One Server vs Cluster: What Your Homelab Actually Needs

Hey guys, I've got a question for you. Hey guys, I've got a question for you. Have you ever finished a big home lab Have you ever finished a big home lab Have you ever finished a big home lab project where you invested a lot of time project where you invested a lot of time project where you invested a lot of time and money to get everything working and and money to get everything working and and money to get everything working and then suddenly realized, well, maybe I then suddenly realized, well, maybe I then suddenly realized, well, maybe I just over complicated everything and the just over complicated everything and the just over complicated everything and the whole project wasn't really worth it. whole project wasn't really worth it. whole project wasn't really worth it. Honestly, that's exactly the feeling Honestly, that's exactly the feeling Honestly, that's exactly the feeling that I had after I finished my FreeNAS that I had after I finished my FreeNAS that I had after I finished my FreeNAS Proxmox cluster project because at Proxmox cluster project because at Proxmox cluster project because at first, it seemed like a really cool way first, it seemed like a really cool way first, it seemed like a really cool way to increase the power of my entire home to increase the power of my entire home to increase the power of my entire home lab, being able to spread workloads lab, being able to spread workloads lab, being able to spread workloads across multiple smaller, more power across multiple smaller, more power across multiple smaller, more power efficient machines. But honestly, it efficient machines. But honestly, it efficient machines. But honestly, it also introduced many new challenges also introduced many new challenges also introduced many new challenges because I needed to buy three new hard because I needed to buy three new hard because I needed to buy three new hard drives, more memory for my servers, one drives, more memory for my servers, one drives, more memory for my servers, one entirely new server, get fast networking entirely new server, get fast networking entirely new server, get fast networking up and running, and of course, I now up and running, and of course, I now up and running, and of course, I now have many more things that I need to have many more things that I need to have many more things that I need to maintain and troubleshoot all the time. maintain and troubleshoot all the time. maintain and troubleshoot all the time. And when I recently just bought a brand And when I recently just bought a brand And when I recently just bought a brand new server, the HP ML350, new server, the HP ML350, new server, the HP ML350, originally just because I wanted to originally just because I wanted to originally just because I wanted to experiment with enterprise hardware, experiment with enterprise hardware, experiment with enterprise hardware, that got me thinking, well, this one that got me thinking, well, this one that got me thinking, well, this one server has so much room for expansion, server has so much room for expansion, server has so much room for expansion, wouldn't it have been much easier to put wouldn't it have been much easier to put wouldn't it have been much easier to put my entire home lab on this one huge my entire home lab on this one huge my entire home lab on this one huge server instead of building a complex server instead of building a complex server instead of building a complex cluster architecture? Of course, that cluster architecture? Of course, that cluster architecture? Of course, that would make many things easier, and in my would make many things easier, and in my would make many things easier, and in my opinion, it leads to an interesting opinion, it leads to an interesting opinion, it leads to an interesting question about what strategy should you question about what strategy should you question about what strategy should you actually follow in your home lab. So, do actually follow in your home lab. So, do actually follow in your home lab. So, do you keep expanding on one capable server you keep expanding on one capable server you keep expanding on one capable server or do you add more physical machines and or do you add more physical machines and or do you add more physical machines and build storage, networking, and high build storage, networking, and high build storage, networking, and high availability around them? Ideally, you availability around them? Ideally, you availability around them? Ideally, you should think this through before making should think this through before making should think this through before making an investment into the wrong equipment.

