Hacker News Reader

Comment by vkat | Hacker News Reader

For anyone who doesn't know, VMware has doubled down in open source in the past 3 years, especially Kubernetes. I don't think Broadcom really understands how investment in open source works or for that matter cares.

Half of my org of ~500 engineers works on upstream projects.

Disclaimer: Current employee of VMware

View on HN

After being acquired by Avago, Broadcom ceased to be a single company. It’s a conglomerate of entities/BUs that are explicitly asked to NOT work cooperatively, so that each entity can be sold at moment’s notice if it doesn’t deliver the right financial results anymore.

Chances are that the current leadership doesn’t care about how VMware operates, as long as it cuts enough costs to increase metrics that Wall Street considers important.

Expect major cuts in R&D, receptionists, travel expenses, general benefits, and virtually no hiring of college grads or interns. You’ll be on the phone in a queue for hours to get outsourced IT support.

Support of open source should be the least of your worries.

Stock price might go up though…

I was part of Symantec when we were acquired. I didn’t agree with all Broadcoms moves, but they empowered us to really cut the fact and focus on what we were good at.

I was very surprised when the product quality went up, for the Symantec software I use.

To be fair, prior to broadcom purchase symantec was not known fo high product quality

In the sysadmin world, Backup Exec and Norton products are meme's at this point, examples of the worst kind of software

so I am not sure how it could have gone down in quality

The consumer grade Symantec software and enterprise Symantec software were always different.

Norton never had any real bearing on SEP.

The consumer side of Symantec was not sold to Broadcom and became Norton Lifelock.

Cut the fat, I think you mean.

Not sure if that was a typo, eggcorn, or pun on truth and fiction post-acquisition...

Thank you, I did indeed mean cut the fat. I need to cut the fat on my fat fingers!

Thanks for the clarification - it's the corporate-shoggoth of HP>Agilent>Infineon+LSI>Avago "Broadcom Inc." not underground-dungeon-scandal "Broadcom Corporation". Their NASDAQ ticker is still AVGO. :)

… and here all I wanted was updates to fusion and decent HBAs…

> not underground-dungeon-scandal "Broadcom Corporation"

TIL: https://www.vanityfair.com/news/2008/11/nicholas200811

Broadcom Inc. bought Broadcom Corp. in 2015 (and presumable got its name from it).

Consistent with my experience as a customer of other Broadcom purchases.

This is exactly correct. Hock (love him or hate him) knows exactly what he is doing, and is entirely unapologetic. He looks only at margin. You have $10B revenue stream at 2% profit margin? Hock will happily cut that in half for a 10% margin he can depend on. It's all about ARR. Great for shareholders and those with a mountain of RSU's... but not so great for employees. Was at CA...saw it happen.

> It's all about ARR.

You argued the exact opposite? You're saying it's all about EBITDA...

I don't think so. Certainly, EBITDA is part of any story. What I was trying to convey was that Broadcom very much wants their customers to be engaged in a strong subscription model... big, recurring license agreements. Customers who, year after year, can be relied upon to send Broadcom money...even if they buy nothing new. Broadcom likes those Big Companies who can be counted upon to spend a consistent (but never shrinking) amount every year. Hock is ruthless in this...and has a pretty good bullshit detector... again, he probably makes a lot of people uncomfortable. Nothing wrong with that approach... it's very pragmatic... especially for shareholders. Broadcom might give you more for your money every year, but you will never spend less of it.

It's amazing our system rewards all of those things. Cutting r&d, cutting jobs, cutting downstream contractors, using profit for stock buybacks, etc

Optimizing for efficiency isn't a bad thing. The real problem is the larger economic feedback loop based around keeping the number of hours worked the same. So the efficiency gains are effectively lost, and the average worker is left still depending on the ever-optimized structure rather than being able to accumulate enough wealth to escape it and create new structures.

VMware learned their lesson from OpenStack, where emerging upstream technology threatened their virtualization monopoly. They also were forced to go in this direction by their enterprise customers, who are very dependent themselves on open source technology.

One of the opportunities at VMware is monetize the open source investment more effectively, especially Kubernetes. I still don't totally understand how they plan to do that but wish them (you) all well.

Disclaimer: Former employee of VMware

Given the situation of Docker, it doesn’t look good for the monetization of Kubernetes, or am I missing something (I‘m not too familiar with Kubernetes)?

The cloud vendors, that cooperatively are building Kubernetes, are making bank off of Kubernetes.

Kubernetes is fine.

Multi billion dollar businesses. Just look at AWS EKS.

