Apple dumps VMware for KVM

kdh

Gawd
Joined
Mar 16, 2005
Messages
806
http://finance.yahoo.com/news/apple-reportedly-dumped-vmware-big-192002459.html

My first thought was.. Good luck with that. The comments section are filled with infrastructer nerd rage. One suggested comment that had me fall out of my chair was for Apple to switch to Hyper-V. While Hyper-V is a great product, you are a complete fool if you think Apple would give any more money then they have to over to MS.

If you think about it.. 20M over 2 years is a drop in the bucket for Apple. Apple would move to KVM, then just dump 20M into it, fund it, build it and then make it the product they want.

anyway, discuss.
 
That's a pretty big deal for KVM, it will put it on the minds of a lot of people in enterprise circles who normally wouldn't have considered it.
 
That's a pretty big deal for KVM, it will put it on the minds of a lot of people in enterprise circles who normally wouldn't have considered it.

Agreed, but you have to be very careful with that sort of rational. You can hear tons of CIOs saying, well if Apple can do, so can we! But thats the sort of very short sighted mentality that completely decimates medium IT sized organizations. Organizations that already run lean and mean, but don't have the resources to rip out an already well established product for one that still has a long way to go. You may save on licensing fees, but you'll crush your opex with the added head count to make it work.

Where I struggle with this move is in the support of the product itself. If VMware eats itself to death for what ever reason, you have a valid 1800 number to call, and get 4 hour support. KVM? Not so much. Why does this matter? If your VMWare cluster eats itself at 3am, you have a support path and a very good chance that there is a way to fix it. KVM? There doesn't appear to be a good way to get support for it other then mailing lists, IRC, or reviewing bug tickets opened on source forge.

In the end it depends on your environment, how critical you think the data is, and what you can and can't take an outage on.
 
The open source hypervisors (KVM and Xen) have a higher combined market share (over 50%) than VMware according to published reports in 2014, which was the last time I checked.

Some agencies (with thousands of VMs) are also migrating away from VMware to Linux containers too, as a response to problems, such as "What do you mean you (tech support) can't identify the physical host the VM is running on?" :eek:
 
As a user of KVM/Qemu in a business environment I approve of this. Honestly, it's very capable.
It even doesn't need tutorials with 20 steps of cryptic commands.

In my case, I just take a bare minimum install of Ubuntu Server, choose the roles 'virtual machine host' and 'sshd server', and that's.... it.
You can launch client software - virtual machine manager - from any place and connect to the server whereupon a monitor welcomes you showing loads for running VMs.
You can easily point-and-click out virtual machines.
Almost just like in VMWare.
It does actually work well enough so you don't need to enter the CLI. The GUI is very capable, you can create, manage, remote desktop, snapshots - the works.

Also, KVM is one of the ways you can virtualize on Linux. The 'big picture' is "libvirtd" which is an attempt to unify the interface so that it will encompass other virtualisation technologies like Xen.

Seriously, I was pleasantly surprised. Lots of drivers available for lots of OS, Windows 7 ran smooth like butter with 200+ MB/S IO throughput.
I can perform heavy IO operations like backing up a VM and work via VNC on another and I feel no slowdowns.

This is on a lowly quad core (no HT) E3-1220 v3 @ 3.10GHz, 16 GB ECC and four WD 7200RPM SATA drives in RAID 10. Fakeraid.

root@ubuntuserver:~# uptime
11:41:04 up 176 days, 4:07, 2 users, load average: 0.10, 0.10, 0.07

I'm a sceptic when it comes to ease of use of various open source projects but that is a decent product.
 
This move would make perfect sense if your IT staff are pretty linux savy.In my environment, the windows team maintains their own vmware clusters. None of them know much about linux, and when they end up on the linux command line of vmware for trouble shooting, its like greek to them. guess all depends on the environment, and the skill set or lack of skillset of the folks who maintain the environment.
 
