
Good
Couldn't happen to a 'nicer' group of bastards.
Broadcom has lost another significant customer after UK-based cloud operator Beeks Group decided to adopt the open source OpenNebula stack. Beeks offers virtual private servers and bare metal boxes for financial services providers and emphasizes low latency for both – and for the trading network it operates. The biz told The …
What's funny is that ARs has cross-posted to this story with the lede "Company claims 1,000 percent price hike drove it from VMware to open source rival"... but the wording here "A bill from Broadcom for ten times the sum it previously paid for software licenses" means that the price hike is 900%.
Yeah, I'm being pedantic... but from a technical site???
"Migration wasn't easy. Beeks runs proprietary software that it had tied to VMware APIs, and that had to be rebuilt to interface with OpenNebula.
The open source project's tools to collect metrics Beeks and its clients value – regarding CPU performance, and utilization of CPU, disk, memory, and network resources – were not strong. That's an issue because an overtaxed CPU or disk can be the difference between pleasingly fast trades or missing an opportunity. As OpenNebula is open source, Beeks felt able to develop tools to collect that info – and succeeded to its satisfaction."
[Anakin Meme]They provided their changes back to the project?[/Anakin Meme]
You ignore the small customers, and the big ones find that your billing is outrageous, your quality insufficient, and they are ready to rewrite their own tools to move away from you.
Remind me again why the hell you bought VMware in the first place ?
Because if it was to kill it, you're doing a brilliant job.
Not sure why the down vote. I hold many VMWare certs and up until around a decade ago, I was a big proponent of VMWare, but other stuff came along that was objectively better and I could no longer recommend VMWare with a straight face...there was just no compelling reason to recommend them anymore. Was it the fastest product? No. Was it the easiest? No. Was it the most stable? No. Did it offer the most features for the best price? No.
These days, Proxmox is the king. It sits in that sweet spot that you want wit ha hypervisor...it's low cost, pretty straight forward, has no limitations out of the box and performs extremely well especially with LXC containers.
The absolute bullshit aspect of VMWare for me was copying VMs to and from an ESX box. That dumb transfer speed cap made using VMWare a fucking chore (no matter the speed of your network or how many connections you bundled together or how fast your storage was, you were capped at around 30MB/s. It was so arbitrary as well...designed to sell you licenses for third party backup products or additional VMWare solutions that should have been entirely unnecessary.
The cap seems to be gone (or at least higher) in newer versions, but it was too little too late.
There was also the crappy snapshot feature that a lot of third party backup solutions relied upon. Every now and then you'd get a snapshot that was stuck and it would be an absolute pain in the arse to resolve, because you wouldn't know it was stuck unless you made a point of checking every VM every day after every backup / replication...especially on multi-terabyte VMs with several virtual drives.
It was just straight up shit by the time viable alternatives rolled around. It was too proprietary for it's own good because the door was wide open for competing products to swoop in and plug arbitrary gaps for free.
Even things like USB pass through were trash. USB pass through on Proxmox works first time, every time...there hasn't been a device that hasn't worked...VMWare on the other hand, randomly loses mappings, occasionally requires a reboot to resolve USB issues etc etc...it's a headache that you don't need.
I mean, if you need a static as fuck, bog standard setup...i.e. you're virtualising a Windows server that will act as a print server or something or you just need to host a bunch of VMs that don't really need to be backed up, then VMWare is fine I guess...but if you need more than that, it's straight up shit.
Company I work for increased prices on one of our SaaS products by 40% earlier this year _and_ removed bulk discounts. A few of us advised against it (or at least wanted existing companies to be grandfathered in on the old price) but were overruled. Just checked and revenue is up by 40%...!
We lost some of the biggest ones, but the ones that stayed have made up for it
The key statement there is "earlier this year"
You can be absolutely certain every single one of the customers is considering leaving at renewal, and many of them will.
The bump in revenue from such large increases is almost always short term, followed several sudden declines as contracts end. You have to boil them slowly enough that the purchasing managers move on and don't notice the increases stacking.
This. The devil is in the contract renewal not the price hikes....at that point though, the bump in revenue might be enough for the provider to kid themselves into thinking they don't need the customers that are leaving because the cost of running the service will drop with fewer customers, but enough will remain that the revenue still seems higher...that will then lead to layoffs and eventually you have the same revenue you always had, but a much more "efficient" business underneath, which looks great on a pie chart...then your product gets old and crusty and you realise the folks you laid off were quite important to the ongoing development of the product and you can't get them back, the market has moved on and you can't get new business...now you're fucked and you sell to Broadcom / Oracle / IBM etc...they strip mine your project for assets and move the remaining customers to their competing product or a similar offering....the original staff are sent packing and your product ends up in a "legacy" folder on an FTP server somewhere that is really hard to find...FOREVER!
10 years later...
A tech nostalgia youtuber digs up the final build of your dead product and reminisces about how crap your product actually was and breaks down in fine detail why the product failed and who the key decision makers were...who are still active in the industry, but could probably do without the random 10 year later scrutiny and roasting.
Yes, but isn't it the case that these are multi-year contracts?
If that is the case, then you've just received a few windy days, before the hurricane really rolls in.
No amount of storm chasers will save you when the bulk come in for renewal later.
Some will renew, simply because they haven't yet got a plan to migrate in the short term.
They will have one long term for sure, unless your product has no competitors.
of those who have managed to wean themselves off the VMware tit.