Given how pluggable k8s is, it should not surprise you that there are companies at every integration point. Tigera for network policy. Security policy vendors using admission controllers to secure clusters. Linkerd/Istio for service mesh. Storage vendors. Etc. Each integration point provides a hook for someone to do better than the default behavior and charge for it.

The ecosystem is thriving and because the k8s platform was built to be pluggable; it is an actual platform. Docker (the company) tried to monopolize and monetize when they built Swarm, and that’s why they lost the cluster orchestrator space IMO. Docker (the tool) is just a tool, not really something you can monetize easily. Docker (the image format) is where most of the value lives, there isn’t much value to extract there as it’s an open protocol.

Docker (the tool) was commoditized by k8s, and now there isn’t really any way to charge for a differentiated implementation. Providing a registry and a best-in-class Docker for Desktop experience isn’t going to be a huge TAM.

So you mean to say it's layer violation soup.

I want to "Integrationware" software that just exists for various vendors to cut deals and piggy-back with oneanother.

> Given how pluggable k8s is, it should not surprise you that there are companies at every integration point.

Of course, this is pretty much what ended up destroying OpenStack outside a few niches. HP OpenStack that only worked with 3PAR SANs was a great example - every finger in the OpenStack pie was trying to push OpenStack to have corners that they could monopolise and tie to their proprietary products.

Docker didn't really attempt to monetize Kubernetes at all until the writing was on the wall, instead they had their own thing (Swarm) that they were pushing but most of the ecosystem went in the other direction.

There's stuff like Openshift and Rancher as commercial Kubernetes distributions as well as a lot of Kubernetes "addons" like weave, portworx, etc. Monetisation of Kubernetes is totally possible.

Docker failed because they tried to compete with Kubernetes, lost and then didn't have a clue how to find a niche where they could enrich the Kubernetes ecosystem.

If you're looking for an example of successful monetization of open source addons for K8s, you could do much better than the example of Weave Net.

The lack of monetization was mostly on Docker itself as far as I can tell. For example, they could have offered secure container supply chains years ago. Instead, companies like JFrog got there first. Public cloud container repos are another example. Amazon and others got there first with products like ECS.

I would be mildly interested in a 5 minute pitch of what VMware is offering these days, it's been 15 years since I've touched any of their products and the way they were doing things then really seems like it is becoming increasingly irrelevant so what have they been doing instead? It looks like my vague guess of "the same but also on top of cloud providers" seems to be about right given a quick glance at their website, but I'm curious to know a bit more.

They have this VMware Tanzu platform which is like Red Hat Openshift. A heavily customized application packaging , deployment management solution on top of Kubernetes. It can be deployed on any cloud and on-prem DCs. So kinda multi cloud bet. They are selling it in a big way.

They also offer tanzu as on prem cluster where you can somehow use kubernetes to provision kubernetes virtualised kubernetes clusters.

It uses cluster-api to provision clusters. It's very cool and isn't a VMware thing. It's open source, and you can easily run it on regular k8s clusters.

https://cluster-api.sigs.k8s.io/

Well, Cluster API depends on underlying providers and the VMWare provider is definitely a VMWare thing as it's maintained by them:

https://github.com/kubernetes-sigs/cluster-api-provider-vsph...

Yes it is a community project but VMware engineers play a big part in maintaining as it is foundational to the kubernetes product that VMware sells. Open source does not mean it runs by itself.

What does "They are selling it in a big way" mean? I've never heard of it, and neither have any of the software engineers who are next to me at the office. RedHat OpenShift also bombed. They had a great start 11 years ago, but then didn't innovate, and worse, released a 2.0 that wasn't backward compatible. They used the 2.0 release as an excuse to kick off all free plans. What a joke. There's a reason Red Hat isn't a leader in Cloud.

They also released a 3.0 and 4.0 of OpenShift that weren't backwards compatible :-)

Yet OpenShift is making bank. You'd be surprised how much money they make on AWS, Azure, and Google.

OpenShift is fantastic and I wouldn't hesitate to recommend it to any medium sized startup and up. I run it in my closet. Most medium sized startups end up building out an OpenShift-like K8s but they use 6 months to a year plus to do it.

Upgrades within major versions are totally painless, one-click affairs.

I have customers running Tanzu.

It is slick, but niche. Incredibly rock solid - in ways K8S is not - and you pay for it.

It is on its way out as a product.

Tanzu is a loaded term :) Are you talking about TAS (ex PCF) or TKG?

PCF

i wouldn't say it's on its way out. it's still in active development and very much supported

Why is it on its way out as a product? Is it simply not making money?

The scuttlebutt I hear is Kubernetes is stealing its lunch money.

