Enterprises can't take them seriously...
... until they have at least 3 tiers of certification programs and employers can put "must be Proxmox Grand Ultimate Professional with 20 years of experience" into job ads.
Open source virtualization project Proxmox has delivered the first full and stable release of its Datacenter Manager product, making it a more viable alternative as a private cloud platform. Proxmox’s Virtual Environment, a platform that hosts virtual machines and containers, and includes software-defined storage and …
I would have been able to claim an extra year experience with two or three product launches over the years, as I was involved with the alpha site setups and testing before even the early adopters were going to have access. Ah, the times of youth and glory and absolute naive dedication to my craft and believing that I would be rewarded for diligence...
The one I'm most pleased with of such tools is Java - I was lucky enough to work on a contract with a Big Corporation that had early-access to the Java development images for evaluation purposes. I must admit - even then, Java was surprisingly stable and feature rich compared to the C++ programming environments that were in vogue at the time; I'd already moved beyond the 'C' phase of my career.
So, yes, I worked with Java for six-nine months before the first public release. Too bad I can't even name the corporation 'cause it's all under non-disclosure agreements that don't terminate as long as either party exists. I guarantee you know them.
Less a point of pride was being involved with early alpha testing of drivers for the yet-to-be released 386 version of SCO Unix.
"must be Proxmox Grand Ultimate Professional with 20 years of experience"
It's imperative that this appear on job ads 6 months after the associated cert becomes available to ensure nobody applies for the jobs and nobody qualifies for the jobs.
I remember getting the VI3 VCP cert from VMWare not long after the exam existed and relative close to when the product existed and I was seeing job ads that required 10 years experience using the product. I don't think the first version had been around 10 years at that point...it certainly wasn't in mainstream use, because VI3 was the "golden age" of ESX. That's when it started appearing everywhere.
I know for a fact I was one of very few people that had the cert in the UK...there were less than 200 people with it. I was the first person to sit it in the test centre I went to because they had issues setting it up for me...I had to start around 45 minutes late and about 20 minutes in, it stopped working.
There was also only one book available, the official one...because nothing else had been released yet and wouldn't be for at least 6 months.
I passed, squeaked through at 87%, and I applied for around a dozen jobs...got knocked back from all of them, a couple responded saying I didn't have enough experience. I had about 7 years experience as an engineer at that point, and I'd been using the product since release, approx 9 months and I'd been using other VMWare products for many years.
I never went into VMWare work ever again. I went into Citrix Xen instead...without a cert...and earned more money.
Weirdly though, a colleague of mine at the time...a chap that constantly wrote off virtualisation and thought it would never take off...eventually ended up working for VMWare...which is probably why it ended up where it did...if you have a whole swathe of staff that never believed in your product, you're going to go down the tubes.
For sure. At least so far, it seems nothing is held back by Proxmox and nothing is crippled like ESX/VMWare.
I've been using Proxmox for many years (but I go way back with virtualisation) and it's just a dream product for me. I haven't once hit a snag that forces the purchase of a license. I have purchased basic licenses for my own setup, purely because I wanted to support the project and I feel they deserve it...some of my clients have enterprise licenses, not because they need it for functionality, but because I was able to sneak it in and support packages are easy to wave through, it's that reasonably priced.
I don't agree with per socket pricing usually, I think it's a bit unfair (purely because a company with more money can afford CPUs with higher core counts and dodge the per socket licensing to a certain extent, socket count is a bell curve most of the time, small companies have a single socket usually with a low core count, large companies have the resources to have high core count single socket gear for higher density, but companies in the middle tend to have multiple sockets with modest core counts and the worst density)...but in the case of proxmox, it's cheap enough that it's not a concern.
I was going to say it didn't do everything but did a quick check, because my knowledge was pre-Covid and it seems to be better. It's still a big beast private cloud and I wouldn't try it on a smaller estate, whereas Proxmox looks like you could build up and environment to match a small ESX cluster.
