It looks a lot like VMware just lost a 24,000-VM customer Global stock-market share registry operator Computershare looks like it has just decided to bail from VMware rather than suffer Broadcom’s latest licensing regime and price hikes. Speaking during the closing keynote of Nutanix’s Next conference in Barcelona on Wednesday, Computershare's CTO Kevin O’Connor was asked how he …
Don't take this as condoning Broadcoms practices, but I suspect that their pricing is such that they make 3-4x the revenue off VMware for the next 2-3 years as the reluctant migrators wait for Broadcoms to see sense and then finally move over the next 1-2 years.
At which point Broadcoms have made a massive profit and can sell VMware onto someone who wants to rebuild it.
That would be an extreme example of short termism. Why buy a company for (guess) $billions, gouge its market for a few extra $billions in a couple of years, and then have nothing but a dry husk of a long forgotten product? Why not keep the company running for the long run and make far more?
And if the strategy is to harshly squeeze a few billion out of a dying product, why not simply get into the technology to where the product's former customers are moving, and secure their future business and make more money that way?
All ways round, it feels like they've deliberately set out to make less money than they otherwise could. That's not a great business strategy, it's called a "lost opportunity cost", and shareholders do notice those regardless of whether they're on the balance books or not...
Not that I've ever considered Broadcom to be rational...
Long term profits don't help the executive when their bonuses are based on short term revenue (not profit!), and they intend to be long gone when the ceiling falls in.
Large acquisitions almost never turn an overall profit, and the larger they are the bigger the eventual loss.
But as the pay and bonuses of everyone involved in the decisions are based on the scale of the acquisition, and none of it returned when the inevitable write-down happens a few years later.
When a company becomes more focused on extracting cash from its customers instead of meeting their needs, it's time to go. It will not get better; it will only get worse. It will only be temporary if it does get better until they figure out a way to lock you in. Once you are locked in, it will get far worse.
Back in the day when cloud was on the horizon rather than above your head, a typical organization had about 1 (Windows) server for every 6 or 7 personal computers.
Or about 1400 for this organization, if everyone has a PC. I wonder what the other 22,600 virtual machines are for?
Just asking.
I expected "Computershare" would rent out ... shared computers, but wikipedia says "Computershare primarily provides stock registration and transfer services to companies listed on stock markets, but also offers technology services for stock exchanges, investor services for shareholders and employee share plan management."
So it's a financial services company? Then the number of VMs is indeed quite high.
I don't know about the specifics here, but fancy VMs are used to pack a large number of virtual devices into a smaller number of powerful physical devices. Memory dedup and increasing CPU utilization are the big win. Some fancy VMs can hop from host-to-host to balance loads.
The less cool use is running ancient software needing ancient operating systems that won't boot on current hardware.
That still leaves the question of WHAT they're doing with 24,000 servers, even if they've been made virtual, when the calculation is that a company that size ought to get along with 1,400. (Whether that's a truly valid number, I don't know. It would depend on the company.) One contributor to that might be simply separating out workloads onto different "servers" for ease of management where a company in the past would have needed to lump them all under one OS/host simply because that was the only way to take full advantage of the hardware. If you can afford one physical server and can't run virtual servers on it, all of your workloads end up running on that one host even if it's harder to maintain, like having to take down the entire infrastructure just for one update. You'd only put in a separate physical server if it was absolutely critical that the workload on it be separated, because you'd be wasting a lot of hardware cycles and electricity. Once you can run virtual servers, you can run one workload on each VM and they won't affect each other if one has to go down, and you can even run two VMs performing the same task so that service is never down unless the actual hardware goes down, and you take advantage of as much of the available system performance as possible which is more efficient. So 1400 physical servers could potentially be broken down to 5000+ virtual machines, and even more if they're spread around the country. They may also run a lot more development machines now.
But VMware licensing for the hosts for servers on this scale isn't based on number of VMs usually, is it? So that would mean the majority of that 24,000 is actual physical hosts, which could be running hundreds of thousands of VMs, which would be even harder to understand.
"But VMware licensing for the hosts for servers on this scale isn't based on number of VMs usually, is it?"
If they were using VSPP licensing it was. No hardware licensing costs, only VM RAM reservation assignment, up to 24GB limit, charged at an hourly rate. We used this, we would have had a bill 10x what we had with the new licensing agreement if we hadn't cut all our hardware to the bare minimum to run what we have until we decide on what to move to.