Again, referring to what used to be PCF. Hence why the more modern Tanzu is K8s based.

PCF is still making bank but the industry has rallied around K8s. So growth has stalled but it has a happy customer base.

Unfortunately there’s no real equivalent (OpenShift is not really “it”) on K8s unless you’re assembling several products together with a lot of glue. Which is what everyone does, with varying quality.

vmware enterprise footprint and business still seems to chug along. most of the world markets lag behind hyperscale technology progress.

They still make the best virtualization software for OSX. I use it daily to run desktop Linux on OSX. Video drivers seem to be having a tough time with Electron apps these days, sigh.

I run esxi on my (admittedly beefy), home server. It is a fantastic piece of software that can be completely controlled through a web browser. I can easily spin up a couple of VMs, interact with the install without having to use a remote virtual keyboard/video switch, and the whole thing is rock solid.

I could probably do a lot of the same thing with Linux itself, but the web interface is so nice, I'm keeping it as long as it continues to be free for home use.

I’d be really interested in an “i used both” comparison between esxi’s free offering and proxmox for this sort of purpose. I imagine esxi might be more polished, but also more limited.

I use Proxmox and can do everything OP just described. I would assume the esxi is more polished, but I don't spend a heck of a lot of time on my hypervisor; mostly on what it's hypervising.

I've used both, though not for anything especially complicated. Your guess about the polish level is correct; ESXi blows Proxmox out of the water interface-wise, especially for some specific tasks like configuring networks or managing PCI passthrough devices. As far as "more limited" goes, I can't say I noticed much of a feature discrepancy, although being open and just sitting atop KVM, Proxmox doesn't have any of the artificial limitations imposed by VMware (like a cap on number of vCPUs in the free version, eugh).

One pretty big one, if you're looking at homelab use, is that Proxmox is just Debian, so it's waaaay more likely to run on crap-tier commodity hardware, whereas VMware has all the usual enterprise-level support commitments and so you're very much on your own if you try to force it to run on hardware that isn't on their official compatibility list.

Excellent point on hardware support. I'm running several years old Xeon chips in my server, and ESXI is already complaining they might not be supported in a future release.

> One pretty big one, if you're looking at homelab use, is that Proxmox is just Debian, so it's waaaay more likely to run on crap-tier commodity hardware

That was my first (and last) hurdle when wanting to try out ESXi...

I booted up the installer only to be greeted by a "not enough RAM" message. I believe it's 4GB minimum, which my tiny machine had, but ESXi read it as 3.9GB instead of 4.0GB and refused to budge.

Very happy with Proxmox on the other hand! Having access to the underlying Debian is also quite nice if you need to do something that's not supported by the interface.

Proxmox also has native container workload virtualization with LXC container support built right in and ZFS filesystem. Two killer features.

You are better off running a VM for dockers. That way that entire VM is portable.

The ZFS Filesystem is a nice feature

> Proxmox doesn't have any of the artificial limitations imposed by VMware (like a cap on number of vCPUs in the free version, eugh).

These were what i was thinking of yeah.

You can do the same thing with Proxmox VE. IMO, Proxmox is a much better choice for stand-alone VM hosts. The UI alone is certainly miles better.

Where VMware shines is when you want to virtualize multiple datacenters and manage them all from a single browser window. Getting hundreds of hosts and datastores managed all under one vCenter umbrella all while eliminating single points of failure, while not trivial, is the bread and butter of VMware and it works very well once it's all up and running.

not super critical but is there a terraform provider for proxmox? I looked at proxmox a while back when building my home server but went with esxi free because of the community provider - https://github.com/josenk/terraform-provider-esxi

As Someone that runs Proxmox and ESXI.

That is hardly true.

The ESXI UI is vastly cleaner and more VM Focused. The network stack is easier to work with, and setting up passthrough is dead easy. A big plus to ESXI is how easy it is to export a VM and move it to another machine. The Export/Import method on Proxmox sucks, and all CLI. Plus there are a lot of premade VMware images out there that spool up in minuets on ESXI, getting them up and running on proxmox is a chore.

Proxmox is not bad if doing everything from scratch in Proxmox, and you don't need to move VMs arount.

Plus don't get me started on lack of a filebrowser on proxmox.

Performance is better with ESXI anyways. Only downside to ESXI is hardware limitations.

Proxmox fall down on Enterprise tooling and integration with other software enterprises use to run the business

Things like Monitoring Platforms, Backup Solutions (veeam), Automation tool kits (powershell), etc.

