OmniCloud

OmniCloud is just the name that I gave to my home server, but of course there was a reason for it. In this post I’ll be showcasing it a bit and explain what it is, its purpose and kinda the role that it fits into my digital life! (which, if you couldn’t tell already, is quite exciting and never boring xD).

Before getting into detail of what specs this wild beast is running currently, first I wanna talk a bit about how it started and its evolution so that you can kind of get a glimpse of how the concept of it has been changing.

It first began as a cheap, second-hand Dell office computer with a different and bigger chassis so that I could fit a GPU correctly in there. Yup, that’s how it started, an i7 of maybe 3th or 4th gen probably, simply 8GB of RAM and an NVIDIA GTX 1050 Ti. Yup, that was it, because it didn’t need to be more as its first concept was just a machine I could do my remote gaming with. Yup, you heard that right, remote gaming; there’ll also be a post about it since it’s still what I use to play my games. Just… on a completely different level now, of course.

Back then I had to buy one of these HDMI dummy plugs (that I still have laying around) to keep the GPU active as when a graphics card doesn’t have an active display connected to it, it tends to function with less capacity and overall worse, but also because if I had no display, I had no display to stream remotely either xD.

Short story short, it got slowly upgraded overtime with spare parts of people around me that they gifted because they didn’t need anymore. It got 24GB of RAM, better CPU, etc. All until my dad on one of our scavenging hunts in the trash accidentally found a crazy gaming PC that was completely functional and was better, of course, so I started to use that one which included an NVIDIA GTX 1060 6GB. But I also used yet another 1060 from a friend that wasn’t using, and fused the RAM that came with the trash PC with the one I already had, making for 64GB.

Buuut even that isn’t used anymore, as when I got a better financial situation and eventually better jobs, I upgraded the motherboard, changed its CPU 2 times, and replaced the 1060s with two 1080 Ti’s of 11GB each. Storage had been upgraded too in all this time to 7TB. I’d say around the time my dad found the PC and i got more pieces from friends is when the true concept of OmniCloud came into existence.

It was no longer just a machine for silly videogames, oh no. It was now a full fledged Proxmox Virtual Environment server, with lots of resources. The most power I’ve ever had, really.

Andddd….. for what? What are you gonna do with all that power? Are you going to nuke my house? O_o

No, of course not! 😛 All that power goes to virtualization, obviously. I won’t go into much detail about virtual machines, Proxmox itself or KVM/QEMU but the point of the server is to host a lot of VMs running at once.

Now that you know what OmniCloud  is supposed to be, let’s take a closer look at it shall we~?

Ain’t she pretty~? :3

Well don’t eyeball it too much, it’s mine >:D

Yup, indeed. It’s final and current specs as of today are those. Pretty impressive, right? And to think this just started as a cheap office computer with a 1050 Ti slapped on it… We really have come far, but if you thought this is all there is to it, you clearly don’t know me yet~

Apart from the impressive hardware, this is not all I have in store for you. Let’s take a look at the um… software tweaks I may or may not have hacked in.

For starters, those of you already familiar with virtualization already know that most things can detect the presence of a VM when scanning for certain things in the system. Some malware for example will refuse to run upon discovering it’s running on a virtual machine, because more often than not, it can mean it’s being analyzed in a safe and prepared sandbox. any.run is a good example of this. Well, one of the two hacks I implemented on OmniCloud is precisely that:

Manually patching the hypervisor so that no software inside a VM knows it’s running in a VM. Though, of course, there’s some exceptions to this like a heavy kernel-level rootkit or anti-cheat for a game (yes, Vanguard, I’m talking about you). But anything else including regular anti-cheats for any game like the most widely used (EasyAntiCheat, BattlEye, etc) all work fine inside my VMs, meaning I can play basically any game I want to play already. Or, if I want to do malware analysis, I pretty easily can! I bypass most of the checks really, but there’s some which are really difficult to impersonate real hardware properly.