I'd ask about tools to assist VM porting and also application support but if you're going the OSS route it may not matter.
Which version of open stack? It seems like there are loads to choose from, and they are all quite a bit more complicated than proxmox.
Proxmox has it's quirks, but once one has them figured it out, it's quite stable and useable,
One thing I am still unhappy about is the placement of the VM config files in the clustered file system in the OS as opposed to shared storage. It makes storage based replication of large environments very difficult compared to Hyper-V or vSphere.
As I understand it, OpenStack has traditionally been less of a product and more of a project. Like Kubernetes, it's not so much a thing an individual sysadmin installs as a complex monstrosity wrangled by a team. The project's software page illustrates this fact: to implement OpenStack, one needs to understand all the cutely-named components and how they interoperate, whereas Proxmox appears to be much simpler.
While I can't personally speak to the Kubernetes analogy (maybe I should be thankful for that ;-) ) I do agree with the "product" vs "project" take.
E.g. Proxmox IME is probably a good fit if you're replacing a (few) standalone hypervisors (like ESXi) or even clusters (like vCenter). At that level it scales decently, and has enough features (VM cloning, shared storage, reasonable clustering etc.) that it checks the table-stakes boxes you'd expect from a modern hypervisor setup. A somewhat experienced systems person can work out the basics of Proxmox in a day or two, and end up with something at least functional to get started.
OTOH I don't have enough experience with Proxmox to know how well it meets the needs of infrastructure on a grander scale, e.g. corporation-wide virtualization from server room to desktop, perhaps spanning multiple geographies and such. I expect it's more than a 1-sysadmin job at that level, much as OpenStack (or k8s) or even VMware would be.
From personal experience (not all of it positive) I know that Broadcom's VMware "Horizon" VDI product is popular with some big corporate IT environments. I dunno if Proxmox has an answer to that yet. I'm glad I don't need to worry about that part these days.
I've been running it for somewhere around a decade and I really like it.
I moved after VMWare pay-walled backups on their free offering.
Shopped around, tried out a bunch that were available at the time (xen, hyper-v etc) and Proxmox "checked all the boxes" that I wanted.
-built in backups (there is even a separate free backup server)
-does HA clustering and auto fail-over
-free (but you can pay for support and I encourage you to do so to ease the version update process)
-has a web interface and does not require a separate web node (i.e vSphere). There were some commands you had to run in CLI, but I think they are in the GUI now.
-has KMS memory balancing (similar to VMWare balooning)
-now has ZFS (but I still use a couple OmniOS ZFS NAS's via NFS)
It is stable, I go over a year between reboots sometimes on some nodes
Upgrades can be done without paid; you essentially backup, migrate guests to new nodes, install new version with new name, add to cluster, migrate back or restore.
Yes I am a fanboy, but only because Proxmox has not ever burned me.
Backups...absolutely...VMWare has always been the worst for this...last time I setup VMWare backups (and I'm showing my age here) was when vRanger and Platespin were a thing...and they fucking sucked...they were so fucking slow and you'd inevitably end up with snapshots that get stuck from time to time without you knowing and you'd end up with snapshots of snapshots of snapshots which slowed the whole server down, it was absolute hot garbage).
The important thing about Proxmox and backups is you don't even need to use built in tools, you can roll your own solution without a performance penalty. ESX was the absolute worst for this...you essentially had to use third party solutions because the throughput was crippled (more so) if you tried to use your own solution...straight up SSH transfers were capped at 35MB/s for some fucking insane dumb fucked up reason...I hated it so much...you couldn't even just plug a USB drive in and it "just work".
You can have a lot more variety with Proxmox with ease...I have a local NAS that I have mounted as a storage unit in Proxmox that contains backups, from that NAS I push to Glacier...I also have a local dual slot 2.5" USB drive plugged in that I also backup to for any "quick" restores that I need to do.
Finally, I have a bash script on my workstation that I run once a month that does a local sync to a Sabrent 4 bay USB enclosure for the absolute last resort disaster backup.