Agreed, but you have to be very careful with that sort of rational. You can hear tons of CIOs saying, well if Apple can do, so can we! But thats the sort of very short sighted mentality that completely decimates medium IT sized organizations. Organizations that already run lean and mean, but don't have the resources to rip out an already well established product for one that still has a long way to go. You may save on licensing fees, but you'll crush your opex with the added head count to make it work.

Where I struggle with this move is in the support of the product itself. If VMware eats itself to death for what ever reason, you have a valid 1800 number to call, and get 4 hour support. KVM? Not so much. Why does this matter? If your VMWare cluster eats itself at 3am, you have a support path and a very good chance that there is a way to fix it. KVM? There doesn't appear to be a good way to get support for it other then mailing lists, IRC, or reviewing bug tickets opened on source forge.

In the end it depends on your environment, how critical you think the data is, and what you can and can't take an outage on.

I completely agree with you there. At my last job we ripped out Xendesktop/Citrix to implement VMWare Horizon because the IT Director saw it at another hospital and liked the way it looked.. I shit you not.. It ended up costing a lot more than initially projected not to mention the support headaches and user frustration when the way they've connected for over a year just stops working.

I just like to see another player being brought into the big leagues. It seems like VMWare has been the only major player in most executives minds for awhile. Where I'm at now we run Hyper-V and I've been pretty happy with overall.
 
Where I struggle with this move is in the support of the product itself. If VMware eats itself to death for what ever reason, you have a valid 1800 number to call, and get 4 hour support. KVM? Not so much. Why does this matter? If your VMWare cluster eats itself at 3am, you have a support path and a very good chance that there is a way to fix it. KVM? There doesn't appear to be a good way to get support for it other then mailing lists, IRC, or reviewing bug tickets opened on source forge.

In the end it depends on your environment, how critical you think the data is, and what you can and can't take an outage on.

That's what Red Hat is for. If you want support you can get it through Red Hat.
 
This move would make perfect sense if your IT staff are pretty linux savy.In my environment, the windows team maintains their own vmware clusters. None of them know much about linux, and when they end up on the linux command line of vmware for trouble shooting, its like greek to them. guess all depends on the environment, and the skill set or lack of skillset of the folks who maintain the environment.

They may want to start learning Linux because that's the direction the enterprise/datacenter server world is increasingly moving.
 
They may want to start learning Linux because that's the direction the enterprise/datacenter server world is increasingly moving.

I am surprised this is not old news... The cloud IS LINUX, by in large.

And yes, if you need an 800-number to call for tech support, there is Red Hat or even IBM haha. VMware hasn't been the only game in town for quite some time... .;)
 
We run a fairly large ESXi farm. 60+ hosts and over 1000 VMs. I don't see us moving away from VMWARE just to save a few bucks. Plus most of our Admins run when they hear Unix/Linux :)
 
This is great to see, honestly I want to switch to KVM but it's just a lot more work to setup and manage, while Vmware is just turn key. When I built my new VM server I tried to use it and realized just how much work and how much reading I need to do just to do basic things like setup vlans and shared storage, I just gave up and went Vmware. Though with Vmware I'm more restricted unless I want to pay. Eventually I want to look into what it would take to write a decent front end for it though. Nice thing with something open source like KVM is that it's rather versatile in how you can interface with it if you know what you're doing. (not that I do... lol) There's virt-manager, but every tiny little change you do to the VM requires that it be rebooted, even something basic like changing the CD, that alone had pretty much got me to look at vmware. I'm sure there are ways around that, but I could not figure it out and I just wanted something that works out of the box at the time.

Maybe with Apple being involved we'll see more development happen to make KVM more user friendly like vmware. At least we can only hope. But regardless if it's a success for Apple it will be saying a lot about a free and open source product's capabilities.
 