Second thing would be that GT 710 that you’ve seen there too. Its purpose is more simple but important nonetheless as well. For those of you familiar with hackintoshing, that GPU over there is just there to provide smooth graphics to any macOS VMs I create as well. It makes my Cloud more Omni as it allows it to do more stuff, and Omni means being able to do anything and everything, so it’s important to me to have a GPU with such a high compatibility with new macOS versions for hackintoshing.

The third thing and probably the coolest one I implemented is a hack for vGPU. Meaning, I hacked the vGPU NVIDIA driver for Linux to think my consumer-grade cards are essentially capable of vGPU (which, they are, just that NVIDIA forcibly locks it down for you so that you can’t use it and you need to buy an actual enterprise card + a license for vGPU… you can kinda get an idea on how expensive all that would be). After that’s done, I also have my custom virtual GPU profiles as the way it works, you can only split the VRAM equally to each vGPU. Like, you can’t assign 2GB to one, 1.34 to another and 5 to the last one, it has to be equal splits, so if you decide splitting by 2GB, ALL vGPUs get 2GB until you can’t create any more virtual ones. So, if you had a 6B GPU and you decided the split is 2GB, you’d get 3 vGPUs… easy to understand, right? :3

As an example, take a look:

All those vGPUs later then can be passed through to your VM as if they were regular graphics cards, making a single physical GPU able to be used by multiple virtual machines, all sharing its resources at once and allowing for hardware acceleration. In short and dumb: me have graphics for all VMs, VMs run smooth and can do heavy task like game, AI, 3D, etc; me happy :3

The alternative to that for those of you who don’t know much about any of this, is I’d have to pass through the whole graphics card to a single VM, meaning I can’t split its resources to more than one, and since I have only 3 GPUs I’d only be able to run 3 VMs, but I haven’t given a number on how many I can run right now.

Simple overview: the 2 GPUs that I can split are the 2 1080 Tis, as the GT 710 is unsupported and nonetheless has a different role on OmniCloud anyways. That means we have 22GB total of VRAM we can split. If we split each one by portions of just 1GB we’d get 22 VMs running with GPU hardware acceleration at once, offloading all the heavy graphics tasks of just running the VMs to the GPUs and not the poor CPU which already has enough on its plate… but we can go further, as we’re not limited by GBs~

We can also split each one by 512MB portions, making a whopping total of 44 VMs!! :O

This means I have enough power to run the practical limit of VMs I can even run on this CPU, as if I really wanted to stretch it, I’d just assign 1 single CPU thread to each VM to a total of 36 threads I have, 36 VMs; because if I start running more, they start overlapping resources, something I really want to avoid. But I can also split by 5.5GB (half of each GPU) and have 4 gaming VMs ready to play remotely at any time (more on that cool setup in a future blog post ;P), running an AI LLM, or multiple of them, etc etc… but it’s true use is for giving graphics to a lot of machines for a lot of big VM labs I plan to setup to learn from them, hack them, reconfigure them and of course post them in this blog! This is a lab after all, it’s meant as my playground to learn, hack, game… anything I want.

In fact, those of you with keen intuition might have already guessed that The Green Pixel is also running as a VM inside OmniCloud, and if you didn’t, well now you know >:3 (indeed, this is only a fraction of my power, mwahahaha! >:D)

I hope this post gave you a better overview on the types of things I do with tech, but also how I usually run my power-hungry things and what OmniCloud is all about which will extend to this blog, of course. Anything I talked about in here is worthy of its own long-rambling nerdy post but as you can understand I can’t fit all of that in here. So, if any of what I talked about sparks your interest… tell me so! If you want me to talk about virtualization, hackintoshing, vGPU or my cool 4 gaming VMs setup just let me know as a comment. Any feedback is appreciated and I’ll get to better know what you want to see (though the gaming setup I already will confirm to you that it’ll be its own separate post, and a full guide at that, so stay tuned).

See you in the next one! ;3