Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On
-
@scottalanmiller said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@stacksofplates said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@scottalanmiller said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@dyasny said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
Glad to see you stopped being such a die hard Xen fan
I remember you fighting tooth and nail against KVM on spice forums.Not so much a fan, as choosing the best products at the best time. Xen was superior for a long time. KVM had a lot of deficiencies. KVM has pulled ahead in a lot of ways. I don't choose or promote products because I like them, I do so because I believe that they are fitting the specific need best.
That means, that naturally, as the products change over time their place in the universe changes. And nothing is for everyone. Xen still remains a viable player, KVM just makes sense for more shops, more of the time.
KVM has been the same since ~2013 she I started with it. The only difference is Cockpit. All the other tools have existed for quite some time.
Under the hood to some degree, but tooling and market support have changed a lot. KVM was doing okay, but wasn't the big leader in 2013 like it is now. It's place in the market is important, too.
What tooling? Libguestfs, virsh, and Virt-Manager have been around forever.
-
I've not tried XCP-NG/XO since they released it. But my go to is ESXi.
HYPER-V I just found a pain in the rear to admin. Just found sometimes I could manage it with manager built into windows, then when trying on my laptop it wouldn't then it did, then when on VPN again hit or miss.ESXi For me. Is the simple install. Boot. Hop onto the web interface to administer. From there create and change VM's.
Like other solutions backup is separate.
At the moment I'm using Unitrends free appliance and doing agent less backups as all the VMs are Linux based so simple.But of course this is my preference and experience.
-
Cockpit doesn't let me setup the networking I want though.
But VMM does.
Once changed, cockpit, of course shows it. But is stilled greyed out.
-
Not the one you want, you are limited to NAT for now
Really hoping that that is included in 30. In 28, even the menu didn't exist.
-
@scottalanmiller said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
Not the one you want, you are limited to NAT for now
Really hoping that that is included in 30. In 28, even the menu didn't exist.
. . .
-
@Dashrender said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
KVM - you need??? in case of VMWare, you need (let's say Veeam - or anything else you buy/Open Source), in
- GhettoVCB is open source and has been around for years for VMware.
- KVM lacks a common CBT API. You have one off's of proprietary overlays and forks. It also lacks anything like VAIO as a kernel API enabling IO split and data service insertion (handy for crazy low RPO/RTO stuff).
It's worth noting that people running 2TB and 5VM's are normally just going to use an agent-based solution anyways...
-
@JaredBusch said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
Once changed, cockpit, of course, shows it. But is stilled greyed out.
Not to rub salt in the wound, but this is stuff that ESXi's had for a decade. It's why I hate "feature checklist" or over-focus on 1-2 aspects of a system. It ignores the operational realities of what most people's day 0, or day 2 operations look like.
If Free KVM UI's were full featured, and intuitive, and had a low-cost support option for backing them I think we would see Scale Computing and other KVM appliance offerings have zero VC/Market Cap. Given these products have existed for 10 years, and they struggle in these most basic of ways I don't feel there is a huge amount of money going into solving this problem (and by proxy) not a lot of market demand.
The fight for the management plane has moved on from Hypervisors (That war is frankly over) and has moved on to containers, hybrid cloud management, networking/security and a host of other things.
In containers Kuberentes has "won", but there's a lot of adjacent product space for making things like Networking and security, not a dumpster fire. The reality is that ESXi "won" the on-premises datacenter war, Hyper-V's entire focus is on Azure/AzureStack now (fighting VMware on a full SDDC stack and Hybrid Cloud). Talking to one large OEM recently Microsoft has pulled all funding for headcounts on competitive Hyper-V trying to displace ESXi on premises. They recognize that the fight has moved on.
KVM is showing up in a few turnkey appliance vendors (Scale Computing, NTNX) but from what I'm seeing adoption numbers are a rounding error of the total addressable market (Some could argue though that everyone underestimates the growth of on-premises IT, especially on the edge making this still an underestimated TAM). I suspect we'll KVM in IoT platforms on a net adoption rate, but not in people actually having SSH to the platform or consuming it at large scale as a pure open source, roll your own. KVM in public cloud (like AWS) still has solid market share growth going on (As it slowly replaces the massive amount of Xen especially at AWS).
Meanwhile OpenStack continues to lurk in Telco where it's gotten a solid footing despite the rest of us all forgetting it exists
-
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
The reality is that ESXi "won" the on-premises datacenter war, Hyper-V's entire focus is on Azure/AzureStack now (fighting VMware on a full SDDC stack and Hybrid Cloud). Talking to one large OEM recently Microsoft has pulled all funding for headcounts on competitive Hyper-V trying to displace ESXi on premises. They recognize that the fight has moved on.