This is great to see, honestly I want to switch to KVM but it's just a lot more work to setup and manage, while Vmware is just turn key. When I built my new VM server I tried to use it and realized just how much work and how much reading I need to do just to do basic things like setup vlans and shared storage, I just gave up and went Vmware. Though with Vmware I'm more restricted unless I want to pay. Eventually I want to look into what it would take to write a decent front end for it though. Nice thing with something open source like KVM is that it's rather versatile in how you can interface with it if you know what you're doing. (not that I do... lol) There's virt-manager, but every tiny little change you do to the VM requires that it be rebooted, even something basic like changing the CD, that alone had pretty much got me to look at vmware. I'm sure there are ways around that, but I could not figure it out and I just wanted something that works out of the box at the time.

Maybe with Apple being involved we'll see more development happen to make KVM more user friendly like vmware. At least we can only hope. But regardless if it's a success for Apple it will be saying a lot about a free and open source product's capabilities.

Check out Promox. It's based on KVM and LXC with network and storage management.
 
This is great to see, honestly I want to switch to KVM but it's just a lot more work to setup and manage, while Vmware is just turn key. When I built my new VM server I tried to use it and realized just how much work and how much reading I need to do just to do basic things like setup vlans and shared storage, I just gave up and went Vmware. Though with Vmware I'm more restricted unless I want to pay. Eventually I want to look into what it would take to write a decent front end for it though. Nice thing with something open source like KVM is that it's rather versatile in how you can interface with it if you know what you're doing. (not that I do... lol) There's virt-manager, but every tiny little change you do to the VM requires that it be rebooted, even something basic like changing the CD, that alone had pretty much got me to look at vmware. I'm sure there are ways around that, but I could not figure it out and I just wanted something that works out of the box at the time.

Virt-Manager is pretty simple. For CDROM drives you just add the device when you create the VM. You can switch out isos or disk drive media on the fly after you do that.
 
Some agencies (with thousands of VMs) are also migrating away from VMware to Linux containers too, as a response to problems, such as "What do you mean you (tech support) can't identify the physical host the VM is running on?" :eek:

If you have tech support that can't identify a host a VM is running on you either haven't given them the right permissions, or the right training. Or failing those two, you don't have the right tech support.
 
If you have tech support that can't identify a host a VM is running on you either haven't given them the right permissions, or the right training. Or failing those two, you don't have the right tech support.

Agreed. 800-tech support can leave you wanting more often than not. It's best to know your systems inside out.
 
That's what Red Hat is for. If you want support you can get it through Red Hat.

Assuing you are running Redhat. KVM is supported to run 2.6.20 main line. Redhat isn't the be all end all linux.
 
They may want to start learning Linux because that's the direction the enterprise/datacenter server world is increasingly moving.

This is an incredibly short sighted statement. People have been saying this for 20+ years. Yet Microsoft in the data center persists.
 
I am surprised this is not old news... The cloud IS LINUX, by in large.

And yes, if you need an 800-number to call for tech support, there is Red Hat or even IBM haha. VMware hasn't been the only game in town for quite some time... .;)

I doubt apple runs Redhat or any other OS vendor unless they have to. I know they run Windows for their SAP environment, but they had no other choice. Apple does sell thier own Server OS called OS X. It wouldnt be a good look for Apple to sell OS X Server, but then thier own backend systems run a 3rd party Linux OS for the majority of day to day activities. You can download the KVM source, and then compile it into your flavor of linux.
 
This is an incredibly short sighted statement. People have been saying this for 20+ years. Yet Microsoft in the data center persists.

How is that short sighted? He didn't say Microsoft will exit the market, just that *nix will continue to grow in datacenter use as it has for a long time. Even Microsoft knows this and built Azure to take advantage of these operating systems.
 
When it comes to public web servers, Linux already dominates the market. I would be curious to know the market share when it comes to applications that run in businesses.
 