an investment into the wrong equipment. an investment into the wrong equipment. Though, I'm not saying that there is one Though, I'm not saying that there is one Though, I'm not saying that there is one correct answer because I know every home correct answer because I know every home correct answer because I know every home lab is different and changes all the lab is different and changes all the lab is different and changes all the time. There are probably a lot of pros time. There are probably a lot of pros time. There are probably a lot of pros and cons for the one or another and cons for the one or another and cons for the one or another decision. So, I'd like to make this decision. So, I'd like to make this decision. So, I'd like to make this video not as a tutorial, but as an open video not as a tutorial, but as an open video not as a tutorial, but as an open discussion with you guys and talk about discussion with you guys and talk about discussion with you guys and talk about the pros and cons of building your home the pros and cons of building your home the pros and cons of building your home lab on one giant server versus building lab on one giant server versus building lab on one giant server versus building a cluster. So, I hope you're up for this a cluster. So, I hope you're up for this a cluster. So, I hope you're up for this discussion. Before we continue, I discussion. Before we continue, I discussion. Before we continue, I quickly want to say thanks to AMD for quickly want to say thanks to AMD for quickly want to say thanks to AMD for sponsoring this video. AMD not only sponsoring this video. AMD not only sponsoring this video. AMD not only builds the fastest desktop and efficient builds the fastest desktop and efficient builds the fastest desktop and efficient mobile CPUs, they also build AMD epic mobile CPUs, they also build AMD epic mobile CPUs, they also build AMD epic processors, which are enterprise-grade processors, which are enterprise-grade processors, which are enterprise-grade server CPUs for serious infrastructure server CPUs for serious infrastructure server CPUs for serious infrastructure workloads. And I already talked about workloads. And I already talked about workloads. And I already talked about this in previous videos. Enterprise gear this in previous videos. Enterprise gear this in previous videos. Enterprise gear can matter even in a homelab because you can matter even in a homelab because you can matter even in a homelab because you get access to hardware features that you get access to hardware features that you get access to hardware features that you usually don't get or only get in a very usually don't get or only get in a very usually don't get or only get in a very limited way on normal consumer limited way on normal consumer limited way on normal consumer platforms. With AMD epic, you get lots platforms. With AMD epic, you get lots platforms. With AMD epic, you get lots of CPU cores, huge memory capacity, and of CPU cores, huge memory capacity, and of CPU cores, huge memory capacity, and high bandwidth, and many PCI Express high bandwidth, and many PCI Express high bandwidth, and many PCI Express lanes for expansion cards like fast lanes for expansion cards like fast lanes for expansion cards like fast networking, storage controllers, NVMe networking, storage controllers, NVMe networking, storage controllers, NVMe GPUs, and so on. That gives you much GPUs, and so on. That gives you much GPUs, and so on. That gives you much more power to experiment in your more power to experiment in your more power to experiment in your homelab. And for enterprise companies, homelab. And for enterprise companies, homelab. And for enterprise companies, that of course matters even more because that of course matters even more because that of course matters even more because you get consolidate more virtual you get consolidate more virtual you get consolidate more virtual machines on one host, build machines on one host, build machines on one host, build storage-heavy systems, run databases, storage-heavy systems, run databases, storage-heavy systems, run databases, and create server platforms that still and create server platforms that still and create server platforms that still have a lot of room for both cloud and have a lot of room for both cloud and have a lot of room for both cloud and on-premise workloads. And for AI-powered on-premise workloads. And for AI-powered on-premise workloads. And for AI-powered systems, the CPU also matters because it systems, the CPU also matters because it systems, the CPU also matters because it handles data preparation,

handles data preparation, handles data preparation, post-processing, storage, networking, post-processing, storage, networking, post-processing, storage, networking, and host-side work. And depending on the and host-side work. And depending on the and host-side work. And depending on the platform and configuration, AMD epic platform and configuration, AMD epic platform and configuration, AMD epic also includes AMD Infinity Guard also includes AMD Infinity Guard also includes AMD Infinity Guard security features, which are designed to security features, which are designed to security features, which are designed to help protect workloads in virtualized help protect workloads in virtualized help protect workloads in virtualized and cloud environments. So, with AMD and cloud environments. So, with AMD and cloud environments. So, with AMD epic, you get a powerful x86 server epic, you get a powerful x86 server epic, you get a powerful x86 server platform that is fast, efficient, platform that is fast, efficient, platform that is fast, efficient, expandable, and compatible with the expandable, and compatible with the expandable, and compatible with the normal infrastructure stack most of us normal infrastructure stack most of us normal infrastructure stack most of us already use, Linux, databases, already use, Linux, databases, already use, Linux, databases, cloud-native workloads, and all the cloud-native workloads, and all the cloud-native workloads, and all the usual tools. In my opinion, this is an usual tools. In my opinion, this is an usual tools. In my opinion, this is an amazing platform for both serious amazing platform for both serious amazing platform for both serious homelab and enterprise companies. So, if homelab and enterprise companies. So, if homelab and enterprise companies. So, if you want to check it out, then use the you want to check it out, then use the you want to check it out, then use the AMD epic link in the description of this AMD epic link in the description of this AMD epic link in the description of this video. video. video. All right, guys. So, now before we jump All right, guys. So, now before we jump All right, guys. So, now before we jump into the details of these two setups, into the details of these two setups, into the details of these two setups, let's take one step back and look how let's take one step back and look how let's take one step back and look how this decision usually comes up in the this decision usually comes up in the this decision usually comes up in the first place because homelabs don't first place because homelabs don't first place because homelabs don't usually start with a complete usually start with a complete usually start with a complete architectural plan. At least I don't architectural plan. At least I don't architectural plan. At least I don't know a single homelab that does it like know a single homelab that does it like know a single homelab that does it like that because homelabs usually start that because homelabs usually start that because homelabs usually start small. Maybe with an old PC or perhaps small. Maybe with an old PC or perhaps small. Maybe with an old PC or perhaps with a small server. You just install with a small server. You just install with a small server. You just install Linux, a hypervisor, maybe Proxmox and Linux, a hypervisor, maybe Proxmox and Linux, a hypervisor, maybe Proxmox and then start self-hosting a few then start self-hosting a few then start self-hosting a few applications and you just have fun applications and you just have fun applications and you just have fun experimenting with all of this. At that experimenting with all of this. At that experimenting with all of this. At that point you probably don't even think point you probably don't even think point you probably don't even think about building a cluster or running a about building a cluster or running a about building a cluster or running a complicated infrastructure at home complicated infrastructure at home complicated infrastructure at home because want to know how things are because want to know how things are because want to know how things are running and learn how they work. But running and learn how they work. But running and learn how they work. But trust me guys, if you enjoy trust me guys, if you enjoy trust me guys, if you enjoy experimenting, you just keep adding new experimenting, you just keep adding new experimenting, you just keep adding new things over time and your homelab will things over time and your homelab will things over time and your homelab will naturally grow to a point where you want naturally grow to a point where you want naturally grow to a point where you want to just start learning more about

