Hacker News new | past | comments | ask | show | jobs | submit login

A lot of ink spilled to reach these conclusions:

> one of the most important takeaways from the Red Hat acquisition is the admission that IBM’s public cloud efforts are effectively dead

> while IBM will certainly be happy to have the company’s cash-generating RHEL subscription business, the real prize is Openshift, a software suite for building and managing Kubernetes containers

> IBM is betting it can again provide the solution [to cloud lock-in], combining with Red Hat to build products that will seamlessly bridge private data centers and all of the public clouds

To sum up - IBM lost the hardware cloud wars and is making big bets on cloud software. The author doesn't answer the key question posed at the end - does the problem of cloud vendor lock-in actually exist (and the implied question - if it does, how profitable is it)? I suppose that's the info he wants you to pay for ;)

Where I work the answer is most certainly a YES. We have so much stuff using non-portable cloud offerings that switching would be a nightmare. (Step Functions, Lambda, SQS) Sure we could switch to replacements but the cost and time would be prohibitive.

At some point most large companies will play their hand and raise prices to more closely match the cost of switching off the platform. (i.e. the MO of Oracle/IBM) It is really more a matter of when, not if. The problem with doing this is customers wise up this strategy and quit using your products for most new work. Once this happens, the company's best option for growth is to buy new users via acquisition.

I don't see Google allowing this to happen, they will try to offer lower prices/extras because they need to make up for lost time and market share to leaders AWS and Microsoft who are quickly pulling away as the 2 leaders in the market.

Google has enough resources to accommodate any medium/large enterprises that want to switch through price and onboarding incentives. Not to mention they employ many of the world's Kubernetes developers, experts, and maintainers.

The holy trinity has enough cash and willpower to keep a monopoly/monopolistic pricing from happening and there's enough business for them outside of the cloud to encourage them not to collude IMO.

Not to say hybrid cloud architectures are unnecessary - they most certainly are good, and the way the industry is moving, I just don't think there will be a huge pain point over monopolistic pricing/actions.

It doesn't really matter that Google does.

Sure it might be cheaper to use a competitor but if it will cost us 10M to migrate and test the working production workflow then it would have to be 10M cheaper to switch.

This gives our current provider a strong incentive to charge us a bit extra. Say 1/4 the price of the migration. Nobody wants to start a high risk migration that would take 4 years to pay for itself. Easier to just keep paying 2.5M extra. Safer that way and it isn't their money anyway.

But cloud software will continue to come (and go) rapidly, and Google would snap up the new customers in the frothy market.

AWS would have to be certain that these losses weren't more than their up-priced gains

> At some point most large companies will play their hand and raise prices to more closely match the cost of switching off the platform. (i.e. the MO of Oracle/IBM)

I don't see AWS doing this. They've continuously implemented technology improvements that have allowed them to reduce costs for customers 67 times in the last 10 years. The price of cloud services always goes down, not up, over time, as you're able to realize the benefits of improvements in compute power for the same cost (Moore's Law) without upgrading hardware.

Disclaimer: I work for AWS, but my opinions are my own.

You are thinking like a software engineer, not a product manager or owner of capital.

If any of these companies is able to extract monopoly profits without fear of competition undercutting them or government/legal intervention, they will.

Right now all 3 are competing fiercely to win the hardware rental market, but this may not be the case in the future once the market matures, and they may raise prices to match leaving costs (as parent comment pointed out) if they can lock customers in.

I have to +1 this.

The price must always go up over time. If there were a first rule of business, that would be that.

Price decrease is just a longer term strategy where it goes down before it goes up.

The amount of implicit absurdity behind your words is astonishing. Basically outside die-hard Marxists nobody takes such claims seriously, because it implies lack of competition.

Hint: the cloud market has PLENTY of competition.

> IBM lost the hardware cloud wars

This makes me scratch my head. I heard Power9 is getting more and more traction in some clouds, notably Google's.

Maybe IBM lost the "bog-standard VM-based cloud" wars, where the "standard" is what AWS / GCP / Linode / DO, etc offer. But that space is indeed crowded; doing something where you have little competition makes a lot of sense. (I wonder if they'll be able to.)