Variety is the most important aspect of any backup solution in my opinion, you should be taking backups out in a variety of different ways, and you should have a variety of different ways to put shit back in. That way, if you do have a disaster, you have plenty of options and you won't be hamstrung by arbitrary bullshit...you can get things going again fast.
Prior to Proxmox I used to entirely avoid full VM backups and I'd just backup the contents of the VMs using something like Iperius, rsync etc etc...now I don't have to faff with any of that (although database backups I still do separately as well, doing hourly snapshot style backups, purely if I need to roll back a database to a specific hour over the last 120 hours).
I had a bunch of servers blow up about a year or so ago, they were taken out by a UPS going critical (long story) and sending out a huge surge and some tasty smelling smoke....I had enough spare parts lyring around to repair the servers and I was able to re-import VMs and get things going again in under 3 hours...customer was blown away...they went from full bore panic and rivers of sweat to cracking on before lunch time the same day...with VMWare, I'd have fixed the servers, but the import would likely have taken days...all I had to do was simply copy the backups from a local storage server (that wasn't in the same room on the same UPS) to an external NVME drive, whack that drive in the repaired servers, and import the VMs again...was quick and easy...and painless...
Its even more painless if you use containers instead of VMs because the backups are much smaller.
^ Title. They touched on Veeam BR support for Proxmox, but not netapp ontap or Inuvika VDI, or any other 3rd party that have adopted support on Proxmox now. With PDM and it's planned roadmap, there needs to be more indepth coverage on the Proxmox product suit up against VMware and Nutanix.
I have run all sorts in my cupboard cloud... Qemu/KVM, oVirt(RHEV), vSphere/ESXi, Proxmox, Xen/Cloudstack, OpenStack, k8s clusters etc.
Honestly VMware's ready to go product is still years ahead in terms of a vendor-supported suite for *on-prem* compute. If you want/need to cloud then fine, AWS, MS, Google etc.
For in-house though... only VMware passes the main CTO test, and fundamentally that is: hirability. All of the virt platforms I listed are cool etc, but what most Ent IT departments need when dealing with on-prem infra (so not cloud) is a definable/clear salaried resource pool for recruitment, and for now VMware still wins against Proxmox, OpenStack, Zen/XCP, KVM etc.
k8s/cloud is good for this because it transcends the lower-level infra (sad times if, like me, that is the interesting thing) and a recruiting bozo just has to tick the relevant cloud box on the job ad.
But VMware technology paradigms like vDS, DRS, vSAN, VMFS, PowerDRS (and on and on...) do simply not exist in a contractable/supportable/hirable alternate on-prem virt platform
Not just that. ESXi/vSphere is still light-years ahead in reliability and scalability vs Proxmox. For single hosts and small clusters, Proxmox can work very well, but once a cluster exceeds more than a couple of hosts it gets increasingly fragile. Networking can also be hit and miss, often with strange errors and bugs randomly strangulating performance for no obvious reason. Error messages are cryptic, and the UX isn't very logical/intuitive (and many basic things can still only be done through the CLI, such as enabling hardware passthrough).
The main difference is that ESXi is essentially just a hypervisor and in vSphere, everything else runs on top of it as VM, while Proxmox is not just a hypervisor but also a container host, a SAN, management pane and so on, all which runs natively on the host. This makes the host itself much more complex, and this complexity is replicated on every host in a cluster. Proxmox is much closer to Hyper-V than to vSphere/ESXi.
To replace a large scale vSphere deployment, the only viable alternatives are Nutanix, RHV, or OpenNebula/CloudStack/OpenStack on top of KVM. Perhaps HPE Morpheus as well (but the jury's still out on whether that's actually dependable).
For lots of companies, Proxmox with a support contract is all they need. We're holding out because of the lack of support for SAN.
Larger companies may have different needs, but they also have different resources. I know of at least one company that jumped ship from VMWare to Nutanix precisely because of the price increases.