Won't be the last company to dump VMware either. Just working through renewals on our large ESXi footprint while also focusing on migrating a massive amount of our on-prem applications to public/hybrid cloud options you can really feel them getting frustrated that they have no answer to their customer bases shrinking datacenters. We are considering dumping our current infrastructure for half the size footprint with Nutanix and also moving to KVM as well. Like i said, considering, as we know this will be a massive undertaking as well, but it ultimately makes sense for the direction our company is heading. Keeping 24/7 systems online on-prem and bursting workloads to AWS/Other during peak business hours. Just wont require us to run what we traditionally had. Not even going to talk about when our container initiative goes into affect how it will probably let us shrink on-prem even more.
 
How is that short sighted? He didn't say Microsoft will exit the market, just that *nix will continue to grow in datacenter use as it has for a long time. Even Microsoft knows this and built Azure to take advantage of these operating systems.

I'll explain and it’s not a technical reason. Telling someone who only uses Windows that they now need to learn Linux because that how the datacenter world is moving is incredibly short sighted. That is not how most mid to large organizations work. Small shops may be able to pull that off.. But even then. If you are a CIO of company who decides to rip and replace all of their MS stuff to a Linux equivalent because “that’s how the datacenter world is moving” will most likely not be the CIO of said company for very long.

What if I was your manager, walked in your cube in said, “I know you’re a linux guy, but I have this MS product we need to install. You will be responsible for it, you will install it, maintain it, support and fix issues with it. That’s just how the world is moving, get used to it”.. You’d eye roll me so hard, that your cube coworker would get vertigo so bad that he’d fall out of his chair. And before your coworker picked himself up off the floor, you’d have already blasted your resume out on monster.

Point being.. you cant just tell your staff to suck it up, and go a certain direction because you think its going that direction. If your staff is any good at what they do, they'd go find another org that needs what they do. Hence, why the original statement is short sighted.
 
When it comes to public web servers, Linux already dominates the market. I would be curious to know the market share when it comes to applications that run in businesses.

definitely agree with you here. But on the inside back office type stuff? Not many Windows Web servers out there, unless you are running share point, but Exchange is king in most mid to large scale orgs. Ripping out Exchange and replacing it with some linux type mail server, while doable is not a smart move. The end user training alone and the problems with moving to a completely new platform completely out weight the benefits. You want to murder a company quickly? Break their email and scheduling system.
 
The difference in perspectives stem largely from the scale involved. SMB vs. Internet scale/Cloud. The latter tends to gravitate towards open source solutions - e.g. KVM, LXC, Apache, Postgre, etc.

small sandbox vs. LARGE Sandbox requirements and commercial considerations.
 
Point being.. you cant just tell your staff to suck it up, and go a certain direction because you think its going that direction. If your staff is any good at what they do, they'd go find another org that needs what they do. Hence, why the original statement is short sighted.

It's not about your manager telling you to go learn new stuff, it's about you, the employee, deciding to learn new things to enhance your personal growth and career options. No one told me I should go get a Masters degree, but if I hadn't, I'd be at a huge disadvantage compared to where I am now. I could still have a fine career, but the one I have now is great as a result of the doors opened to me by learning new skills.

If I could hire an admin to work in my datacenter who only knows Windows and another who is comfortable with Windows, Linux and more, the guy with more breadth is getting the job every time. Being an admin is about way more than knowing the OS, but if you've only bothered to learn one because no one ever told you to learn more, you're about as valuable to me as a developer who only works in Ruby.
 
Assuing you are running Redhat. KVM is supported to run 2.6.20 main line. Redhat isn't the be all end all linux.

Never said it was. When it comes to Linux support there are tons of places to get it. However, there are benefits w/ Redhat just as any other. However, when it comes to virtualization you'll get access to virtualization tools either not offered or not supported elsewhere and vice-versa.
 
It's not about your manager telling you to go learn new stuff, it's about you, the employee, deciding to learn new things to enhance your personal growth and career options. No one told me I should go get a Masters degree, but if I hadn't, I'd be at a huge disadvantage compared to where I am now. I could still have a fine career, but the one I have now is great as a result of the doors opened to me by learning new skills.