From what we see here in the SMB, this seems to be the opposite. There was a time when this seemed true, now it seems the opposite. ESXi lost the war and is now only in the places that haven't updated in a while. It's still lingering, but you don't hear people talking about deploying it again. Sure, because it was so many places and so many shops just never update, it's still around a lot. But we used to discuss it constantly as a key consideration, now it doesn't appear to be any almost anyone's deployment radar.
To say that they won the war, when they appear to have given up, seems odd. In the enterprise I assume that ESXi is still dominating, that's a different game. But in the space where no one is keeping track of deployment numbers, it used to be ESXi that was always being replaced. Now it is Hyper-V.
-
@scottalanmiller said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
But in the space where no one is keeping track of deployment numbers, it used to be ESXi that was always being replaced. Now it is Hyper-V.
It's a misnomer that people don't try to track free installs. There's phone home telemetry for vendor products, download tracking and a host of other voodoo to at least make an attempt at reconciling these data sets. IDC and the like tend to only report revenues, but there are ways to track embedded ESXi free edition at the OEM and distribution level as well as the phone home level etc. Looking at unique downloads of security patches isn't a terrible proxy for active installs that are being maintained. There are also groups like IDC who conduct phone and other survey's in line with statistical models. Vendors who work in the ecosystem and have to make product judgments of where to invest in also do their own private tracking (Backup vendors do some fairly large stuff).
Are you sure you are doing tracking and statistical modeling of this, or are you just reviewing anecdotes and assuming the plural of an anecdote is data? I feel like we were just arguing 3 years ago that Xen going to take over the world.
@scottalanmiller said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
To say that they won the war, when they appear to have given up, seems odd. In the enterprise I assume that ESXi is still dominating, that's a different game. But in the space where no one is keeping track of deployment numbers, it used to be ESXi that was always being replaced. Now it is Hyper-V.
To clarify, the "war" was claiming a major stake in revenue and market share when Hypervisors and Hypervisor management software was still in Hypergrowth and a double-digit CAGR. That war is over (It's single-digit growth at best).
I'd argue It's the other hypervisor vendors that have largely given up in this space...
-
Microsoft shifted licensing for datacenter and VMM to per core to drive up costs for server edition and applications and shift customers to hosted providers (making SPLA look a lot cheaper at small volume). They've cut their bounties for partners to deploy Hyper-V, removed partner competitive resources and adjusted field compensation to focus entirely on Azure/AzureStack. Hyper-V as a stand-alone free entity isn't a priority beyond it being a sum of the parts that they want to sell. I'd argue this was a smart move (Stop trying to get a piece out of a market that is in single digit growth which is where revenue is for stand-alone hypervisor management products is today). I would summarize what I've seen of their strategy as "Go to Azure, or go somewhere else". Oracle's pursuing a similar strategy with their cloud (Field Rep's only get paid on cloud consumption).
-
RedHat failed to gain any meaningful market share for KVM and their management platform. I'd argue their new focus is more around containers, and PaaS things. Curious if the marriage of this could make IBM cloud a much more diverse beyond IaaS competitor in the field. Note, IBM Cloud has a massive amount of VMware in it (SDDC as a service in the form of VCF popular for the bigger shops). IBM's service heavy leanings and existing customer base put them at the least married to the hypervisor that they de-facto own/control. KVM could become more strategic (Get everyone off LPAR's and VMware) or it could become less strategic (Redhat has a lot of other IP that will work fine on multiple hypervisors and makes plenty better margin). My bets on the later, but we will see.
-
Citrix last I heard has been losing market share with XenServer and Xen itself is losing market share in public clouds in general.
-
There's a massive long tail of other niche vendors, none of which I"m aware of having 1% market share in Hypervisor management revenues. The Cloud Service Providers (AWS/Microsoft/Google/IBM) are sucking all the air out of the room as public cloud expands. There could be an argument about Hybrid Cloud revenue being a market split between both, but that's still a growing market.
There are other ancillary markets still with a lot of growth that there's still strong battles going on in. There are a few minor players trying to bundle HCI with higher level functions (Containers etc) but I'm seeing more of these falling off analysts and revenue tracking radar than being added on.
-
HCI in Hypergrowth, which is mostly being deployed on vSphere. Analysts underestimating the TAM on this one should be a meme. I do think we are past the uniform BBQ phase.
-
Cloud Management Products (CMP) - VMware still leading this and outgrowing market but there's a decent bit of diversity and I'm convinced this one will continue to fracture for a while until we see more consolidation. It's worth noting that the CMP's were chosen for app's today will likely Zombie for another 20 years if CA and BMC's continued relevant irrelevance says anything.
-
Advanced micro-segmentation and security services - There's a lot of growth but this is currently a small space due to the massive capital requirements of this arms race. I expect security, in general, to end up with fewer players in the long run. Too many vendors cause much confusion on operationalizing them and gaps form.