I like Proxmox, I use it personally, but in its current form I do not see how any medium or large enterprise that has is management tooling around VMWare could just drop in proxmox.

VMware has several billion-dollar or more businesses...

- the core vSphere hypervisor and management software

- Vmware Cloud on AWS which is pretty big on its own, plus other cloudy things like datrium / disaster recovery as a service.

- cybersecurity (carbon black and secure state), which is similar to what Symantec does, will be an interesting overlap

- endpoint management (security for desktops, mobile) which was AirWatch

- management (monitoring & automation) aka. vRealize and Cloud Health

- workspace / single sign on, similar to okta (Workspace ONE)

- virtual desktops (Horizon) which compete with Citrix

- software defined networking (NSX) which competes with F5, Cisco, Juniper etc.

- modern apps (Tanzu) which is largely open source... containers & developer tools & data services & agile consulting, e.g. Spring, Apache Tomcat, RabbitMQ, Greenplum, Wavefront, Bitnami / Kubeapps, Kubernetes, Cloud Foundry and Pivotal Labs (now Tanzu Labs)

Fascinating … your very comprehensive list does not contain the only VMware products I have ever cared about: workstation and fusion.

And has a lot of "similar to {thing I have actually heard of}"

It reminds me of the first episode of Saturday Night Live after the original 1975 cast all left. In the first episode, the new cast did a sketch in which they each described themselves as a version/variant of/similar to a member of the old cast.

(Everyone was fired after that season, except for Eddie Murphy and Joe Piscopo.)

A few years ago, most of the US dev team for workstation was laid off in favor of a non-US team.

You run a well-known online backup company and I don't, but my reaction was exactly the same: The only VMware products I've ever cared about are Workstation and Fusion, neither of which is on the list.

On the one hand, it's pretty cool that VMware has pivoted so successfully over the past 15 years from being solely focused on a hypervisor (or, I suppose in modern terminology, "self-hosted cloud") business that the cloud largely supplanted. On the other hand, it's sad for someone like me, who used Workstation to run VMs at home (I moved to VirtualBox), and have thought at times about spinning up ESXi.

> a well-known online backup company and I don't

What’s up with this? Did their company steal the name of the tool rsync?

In early 2006 I asked the original author and the current maintainer for their blessing to use the domain "rsync.net" for (re)incorporating the offsite backup component of the ISP I started in 2001[1].

I gave them right of first refusal, etc., and they said my use of "rsync.net" as the name of the company, domain name, etc., was acceptable.

[1] "JohnCompanies"

their main business for two decades has been datacenter virtualization. Individual developers don't care about it, sure, but it's a huge software business.

I think both of those are free now for consumer usage, for at least a couple years now?

No not the Pro versions still

good point… I don’t think they’re billion dollar businesses though. Probably hundreds of millions? Hard to say as I don’t think VMware publishes this info

Workstation and fusion were never big money makers. They were more of a labor of love.

I joined VMware just after the release of Businessware 2.0 - which I agree was 14 years ago.

The big money maker was vmotion because it allowed datacenters to decouple the server from the underlying hardware. Ever after, the money was in the datacenter products.

correction: VSphere 2.0 not Businessware 2.0 (wrong company).

For some reason I could not edit my post so I had to reply to my post.

You covered SDN but they are also the leading SD-WAN vendor through their VeloCloud acquisition 7ish years ago. Workspace 1 is also the core for their SASE products (zero trust network access, secure web gateways, browser isolation etc) when tied with VeloCloud.

Yep I forgot about VeloCloud and SASE. It really is a beast of a company :-)

They own a substantial chunk of the on-premises datacenter virtualization space. Loads of orgs still run their own kit, and VMWare makes some damn nice tools to squeeze the most from that investment. There are competitors in the space, but none with the breadth and depth of VMWare.

Working for a client of Broadcom, I can say that no, they don't understand how opensource works. Have a problem with one of their products which they blame on an upstream bug. They don't allocate an engineer to fix said upstream bug, supplied an workaround that didn't solve the problem just reduced number of occurrences and that's it.

Did they ever stop blatantly violating the GPL in ESX?

Taking driver code out of Linux and just running it inside the hypervisor has to be the most brazen violations possible.

> Did they ever stop blatantly violating the GPL in ESX?

If you mean, did they remove the vmkLinux layer from their product? Yes, they did, back in 2019 soon after they won the lawsuit on procedural grounds (the court ruled the person who brought the suit didn't have standing, and it was upheld on appeal).

It sounds like vmklinux is deprecated but still shipping for older drivers if I'm understanding their documentation. https://blogs.vmware.com/vsphere/2019/04/what-is-the-impact-... vSphere 6.7 (which they're still supporting and shipping updates to as it was only released last June), includes the vmklinux driver stack.