Once they start sharing their experiences of the problems encountered, the solutions that worked and those that didn't, and the alternatives and options considered and those chosen we will have arrived at the 'best practice' for giving Broadcom/VMware the bird. ;)
Look at the other numbers. 20,000 VMs on 3,000 hosts, 20 DCs. That needs a lot of vCentres and all the other bollocks.
Mostly all the other bollocks and there is a lot of it when you do E+ with knobs on. Add in Tanzu if containers are your thing and you need even more. SDN and a funky firewall? lol, more controllers and more stuff.
"Beeks's offerings are a little unusual – and perhaps therefore an edge case for VCF."
Debatable. If an IaaS provider is running off vmware, others may do the same. And for sure, doing the maths, many won't find this magic ROI vs. the huge surge of VMware costs.
"Beeks Group's experience of substantial savings is therefore another example for those considering what to do when their current VMware deals expire – and they face a decision about whether or not to buy into Broadcom's virtual vision."
Yes, all of this. Broadcom is betting on the huge effort to re-platform vs. lower yearly costs. This is dangerous, it assumes most companies don't have what it takes and IT pros are uncapable. They/we won't like this reasoning !
My approach has always been to wrap proprietary API's in an abstraction layer to keep them at arm's length. Its rare for a vendor to offer something that is so truly unique that there is no possibility of there ever being an alternative for it. When the time comes its much more straightforward to swap over to another implementation without having to track down and rewrite great swathes of your code base.
You'd think this would be the standard and obvious approach everyone would follow, but sponsors rarely appreciate the value of having an insurance policy until after the vendor screws them over, and developers just want to mainline directly into the vendor API so that they can get [insert technology here] listed on their CV's.
It's a good idea in theory to introduce an abstraction layer. After all, it makes it potentially much more simple to switch providers, but how much latency is introdoced? It may only be a millisecond or two, but if your application makes a lot of API calls, that could introduce significant latency. If the system is in any way time sensitive, even a millisecond or two could be a significant problem.
Ignoring whether Beeks will impact VMWare or not, ultimately beyond the pale increases in costs, for no apparent increase in services or innovation will always cause users to consider is it worth it.
Enough variations on a theme discussed in this thread for sure, but whoever is advising companies to apply these levels of increases just to improve the bottom line, is still missing the key point: piss your customers off and they will look for alternatives.
What's the logic behind 1000% price hikes? Do they really believe people will put up with this? Are they focusing on a few mega-clients and willingly ejecting the rest of their customers through abominable price hikes?
I mean, at some point their "mind share" (if there's such a thing) will evaporate and after losing one or two of those mega-customers the company will be on its knees and probably on its way to bankruptcy.
Has anyone seen the actual VMWare quotes for pre-Broadcom & post-Broadcom?
Did Broadcom actually raise prices 1000%? Did the line item read $1, and now it reads $10?
Or... Does the line item read $10 on both quotes and the "90% discount" line is missing on the 2nd quote? Software sales is the only business I have seen this type of markup + massive discount. No other industry actually sells their products for ten cents on the dollar in the final deal.
I am not a fan of how Broadcom does business. IMHO one of two things will happen: 1) Hardware-centric Broadcom will eventually 'learn' that how software is sold and capitulate (CA, Symantec, VMware....we're waiting). 2) Broadcom will change the industry and everyone else will do away with these massive discounts.
Regarding pricing, as a cloud provider, originally VMware was priced by the amount of RAM allocated to VMs, capped at 24GB per VM. I think from memory VMware charged about 68p/GB/month. (Consumed, not total so you could have 1TB in a server and still only pay for 400GB if that's all you used). This was policed by a usage meter which is a VMware appliance VM which runs in the environment, connected to vCentre so it knows everything about the environment, hour by hour, day by day.
So for VMware licensing, the number of cores in your environment was irrelevant, so we all had lots of cores because they were cheap. This is good for redundancy, extra hosts cost very little, good for coping with spikes in traffic and workloads and keeping customers happy with snappy performance as well as onboarding new customers quickly. Then Broadcom came along and said we had to pay US$350/core/year. So a small platform could easily go from paying $5k/month to paying $50k/month depending on how many cores they had in their hosts and how much RAM was being consumed. And previously with memory being the cost item, we all had spare servers, lots of redundancy.
This is where the massive increases came from. Changing the licensing model and charging more at the same time. Discounts were not announced for some time.
Then the discounts became known - they are not generous, it takes a single core from US$350/year to maybe US$280/year. So you can see that a 128 core server now costs $36k a year when the RAM based cost was in the region of $8k+/year. Not 10x but a significant multiple to make CFOs thing they're being royally reamed out !
One aspect of the change Broadcom has forced is that the CPU utilisation on a host will be pushed up - typically you don't run at 100% utilisation because there's no room for peak and the scheduler gets very sluggish when it has to schedule CPU time for VMs when the CPUs are very busy. But now expect higher CPU utilisation which will result in lower performance.
Also redundancy will suffer. It was previously possible to have a lot more hosts than were really needed so maintenance didn't affect redundancy and multiple hosts could be taken down for maintenance, but I suspect now many companies will reduce the host count so maintenance might impinge on redundancy. Also if you have a second site running DR you need to licence all of those cores too, if the hosts are powered up. Previously VMware didn't charge for hosts not running customer loads, as the RAM wasn't consumed.
Just thought this might be interesting to some people who haven't had to budget for VMware previously so wouldn't be aware of the issue.
There are two main issues,
- Move from perpetual licensing to a subscription model.
- Bundling of features. They have over 25 offerings so you could pick and choose the functions you wanted. That got stripped to I think 3 bundles though they have backtracked and created a couple of other bundles.