I have to agree with Jeff Ready. The Unicorn BBQ phase is on. We are not going to see large scale growth of new competitors entering now. I think while HCI has plenty of growth, there is going to be an increasing push to show revenue to maintain investments and this is going to leave a lot fewer players at the dance.
-
-
@StorageNinja Net, net you feel/believe ESXi is a better play for SMB's?
-
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
Not to rub salt in the wound, but this is stuff that ESXi's had for a decade. It's why I hate "feature checklist" or over-focus on 1-2 aspects of a system. It ignores the operational realities of what most people's day 0, or day 2 operations look like.
Rub all you want. it is 100% true.
Scott is in a very narrow world, heavily influenced by his bias.
-
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
It's a misnomer that people don't try to track free installs. There's phone home telemetry for vendor products, download tracking and a host of other voodoo to at least make an attempt at reconciling these data sets. IDC and the like tend to only report revenues, but there are ways to track embedded ESXi free edition at the OEM and distribution level as well as the phone home level etc. Looking at unique downloads of security patches isn't a terrible proxy for active installs that are being maintained. There are also groups like IDC who conduct phone and other survey's in line with statistical models. Vendors who work in the ecosystem and have to make product judgments of where to invest in also do their own private tracking (Backup vendors do some fairly large stuff).
Are you sure you are doing tracking and statistical modeling of this, or are you just reviewing anecdotes and assuming the plural of an anecdote is data? I feel like we were just arguing 3 years ago that Xen going to take over the world.How dare you bring in facts!
-
@JaredBusch said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
Not to rub salt in the wound, but this is stuff that ESXi's had for a decade. It's why I hate "feature checklist" or over-focus on 1-2 aspects of a system. It ignores the operational realities of what most people's day 0, or day 2 operations look like.
Rub all you want. it is 100% true.
Scott is in a very narrow world, heavily influenced by his bias.
In the REAL world, MOST people don't touch their virtualization environments. They install and leave it. In the enterprise and bigger spaces, or for MSPs, we touch them a lot. For MOST people, they basically never look at them again. People who work in IT often get obsessed with features that normal shops never look at and we deal mostly with systems that we are touching and forget how little the average system gets touched.
-
@JaredBusch said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
It's a misnomer that people don't try to track free installs. There's phone home telemetry for vendor products, download tracking and a host of other voodoo to at least make an attempt at reconciling these data sets. IDC and the like tend to only report revenues, but there are ways to track embedded ESXi free edition at the OEM and distribution level as well as the phone home level etc. Looking at unique downloads of security patches isn't a terrible proxy for active installs that are being maintained. There are also groups like IDC who conduct phone and other survey's in line with statistical models. Vendors who work in the ecosystem and have to make product judgments of where to invest in also do their own private tracking (Backup vendors do some fairly large stuff).
Are you sure you are doing tracking and statistical modeling of this, or are you just reviewing anecdotes and assuming the plural of an anecdote is data? I feel like we were just arguing 3 years ago that Xen going to take over the world.How dare you bring in facts!
Those aren't facts. Fedora, for example, does not phone home. You can say that people "try" to track free installs, but the FACT is that people do not, an can not.
That's a fact. We all know that our installs aren't tracked. I know that my customer installs aren't tracked. And we know that the non-open source vendors are incredibly incentivized to want the lack of data to be perceived as being a lack of installs.
-
@FATeknollogee said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@StorageNinja Net, net you feel/believe ESXi is a better play for SMB's?
Well of course VMware feels that way.
-
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
- Citrix last I heard has been losing market share with XenServer and Xen itself is losing market share in public clouds in general.
Citrix exited the game, so that's not really meaningful. They were never serious about virtualization. And they've been losing to KVM.
-
@FATeknollogee said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@StorageNinja Net, net you feel/believe ESXi is a better play for SMB's?
Disclaimer: I work for VMware, but my prior gig was an MSP/Consulting and then before that, I was a customer (in a verrry small SMB maybe 4 million total revenue a year).
A few thoughts...
-
It's dead simple to install as it comes pre-installed for no real added net cost by most OEM's (Pre-installed to SD card, SATA DOM, M.2 drives). Even if you do an install it's largely mashing enter with things like creating partitions and sizing them is all automated.
-
It comes with a great HTML5 interface baked into it out of the box (The H5 Host Client is free!).
-
Just because a company is a SMB today doesn't mean it will remain one tomorrow. I"ve seen people grow on VMware from 5 VM's to thousands. While there are dangers to aggressively over buying for growth the Essentials and essentials plus bundles cost less than my houses Starbucks/energy drink habits on a daily cost basis over a lifetime. Powerful central management, alerting, features that can scale with you for less than my last office spent on coffee is being made into a much bigger deal than it is.