For those reading along at home, that's a shim layer to link GPLed code into the closed source hypervisor.

It's absurd that the company responsible for some of the most blatant license violations in the industry (and is still actively continuing to violate after a decade) is trying to push itself as open source friendly.

How do you propose banning this while permitting open source drivers for Windows?

I understand why you might want to, I just don’t understand what the rule would be.

The key is in "derived from". Very few people would say NT is derived from some random drivers that nobody at Microsoft had a hand in writing. Taking existing code and having it form the foundation of a new kernel and claiming it as all closed source is quite a bit more obviously derived from that existing code open source code. "Derived from" is the core concept in the GPL; it actually doesn't use the word "link" at all.

Additionally they didn't publish the source to even the GPLed drivers they shipped as binaries, before even talking about the core hypervisor itself.

I thought they ripped the network stack from BSD.

https://www.neowin.net/forum/topic/381190-windows-tcpip-stac...

That's fair, I should have kept constrainig my statements to GPLed code. That being said, Microsoft complied with all of the license terms of that stack (basically shipping a copyright notice and constraining themselves on how they advertise it's use). If it was a GPLed stack and they never published source, I'd be saying the same thing about them that I'm saying now about VMware.

They did publish the source to the GPL drivers as well as to the vmklinux “adapter”, and I don’t believe either the core hypervisor or the ESX kernel were derived from or adapted from the Linux kernel.

My impression is mostly people are annoyed that VMware drafted off the Linux driver ecosystem and used it to help bootstrap a proprietary ecosystem.

Eventually (once they were successful) they published a native driver SDK and hardware vendors wrote native drivers for it.

To me this seems very similar to the way Linux had for a while a way to run Wifi drivers originally written for NT, though it didn’t lead to the year of the Linux desktop as some of us might have hoped for at the time.

The difference is that it was the original driver stack and the system didn't have a way to run at all without that Linux code until a handful of years ago.

The ndis drivers are the opposite, because the closed source parts weren't written by people even thinking about open source code. No one could say they were derived from Linux. And Linux had it's own network driver stack so saying that Linux was derived from NDIS would be a stretch too. Hell, I don't think NDISWrapper even was ever upstreamed.

You’re referring to the Console OS which they also did open source (it was just Red Hat, sources available upon request).

The relationship between the Console OS and the Vmkernel / hypervisor was very similar to the relationship between Windows/Linux/macOS and VMware Workstation/Fusion, which were also proprietary software not derived from any of those OSes.

No, I'm referring to the vmklinux driver stack which has always had the purpose of running GPLed code directly by hypervisor without the consoleos mediating access to hardware.

Also, does 'they' actually include 'you'? Googling your username seems to connect you with vmware code pretty intimately.

I can only hope they let VMware operate independently and continue growing like they've done in the past decade.

One thing that I always gave Dell/EMC credit for was that they at least outwardly seemed to have left VMware free to their own thing. As opposed to a lot of their other acquisitions that they bastardized in this rush to sell enterprises on everything private cloud. We'll see Broadcom ends up taking the same route.

Quite agree, Dell/EMC treated VMware as independent.

It is best to set your expectations extremely low for this type of acquisition. My personal opinion: Broadcom is _buying_ its way into the Enterprise technology space by hook or crook. When Broadcom acquires a company their M.O. is to immediately install finance-driven leadership and focus strictly on cost cutting measures coupled with accelerated profit-making endeavors. The only thing VMware employees have to look forward to in the near future are layoffs, all types of cost reductions, team consolidations (i.e. you do more work with less resources) and severely reduced R&D. And the only thing that VMware customers have to look forward to are increased VMware costs, more vendor lock-in, reduced satisfaction with support, and almost zero innovation.

Again, this is just my opinion so I could end up be wrong. Or maybe I am just in a bad mood today... meh

Bad mood i hope.

Should Broadcom do all those things, they run a substantial risk of degrading the products offered by VMware, which is a poor business decision.

Given the very profitable state of VMware, it's a lot more likely that they will only introduce minor changes.

Not at all. Broadcom is well known to pimp up license and maintenance prices for all their sw portfolio while cutting fat and flesh in the process.

Disclaimer: forced Broadcom customer here

So Broadcom is just a better run IBM? IBM buys whatever product it wants for a foothold, milks it, chucks it out.

It used to be Product X becomes Websphere Product X, then disappears from their salespeople's lips about 5 years later.

What does the upstream means?

Reply on news.ycombinator.com