If I could hire an admin to work in my datacenter who only knows Windows and another who is comfortable with Windows, Linux and more, the guy with more breadth is getting the job every time. Being an admin is about way more than knowing the OS, but if you've only bothered to learn one because no one ever told you to learn more, you're about as valuable to me as a developer who only works in Ruby.

Thats excellent you went and picked up your masters and I have a lot of respect for that.

Depends on your organization. In smaller shops a jack of all trades makes sense. In larger shops, it doesn't make sense. You get to a point where a jack of all trades doesn't cut it and you need him to focus on a specific skill set. As an example, I do only storage in my org. I'm maintaining close to 3+pbs of storage and 6+pbs of backups on my own. Expecting me to also pull linux sysadmin duty as well? Not going to happen.
 
Never said it was. When it comes to Linux support there are tons of places to get it. However, there are benefits w/ Redhat just as any other. However, when it comes to virtualization you'll get access to virtualization tools either not offered or not supported elsewhere and vice-versa.

Im not sure what you are getting at here..

KVM has no real support structure. Your response was to run Redhat and get support. My response is not everyone runs Redhat. Youre still talking about Redhat support. If Apple runs OSX Server in thier compute farm, Redhat support isn't going to do anything for them. Apple will be right where I originally stated.. Looking at forum posts, mailing lists and lurking IRC chat channels for help. You could google the problem.. But if I have to tell my management that thier infrastructure took a dive, but its all good because I made a post on a forum asking for help? I will be most likely be out of a job by the end of the day.
 
The difference in perspectives stem largely from the scale involved. SMB vs. Internet scale/Cloud. The latter tends to gravitate towards open source solutions - e.g. KVM, LXC, Apache, Postgre, etc.

small sandbox vs. LARGE Sandbox requirements and commercial considerations.

youre only thinking there are 2 kinds of perspective here. My org is fairly small, but my infrastrure is very large. When Im building out my environemnts, I'm not thinking Internet Scale / Cloud. Thats just an entirely different type of buisness then what I am in. In your example, you are 100% correct that Internet scale / Cloud tends to gravidate towards open source solutions. I'm in the finanical sector which is a completely different animal. I don't really care if your chumpy blog goes down for 20 minutes because of some cloud hosting provider. But I guarantee if you lost access to your bank account for 20 minutes when you are standing at the grocery line trying to pay for them, you'd lose your mind. Hence my solutions are very different then what happens at an Internet/Cloud Provider.
 
youre only thinking there are 2 kinds of perspective here. My org is fairly small, but my infrastrure is very large. When Im building out my environemnts, I'm not thinking Internet Scale / Cloud. Thats just an entirely different type of buisness then what I am in. In your example, you are 100% correct that Internet scale / Cloud tends to gravidate towards open source solutions. I'm in the finanical sector which is a completely different animal. I don't really care if your chumpy blog goes down for 20 minutes because of some cloud hosting provider. But I guarantee if you lost access to your bank account for 20 minutes when you are standing at the grocery line trying to pay for them, you'd lose your mind. Hence my solutions are very different then what happens at an Internet/Cloud Provider.

I work for a large financial firm(bn's) as well, we are going full on push to hybrid/public. I dont care what your industry is, you build your applications to survive failure. Its just a comfort level that we are now becoming OK with and are now making a massive push to live out in magic cloud land(AWS mainly).
 