-
Skills wise I can throw a rock and hit someone who knows how to use vSphere and unlike Microsoft flat rate 24/7 mfg support is cheap (~1200 a year for essentials plus). At 3AM when I'm trying to figure out why something isn't working I'd rather call someone than poke around in forums. It's true you can get this from RedHat but it actually costs a good bit more for a subscription last time I checked.
-
If your SMB ever needs to deploy software that has very restricted HCL's and supported configurations (SAP HANA, Intersystems Cache, EPIC, A good number of PBX's) you are often limited to vSphere/AIX etc. You might not be there now, but even some of the smaller EMR's can be. You might say "Well I'm going to control every software purchase we make" but that's not the reality of how company purchasing evolves. Also, you might end up M&A'ing a company who has a large presence on VMware/AIX/etc. It's worth noting that some companies (SAP) don't just say "Linux is Linux". they will often force very specific distributions (Redhat and sometimes SuSE). As companies grow they often own less and less of the code they run and are at the mercy of various black box packages and what they run. Most people would rather have one platform for everything over one AIX box to deal with Caché.
-
When I consulted I worked with a lot of M&A where SMB's merged or were purchased. Given how often vSphere was deployed it made the roll in, and roll out of VM's a lot easier. Anything else (physical, legacy Alpha, Hyper-V etc) all got platformed. I saw a lot of xxxx --> vSphere but never anyone seriously going the other way as part of a M&A.
-
-
@scottalanmiller said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
Those aren't facts. Fedora, for example, does not phone home. You can say that people "try" to track free installs, but the FACT is that people do not, an can not.
IDC survey's, Free installs tend to have default mirrors they check out security patches from (It's true some people run their own repo's and do a fan out, but that fan out doesn't get wide in SMB's), 3rd party inventory tools that collect and anonymize their stats. There are ways (again, I view the war as actually making revenue as well as market share).
-
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
- There's a massive long tail of other niche vendors....
This is misleading because you are looking at the market from the "vendor" perspective - your view is vendors and revenue. But from the IT side, market is by products and deployments. Everything you list is a vendor and a revenue stream, not a product and a deployment. That stuff is meaningful when you are a vendor or an investor on the vendor side and you are looking about where to invest your funds. But we aren't on that side, this is the IT perspective and/or customer perspective here that we represent and none of that matters in the least. What we care about are the products (ESXi, Hyper-V, KVM, Xen), what they cost, how they are used, what we get for our effort or money.
Bottom line, having loads of experience with all four, today I get more bang for the buck both in money AND in time with KVM. Sure, people who haven't learned all four tend to see whatever they are used to as the easy one. But unless you are doing all of them, and in a recent context, you can't really tell.
That's why it seems like it is reasonable to think that deployments can be tracked, because the perspective is "revenue" from deployments. But the biggest deployment factors are the ones that aren't tracked.
In the real world, even many companies have no idea how much of their own stuff is deployed. That stuff is rarely centralized. Even when I was at IBM, IBM didn't know what OSes or hardware was in use in production. It was all isolated in small departments that didn't communicate centrally.
-
@StorageNinja said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
@scottalanmiller said in Why I Feel KVM Is the Easiest HyperVisor to Learn the Basics On:
Those aren't facts. Fedora, for example, does not phone home. You can say that people "try" to track free installs, but the FACT is that people do not, an can not.
IDC survey's, Free installs tend to have default mirrors they check out security patches from (It's true some people run their own repo's and do a fan out, but that fan out doesn't get wide in SMB's), 3rd party inventory tools that collect and anonymize their stats. There are ways (again, I view the war as actually making revenue as well as market share).
IDC surveys of WHOM? Anyone ever heard of someone in the real world getting an IDC survey? It's easy to say things like surveys tell us something, but surveys are designed to get results desired. You can make a reasonable sounding survey to get any result you want. Just look at SW surveys, they are all engineers to have false answers and whatever they want to not be represented just is left off.
Most surveys pre-isolated the desired audience. In SW, as an example, their community is built by acquiring interest through people who used the software, software that was free, collected data illegally (but obviously), and only ran on Windows. So any survey of that audience is almost entirely isolated to environments that are price conscious, but run Windows anyway, but don't take security seriously, and don't try to acquire open source. So the results of any survey are obviously insanely skewed.
How does IDC find people to survey? I doubt that they stop by small shops in rural communities and get universal access to IT departments to ask them questions. It's absurd to think that someone like IDC would have any visibility into the market. I mean... seriously, think about it. They only get access via vendors, no vendor to promote the survey, no reasonable access to people to survey. It's "magic" how then the surveys always favour vendor based relationships.