Since their business appears to mostly involve moving financial instruments around, and they run that in a lot of different countries, maybe they're used to do that? I'm not sure where your number comes from, but a company whose primary product is a service that a lot of servers are needed to implement is going to have a lot more servers than a company whose product doesn't involve servers at all and the servers are just there so that people can do their jobs correctly. There will not be a single PC-to-server ratio that applies to anything, and I already don't know where yours came from.
Are we talking about the same company? I don't know the company, but when I looked up the name I got this Wikipedia article. It only talks about financial market services, not rentable cloud boxes. Either we have two different companies, and the one I found seems to fit what the article says, or someone should update the services offered section of that page.
14,000 / 7 is 2000, not 1400, and they are hyper-converged, so the switches and other things are VMs as well as the file servers, so let's say 4000. The other 20,000?
All I can suggest is that they are a cloud services company, so maybe running micro-services takes a lot of micro-servers?
Could be that their employees are all on VDI platforms, which could mean 1:1 on employees. That alone is a sizeable chunk. Also it appears they do share registry stuff, so they may dedicate a VM for a single corporate share registry to keep things clean and separate so that each one has a completely isolated DB. Numbers can add up quite easily.
This post has been deleted by its author
> Or about 1400 for this organization, if everyone has a PC. I wonder what the other 22,600 virtual machines are for?
maybe for:
serving 40,000 clients and 75 million end-customers.75 million end-customers are going to need some computing resources to do their interfacing with the company (logging into their accounts to see their share holdings and so on). And providing appropriate resiliency/High Availability to those same customers.
Also, often VMs are 'smaller' servers than a traditional server you are talking about. Microservices. So instead of 1 server providing a webserver, local disk, some sort of actual transactional processing (e.g. a Java application server doing work) and databases and whatnot, with VMs you are going to split all those services out into their own VMs - webserver, application server, file store, database, etc. And each of those will be smaller, so instead of having a single web server handling 10000 requests/second (say needing 10 cores and 64GB RAM), you'll split that into 10 VMs each processing 1000 requests/second and having 1 core and 6GB RAM. That makes each individual 'server' simpler and easier to configure and manage - less impact if it crashes, and easier to spin up new instances for more load and shut down unused instances when load decreases, and live migrate them between physical 'boxes' to better manage load across the VMFarm - e.g. maybe you've got 3 physical servers of 24 cores and 256GB RAM that you are paying hourly rent to a cloud provider for, load goes down, you can live migrate all the VMs to only use 2 of those boxes, and stop paying the cloud provider for that 3rd 24-core/256GB RAM box for 12 hours, then as load ramps up you spin up or move instances back to that third 'box' and start paying 'rent' again to the cloud provider for that box you are only using for 12 hours. It also allows beter tuning of each individual VM/server. A webserver needs different tuning parameters than a database that needs different to a file server that needs different to a application server that needs different to a data store. Having them all on the same box means compromise in the server config as the same server is doing them all. Splitting them up so each VM handles a specific type of service (web, database, applicaotin, etc.) means that each server (VM) can be tuned best for its workload without compromise.
Some devs drank a lot of the micro-services kool-aid and split their apps over lots of VMs. I've heard of some apps having over 100 VMs due to this - and not because of scaling. Nowadays you'd use something lighter weight such as containers, etc.
Multiple by your various environments (Live, Dev, UAT, QA, etc) and you could be getting close to 1,000 VMs just for one app.
Crazy, but plausible.
Nowadays you'd use something lighter weight such as containers, etc.
I'm getting the impression that there are a lot of instances out there which are a single containerised app, running on an entire VM. So, basically, you start with 100 VMs, some dev decides it's too complicated to get the app running on "bare metal", so they dockerise/whatever it, then place that on the VM instead.
I just a finished up supporting a major utilities provider. Their corporate "prod" environment alone had 1200 VMs, as did each of the 2 other, DEV and UAT, environments, so roughly 3600 VMs spread across the three environments. BUT, there was also 4 other environments, 2 OT environments, and 2 communications environments (for service provider connection), then each of those had 3 separate environments, DEV, UAT and PROD, again each with roughly 1200 VMs each.
So with 3600 corporate VMs, 7200 OT VMs, and 7200 comms VMs, you're easily pushing 18k VMs. And the utility only services between 6 and 10 million customers. And to top it all off, I didn't include the VDI's for each of the 5 primary environments. So, again, 24,000 VMs is not really a whole lot.
I think the article is not particularly clear - it states at the top "Computershare will migrate 24,000 VMs to Nutanix" while it later states "All of which requires 24,000 VMs, which isn't the largest fleet around but also certainly isn't trivial."
So is it 24,000 moving from VMWare plus whatever is already on Nutanix, or it is 24,000 in total?