IBM bought SoftLayer 5 years ago, and was unable to capitalize on their niche (bare metal). Instead they decided to double down on provisioning VMs, which SoftLayer was never that great at. The set of features IBM's VM's provides compared to any other modern cloud is terrible (headed backwards in Gartner charts). It's barely getting any better, meanwhile Google/AWS/Azure are gobbing on features.

IBM effectively failed in their integration of SoftLayer. They bought SoftLayer but failed to modernize them. Why will RedHat be any different?

Take a look at what's left of SoftLayer today. Most folks who had the necessary skills to keep their ancient php platform relevant have left a long time ago.

I'm sure IBM is on their 5th iteration of their "next gen" VM provisioning system by now, due out any quarter...

Have they fixed their scheduler? For 5 years nothing has improved in the process of launching VMs. Sometimes it just barfs and goes to hell, requiring contacting support to terminate provisioning.

No joke, at one of the previous places we’ve implemented a task which would raise a support ticket via API if provisioning step was hanging for longer than x minutes. This is the case for years!

According to some people from IBM, the reason why IBM purchased SoftLayer was so they could tell their customers who were looking to move to the cloud „you want cloud, we have cloud” ... and keep the hefty support contract runing.

More fatally than that, perhaps: IBM's business model for half-a-century at least including renting time on hardware at as close to commodity costs as IBM could push them (and still make a very healthy profit admittedly). The Cloud model was IBM's invention in a time where computers were expensive. IBM's failure to transition that very business model the company was built upon to a time where computers are plentiful, is like watching an Olympic swimmer forget how to swim in a lake.

SoftLayer is not competitive. It was undercut by Google Cloud since last year, while AWS is not that far off.

Once everyone offers a choice of 1 TB Virtual Machines for rent, SOftLayer will be completely out of the equation. They're strictly worse in services and pricing.


I think we'll see a trend of cloud fragmentation again, once there are more good cloud software stacks. Imagine something like OpenShift + OpenStack + Ceph that does not suck and can be controlled by something like Terraform and support various types of hardware. That's just a crude example I can come up with quickly to demonstrate the concept. Something like that will enable many more types of players in the market and people can use a mix of them or self host.

It will be good for IBM to be in the middle of this, since the first such stacks will probably come from a large player that doesn't have a dominant existing cloud service. It will also be good for IBM's hardware expertise.

I think this trend will come, companies are getting locked into AWS almost deeper than some companies got themselves nailed to IE6 + .Net at some point. This is fine when the benefits appears to outweigh the risks, until it doesn't.

I've never heard of power 9 getting any traction with Google, short of a press release 4 years ago by them saying IF power9 lived up to the hype, they would consider switching. Since then, I've seen no information that they actually moved any noticeable portion. It's certainly not available publicly. Even IBM didn't have power9 in their own public cloud last I checked.

IBM Helped design a lot of the stuff google uses. https://www.wired.co.uk/article/openpower In addition you have this: https://www.forbes.com/sites/patrickmoorhead/2018/03/19/head...

All it said is they've deployed power. That could be 1 server or 1000. I think the fact still stands that of IBM doesn't have them in their cloud, who will?

Vendor lock-in is most definitely a huge thing. And being on the side that wants to avoid it and build portable solutions is never a winning argument. You're always the guy who's setting the perfect up as the enemy of the "works now."

I seriously doubt that IBM can execute on this plan based on my limited experience with them. But if they could, it would be a winning move.

i've read his stuff for years. He puts his best articles in the weekly update, which is free.

It generates the most views and clicks. His other 3 articles per week is...not bad but certainly not as good and meaty as weekly updates.

Paying for it is on some livel similar to patreon sub.

Good insight, thanks, I only see his public stuff which is generally pretty good, but I assumed he would put more quantitative/exclusive analysis in his subscriber stuff

I mean...me too! But IMO that's not quite the case. I think the argument is the best, most insightful pieces are going to get the most traffic.

But still, I'm still pretty happy about the subscription. He's better than most bloggers out there.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact