Hacker News Reader

Comment by duxup | Hacker News Reader

IBM really seems stuck in their old systems / and at some level doesn't understand what to do with new systems.

I was talking to someone who was using the "IBM cloud". They went to spin up some servers and etc and IBM was all "that will take a few weeks, we have to install them". At least to me "cloud" includes the ability to spin stuff up on demand.... IBM was operating it like you were ordering from a very old co-location service or something.

Word has it Watson failed possibly because it was sold as a magic drop in solution where in reality careful data entry and a lot of work was required to get quality results requiring both the customer and IBM to be heavily involved. Maybe had they understood / sold it right and brought the customer's along and tried it in the right places they would have done better.... or at least had a better chance.

There seems to be a fundamental misunderstanding of how things work and thus how to sell / execute at IBM.

View on HN

I hate to defend IBM (I used to work there and have nothing good to say) but its likely they were using the "Private Cloud" which is exactly what you say, co-location in the IBM Datacenter. Its just marketing rubbish.

"IBM Cloud Private" - it's a fascinating product that can be summed up as "We package kubernetes for you, and you pay us 2 million dollars. If you are a huge customer, we'll discount that to 1.6 million dollars."

Trying to keep from maniacal laughing during their sales pitches is always a fun game.

It has been noted that an on-call staff with five years of k8s experience will also cost around $1.6M though.

>they went to spin up some servers and etc and IBM was all "that will take a few weeks, we have to install them". At least to me "cloud" includes the ability to spin stuff up on demand.... IBM was operating it like you were ordering from a very old co-location service or something.

Well, the spin up is instant, but the multiple forms and stakeholder sign offs aren't :-)

One of my close family friends (an endocrinologist) participated in a Watson study at Yale New Haven Hospital. Apparently the amount of effort required to input the patient medical history from paper/other EMR into Watson's system was astronomical compared to just doing the analysis themselves. It sounded like it was mainly a UX issue but that's indicative of IBM's understanding of their market.

I got the impression from other articles that there was a bit of a constant process inputting data, formatting, reformatting and etc to get Watson to digest it for any given problem. Not necessarily a game breaking flaw, but it certainly meant that you would not use Watson for any .... issue that you already could solve with what I'll call "medium" effort. It had to be a really big issue, really big payoff before all that time made sense.

Yes. They kind of claimed Watson would tell you the answer even if you didn’t even know the question.

“Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?"

How many servers were they spinning up? My understanding (which could easily be wrong) is most of 'IBM Cloud' is really SoftLayer rebranded. Softlayer has virtual machines they can spin up for you within minutes, and they can spin up a small number of servers pretty quickly if you don't spec the hardware too weird, but if you want like 50-100+ servers with different specs than they usually do, they often have to order or cross ship parts to the datacenters you want them in. Waiting a few weeks seems out of the norm though.

(I work for a SoftLayer customer)

I don't recall the specifics. Their organization was what I'd call a mid-sized organization so it wasn't small. They were under the impression they could get it on demand from other providers, but IBM did not do that.... but they already chose IBM so there they were.

It is entirely possible they asked for something(s) weird.

Reply on news.ycombinator.com