If that works for your environment, then go for it. In my world? No way in hell I would push to move any of our data out of our data center and into a private provided cloud. Its called possitive control. I know exactly where the data is at all times. If you move your data to a private cloud, you just gave hackers a 2nd attack vector. Being in financial, you must be aware of PCI. No way I would ever release control of our PCI data to live on a 3rd party platform. I say that in the current org I am in today. In my next gig? It could be completely different.
 
kdh, i know what you mean. my day job is at a company that makes servers. a *lot* of our customers are in the financial area, and the anal retentive factor is amazing. i don't blame them at all, given the consequences of outages/hacks...
 
kdh, I don't think you fully understand what a cloud provider is. Besides the fact that "cloud" is just a buzzword. You say Clouds gravitate towards open source solutions. Do you think VMware and Microsoft run their cloud on open source solutions? :rolleyes:

I work for a company that offers cloud hosting, private and public. We use VMware for everything and are using VMware's Service Provider Program which is for cloud providers. What makes us a "cloud" provider is simply that we are not located on-premise at your company. Where are we located? At major datacenters. These are datacenters with class 4 power. Not only does the DC have true dual feeds, our cabs are setup with both feeds. Feed one goes to PDU A and feed two goes to PDU B. Makes it very unlikely for a power outage. As for Internet, for one, we are also an ISP. We always have redundant core routers with a 40Gbps backbone. We peer with other tier 2 and tier 1 ISPs and make good use of BGP. Chance of an Internet outage is very unlikely. Occasionally there are issues with upstream providers, but BGP helps mitigate issues, or we can simply re-route traffic.

Prior to this I worked at a very large financial company. You know what the difference is when it comes to datacenters? They house their datacenter on-premise with class 2 power. Even many companies with their own large datacenter generally don't have better than class 3 power. For Internet they use Verizon and ATT. Not bad, they are both large tier 1 ISPs. But funny thing is, almost all issues I have seen were with these tier 1 providers. Hell, last week one of our customers had an issue that they thought was on our end because a VPN tunnel went down. Nope, their VPN tunnel goes to a company called Armor which is another large cloud provider and the issue was with CenturyLink, a large tier 1 provider.

Point is, these are all datacenters. It comes down to how well they are designed and how well you design your infrastructure within that datacenter. A private cloud could simply mean you are housing your servers with my company instead of in your on-premise datacenter. And in many cases the datacenter we use and our infrastructure design is more highly available than what many companies have on-premise. Now if you are running your own Exchange server and designed it poorly, well that is kind of your problem. :p And I don't mean "you" specifically. I mean a company running their own Exchange. If you want Exchange highly available with DAG in multiple locations, no problem. A lot of people do this and then deal with DNS when failing over from one DC to another. How about geographically redundant IP addresses? I can do that for you.

Infrastructure is only as good as you build it and "cloud" is just a buzzword.
 
Im not sure what you are getting at here..

KVM has no real support structure. Your response was to run Redhat and get support. My response is not everyone runs Redhat. Youre still talking about Redhat support. If Apple runs OSX Server in thier compute farm, Redhat support isn't going to do anything for them. Apple will be right where I originally stated.. Looking at forum posts, mailing lists and lurking IRC chat channels for help. You could google the problem.. But if I have to tell my management that thier infrastructure took a dive, but its all good because I made a post on a forum asking for help? I will be most likely be out of a job by the end of the day.

I thought I was pretty clear. Apple isn't new to open source. It's committed a lot to it. Apparently you think everyone at their Genius Bar is somehow equally keeping their servers running. Apple itself isn't likely to need much in terms of support from outside vendors. It is a a top tier software developer. I think they can handle bugs and software development. They kind of make the most popular phones, the software that runs them and even engineers the chips that drives them. To pretend that their move to KVM forces an administrative assistant to seek help from Google to turn the server on is kind of humorous if not right out comical.
 
Oh, and I also forgot to say, what is wrong with KVM? With the right management tools, it is great. Such as Nutanix's Acropolis gives you HA, DRS, etc like VMware. Linux engineers can make big bucks, so it isn't necessarily a bad skill set to have, especially if you are good at development. We develop a lot of our own tools that work just fine and save tons of money on buying third party products. If you have the engineers and developers to do it, why not?
 
Back
Top