Hacker News Reader

Comment by joshuaellinger | Hacker News Reader

My wife is an ex-IBM and the cut-to-get-growth story rings true.

But it overlooks a core problem that IBM faces. The cloud is a competitive platform for problems that previously required mainframes. They successfully defended (kinda) against Oracle/Sun in areas like running large banks and stock exchanges.

What happens to IBM when you can solve hard real-time transaction problems on 1000 loosely-coupled distributed computers instead of 1 large mainframe (made up of a 1000 processors with shared state)? That day is either here or close.

My take is that they can't ignore the cloud but they can't win it either. Tough place to be.

View on HN

IBM can milk the mainframe for years to come, because the cloud doesn't compete against its legacy enterprise customers. Large enterprises have chosen to retain many of their legacy mainframe applications. Previous attempts at re-writes have too frequently failed due to cost overruns or have been complete failures. So, the safe choice for a CIO is to avoid rewriting legacy systems (many have been around since the 60s or 70s). So, IBM customers will continue paying their premium to run on a mainframe, and in some cases will leverage their existing skills/systems to deploy more onto their IBM systems.

This mainframe strategy for IBM will keep many of their legacy customers, but as Cringley and others point out, it fails to capture new growth.

This is true, but they've been milking it for 40 years, and it's in permanent decline. Mainframe sales are not worth anything like what they were in the 1980s.

Further, a big chunk of IBM's server business is running its own Unix. That market is also in decline, and easier to switch away from. See http://www.itjungle.com/tfh/tfh030314-story06.html

I remember back in the Sam Palmisano days (mid to late 2000s) watching him present in an internal all hands meeting (I'm a past Software Group IBM employee). He spoke briefly about mainframe margins, he didn't state numbers, but boy he made it clear that the margins were significant and they were seeing growth.

So, I don't know about the last few years, but I think IBM can ride the mainframe money train for years to come. My point was simply that it's not going to drive any growth.

Agreed, but the numbers have been going down for decades, and are still going down. In 1985, mainframes brought in $14bn of IBM's $50bn turnover. Today it's less than $5bn of IBM's $100bn turnover. Factor in inflation and the decline is even more dramatic.

IBM used to be able to charge $1 million just to install its top-level mainframe OS (before rental charges) and C cost $1,500 a day. I expect those days have gone ;-)

Here's another view point: I recruited for Mainframe Developers a year ago. It's an increasingly rare skillset to find because developers:

- move on to cloud technology or other newer technologies to stay relavent

- don't want to move back to mainframe since they see less jobs for them

- there are almost no junior level candidates for mainframe roles. Most candidates will have 10-20+ years of experience. It's extremely rare to find anyone less than 5.

It is true that legacy systems aren't going anywhere anytime soon. But as time goes on there's a high cost in finding people that can maintain those systems.

In some ways, I think Cringely underestimates IBM's potential. (Or maybe he correctly estimates IBM management's inability to see IBM's potential.)

One way Google kicked the ass of their competition was to radically their cost of computing through smart use of commodity hardware. It seemed like a small thing, but it gave them a lasting advantage, because nobody could afford to deliver the same search features.

Given IBM's deep pockets and deep experience with hardware, I'd think there must be some strategy there that would let them create cloud computing infrastructure with radically lower cost than the competition. We regular people are stuck with off-the-shelf colocation setups, off-the-shelf hardware, off-the-shelf processors, and off-the-shelf software. IBM can afford to change any of that from scratch. Any of it.

I just can't believe that, less than 10 years into the cloud computing era, we have already happened upon the optimum approach.

IBM no longer has deep pockets: it's spent more than it has earned on buying back its own shares, and now has a ton of debt.

IBM has been the high-cost supplier for almost 100 years, so the idea that it will out-innovate and undercut Amazon, Microsoft and Google in the cloud is an interesting one. Especially as both Microsoft and Google have much deeper pockets and are approaching IBM in annual turnover. (IBM used to have 70% of the IT business. Fairly soon, it won't even be in the top 5.)

What deep experience with hardware? Yes, there was a company called IBM that had a lot of hardware experience, and there is currently a company called IBM. But most of the parts of IBM that have deep experience with hardware got sold off to Lenovo.

Z-series (mainframes) and Power systems (high-end UNIX servers). Yes, both are in a shrinking market, but they've got significantly higher margins on both sales and support than commodity x86. They're also what's keeping IBM's domestic hardware manufacturing going.

Google were innovating new product with the money they didn't spend on hardware. And, AFAIK, the didn't so much reduce existing cost as reduced the future cost of new investment i.e. increased bang for buck.

That's not the same as doing removing every other light bulb in the building (which is what Lucent did at Bell-Labs!)

It isn't just going to come from cloud/distributed systems.

You can scale up to a huge extent just with average x86 servers now and these are competing with what used to be done on a mainframe. So even many companies who don't or can't use distributed systems don't need to use IBM any more.

And it's not just hitting in obvious places.

Consider Hadoop. I know of at two major DB2 customer who are investing in Hadoop rather than expanding their DB2 capabilities to handle problems IBM would love to solve with their solutions.

ibm do distributed computing, their selling point will be that the firms computers will be on location, rather than away in a warehouse in another state.

Latency and concurrency. Mainframes still have an advantage vs. the more big data or multiple task friendly cluster.

Reply on news.ycombinator.com