I've submitted a request for clarification to el reg...
Over the past three years, I've been involved in the migration of some 48,000 server instances, Windows, Linux, from VMWare (hosted at both leased/cloud facilities and hosted locally/on-prem) to cloud native instances NOT using any VMWare product.
And my experience is hardly unique. Colleagues at other firms have done more work in this are than have I.
If I had any VMware stock, I would have sold it the second they announced the licensing changes. I should have shorted it when I saw how they completely obliterated the forums.
Go ahead. Try searching for some VMware related KB article. Lots of top Google hits, only NONE OF THEM LINK THE ACTUAL ARTICLE! Every single one links to the main landing page of the Broadcom forums. No big deal, you say, just search for the same subject? Not a chance. Go ahead, try that too. You'll get 5000 matches and NOT ONE SINGLE ARTICLE RELATED TO THE ORIGINAL KB! I mean, come on. How incredibly inept do you have to be to not configure a 302 redirect to their new locations on the new domain? This is sysadmin 101 stuff I've personally done over 20 years ago. They just lost DECADES of premium Google search rankings.
Whomever is over there at VMware making these decisions needs to be terminated and have whatever salary they are making (and millions in bonuses) given back to the company. They are destroying VMware from the inside. The company I work for has well over 5000 licenses and we won't be sticking with VMware much longer either. Once that contract ends, they can go suck rocks.
Broadcom's executive's thought: If you could just go find the answer on the forums, instead of contacting support, you'd feel less like you were getting something for all that money you're paying them. They think if you're forced to use the support you pay for, you'll be content with the high cost.
> If I had any VMware stock,
Since VMWare is wholly owned by Broadcom, it would be physically impossible for you to own any VMWare stock since, by definition, all VMWare stock is owned by Broadcom. That's why Broadcom can unilaterally make changes to VMWare licensing, as literally no private citizen or company in the world has any say over it. The only one's who could have a say would be governments if they decide there are anti-trust or other regulatory breaches by such a decision.
"I should have shorted it when I saw how they completely obliterated the forums."
"How incredibly inept do you have to be to not configure a 302 redirect to their new locations on the new domain? This is sysadmin 101 stuff I've personally done over 20 years ago. They just lost DECADES of premium Google search rankings."
I am absolutely no expert on VMware, but I can't help to wonder if having the forums crumbling... wouldn't make transitioning away from VMware more slow and painful without easy access to help and tips for the community and people describing what were the bottleneck and lock in features in VMware's products and how to work around them. As another poster wrote: if they can price gauge for a few more years Broadcom might recoup their investment and more.
That's because it's for the end-user bit - which is Carbon Black (the AV product) and some other bits, and not the core parts of VmWare.
Pretty sure it will be a case of "Look, we aren't owned by Broadcom any more, so why not give us a try?". Mind you, my personal experience with Carbon Black hasn't been all that great either.
My assumption is that this customer would sign a multi year agreement with Nutanix as part of their engagement. That doesn’t protect them from farther down the road, but this experience will almost certainly spur in VMWare customers a keenness to spend more now to reduce lock-in later. Any that aren’t already some way down the road with that planning are the precise customers Broadcom has planned their RoI on.
I hope for most of our sake that more customers choose to spend on change now and not pay the Danegeld. Broadcom falling flat would be a huge boon to the whole industry in promoting a positive view of technology change and reducing the power of software licensing money pits.
We've moved way more than 24k VMs from vmware..
Why would you stick around, it's not a great product they offer anyway and you can do the same for cheaper and even free, particularly if you already have your own middleware than can present directly to openstack and other systems.
24000 VMs really is not a lot. You need to be a 50,000+ VM shop to really matter. Plenty of companies out there at the 100k and above mark.
That said- companies like Broadcom, IBM, and Oracle and their subsidiaries will gladly destroy small to medium size businesses to milk the large businesses for more money.
And Computershare is big: the Australian company had revenue of $3.3 billion last year, its 14,000-plus staff work across more than 20 countries, serving 40,000 clients and 75 million end-customers. All of which requires 24,000 VMs – a fleet few orgs will match.While I agree that a company that has a $16B market can't isn't eactly a tiddler, in this day and age of multi-trillion dollar ($US at that, not the Aussie Ruble) companies, maybe from Broadcom's point of view anything less than a $100B company isn't worth their time ...
> I love the link at the bottom.
About un-usable site? He's right. Worse even than my community hospital, which is adding mechanical systems, AND now surrounded by town sewer trenches, but can't post complete information in the website. Yes the side-streets are passable, but Main St is a frickin 10 foot ditch surrounded by excavators and trucks. OK, I "can" park on Park St and walk.... but I'm going to hospital for a Bad Knee--- I don't walk well. Tell me the best path!
Here I thought KVM was a way to connect keyboards and pointing devices, not a convenient measure of large quantities of lost business.
What are they up to? Who’s the target customer they hope to keep? Or is it simply a wring-out-profit then shut down move?
It’s not like hypervisor tech has any huge barriers to entry or exit other than inertia.
KVM is a VM facility in the Linux kernel -- the "kernel virtual machine". Each VM appears as a Linux process, just like running any other program. There are user space applications which then use KVM to implement virtual machines for various user bases. The most common is Qemu. That in turn has front ends, from the desktop GNOME Boxes through to industrial systems like OpenStack. These front ends funnel through a common VM management library -- libvirtd -- which greatly simplifies debugging provisioning faults.
Using a 'real' operating system kernel, rather than a 'hypervisor' kernel which only runs VMs, is seen as advantageous. Since each VM appears as a Linux process, a lot of existing Linux tooling can be used. As a result the management of KVM can seem sparse, in reality for some purposes you use the same management tools as for other processes -- such as network alarming and collecting performance data for capacity planning. Sometimes a VM-specific tool is needed. Often this tool works through libvirtd, making it possible to use one vendor's tool to control another vendor's VM. Many of the technologies for managing Linux containers to also be used to manage Linux VMs, such as Kubernetes and Terraform.
Let's hope that the industry will have learned a lesson, "not to put all your eggs in one basket". Today it's VMware, the next day it can be anyone else, like some popular cloud provider. Sure, Broadcom will profit from companies, who need time to migrate to other alternatives. Just it's a bit sad, that because of someone's intention to make profit in short run a good product ecosystems will die.
This is a repeat of the playbook which Broadcom has run before. The acquired company 'bought the business' with an extremely high discount. They were not a profitable customer. It is known as 'buying logos'. Nutanix collects & brags about them: https://www.nutanix.com/company/customers
Broadcom comes in and does not give the extreme discount. It appears to be a 10-15x price hike. It isn't really a price hike, it is the elimination of a huge discount. After acquisition they are happy to let a money-losing customer go to the competition. 24k seats that are not profitable is business Broadcom is happy to see leave and be a burden on their competition. Frankly it frees up resources to better serve the remaining profitable customers.
I'll put cash on the table to bet that Nutanix worked a sweet deal with Computershare which included their CTO speaking at their annual user conference. Their marketing intention being to generate news articles like seen here.
I have no illusions this comment will be unpopular and will generate a flood of downvotes. But it cannot be denied that walking away from unprofitable customers is a smart business decision.
My company is a very small ~200 VMs VMware customer. Based on the number of time's we've opened a support ticket, once, I'd have to say our contract is one of their most profitable.
re: VMware forum search results, I used to avoid all of the scrapers/indexer sites, but now, those are the only ones whose google search links actually take you to the actual forum posts... I hate it.
>24k seats that are not profitable is business Broadcom is happy to see leave and be a burden on their competition. Frankly it frees up resources to better serve the remaining profitable customers.
We are talking software only. The profit is enormous, it is just the support, telemetry, reccuring licencing and distribution cost that adds and they are tiny. Unless they give the software away they make a profit on any customer.
18 months ago we around 1,000 VMs 95% Windows, we're tiny small fry, we've now got just 50 to go. Moved them all to Azure which is far from perfect but way more controllable and cheaper with volume discounts than VMWare ever wille. VMW used to be very good, but the world caught up and there's simply much more choice is you're willing to make the effort to reap the rewards of staying flexible and mobile. VMW are banking on people not being bothered to move but MS and AWS are willing to cut you some serious bulk discounts if you go all-in.
In the good old days, a large org would have a couple of IBM mainframes running production and test. All the system and platform services would be administered by a bunch of sys progs with an ops team to run it. I know IBM kit was expensive but it was efficient in use of resources and people. Today, virtual estates (on prem just as bad as cloud, if not worse) have truck loads of VM cruft many of which no one has any idea what they are for (I've seen this in large clients), because the CMDBs didn't match the VM reality. Nightmare. This has been going on since the 90's.
I would be interested in their consideration or otherwise of other products which use Linux's KVM to implement VMs. Say, OpenStack or Proxmox VE with Kubernetes and Terraform for orchestration and provisioning. Such a choice could drive their software acquisition costs to not much above the training budget if the base Linux was Debian (a common enough choice by cloud vendors). Maybe they feel they need paid support? In any case, the reasons would be interesting.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
