s not just about what the hardware is, but what can be done with the hardware and how it can deliver more value.
Gordon Moore’s observation about the doubling of transistor density ended some time ago and has led to multiple cores. Launching more than one process on multiple cores is the software trend. Scheduling, managing and controlling those threads becomes increasingly important. IBM spends a lot of effort optimizing its system leveraging many thousands of threads and optimizing instruction sets. Over time, the challenge faced by the vendor community is how to exploit the total of the real estate from 8->12->16 core systems. Threads will double and double again and keep doubling. So the challenge is how to “serialize” the commerical code. There are ways to achieve high levels of throughput optimizing at the middleware levels and deliver commercial workloads that were previously unattainable. It’s heavy lifting, a lot of work and IBM is good at it. Z and P are IBM’s main areas of value add but IBM will do this on Intel. “No one will ever do anything that will match MVS – ever – ever – ever. No way; ain’t gonna happen.”
Steve ranted about Oracle’s Exalogic – compression and indexing and other so-called innovations that Oracle is touting. To this Mills said – “look – Oracle gets its performance from SSD.” That other stuff – things that IBM’s been doing forever – “duuuuhhhh.” The question/issue is not that this stuff can be done but more how can it be integrated. It takes days and weeks to standup Oracle Exadata and Exalogic systems – great for Oracle – but if you’re a customer and you’re on the flip side and your paying through the nose for maintenance and power and cooling it could become problematic.
“Private clouds are dominated by development clouds today”
What about ARM processors in the enterprise? Adding management capabilities to ARM makes the processors more sophisticated but also more complicated. Wouldn’t count ARM out – we’ve seen low end come into the enterprise and eat away at the high end but questions remain as to how large that market will be and how it will play out.