to just start learning more about to just start learning more about certain technologies and setups. And certain technologies and setups. And certain technologies and setups. And this is where I think the architecture this is where I think the architecture this is where I think the architecture of your homelab suddenly starts to of your homelab suddenly starts to of your homelab suddenly starts to matter more and more and you will ask matter more and more and you will ask matter more and more and you will ask yourself, do you continue building yourself, do you continue building yourself, do you continue building around the one server that you already around the one server that you already around the one server that you already have or is it time to add more machines? have or is it time to add more machines? have or is it time to add more machines? Let's be honest, the easiest approach is Let's be honest, the easiest approach is Let's be honest, the easiest approach is just by upgrading the server you already just by upgrading the server you already just by upgrading the server you already have. So you can add more memory, have. So you can add more memory, have. So you can add more memory, install more or larger disks, maybe install more or larger disks, maybe install more or larger disks, maybe replace the CPU with a faster one or add replace the CPU with a faster one or add replace the CPU with a faster one or add expansion cards like fast networking, expansion cards like fast networking, expansion cards like fast networking, storage controllers or when this is not storage controllers or when this is not storage controllers or when this is not possible because you have reached a hard possible because you have reached a hard possible because you have reached a hard limit of your server platform, then you limit of your server platform, then you limit of your server platform, then you just sell your old homelab server and just sell your old homelab server and just sell your old homelab server and just buy a more powerful one. That's by just buy a more powerful one. That's by just buy a more powerful one. That's by the way what I've done many, many times the way what I've done many, many times the way what I've done many, many times as well in my homelab journey and in as well in my homelab journey and in as well in my homelab journey and in professional IT, this is what we also professional IT, this is what we also professional IT, this is what we also call vertical scaling or scaling up, so call vertical scaling or scaling up, so call vertical scaling or scaling up, so making one physical machine larger. So making one physical machine larger. So making one physical machine larger. So this is a very common concept. The this is a very common concept. The this is a very common concept. The opposite approach is horizontal scaling opposite approach is horizontal scaling opposite approach is horizontal scaling or scaling out, so that means you add or scaling out, so that means you add or scaling out, so that means you add more physical machines of a similar kind more physical machines of a similar kind more physical machines of a similar kind and start distributing all the workloads and start distributing all the workloads and start distributing all the workloads across them. However, it is important to across them. However, it is important to across them. However, it is important to say that this does not automatically say that this does not automatically say that this does not automatically give you a useful cluster or high give you a useful cluster or high give you a useful cluster or high availability. Of course, there's a a availability. Of course, there's a a availability. Of course, there's a a more we have to do and we'll get into more we have to do and we'll get into more we have to do and we'll get into this later, but that is also the reason this later, but that is also the reason this later, but that is also the reason why I think scaling up is often the why I think scaling up is often the why I think scaling up is often the easiest place to start because it keeps easiest place to start because it keeps easiest place to start because it keeps the architecture much more the architecture much more the architecture much more straightforward. Of course, every server straightforward. Of course, every server straightforward. Of course, every server has a hard limit, a fixed number of has a hard limit, a fixed number of has a hard limit, a fixed number of memory slots, drive bays, PCI lanes, and memory slots, drive bays, PCI lanes, and memory slots, drive bays, PCI lanes, and so on. That's why in my opinion, a so on. That's why in my opinion, a so on. That's why in my opinion, a refurbished enterprise server such as

refurbished enterprise server such as refurbished enterprise server such as the HPE ML350 the HPE ML350 the HPE ML350 Generation 10 that I'm currently testing Generation 10 that I'm currently testing Generation 10 that I'm currently testing in my home lab or a similar one, maybe in my home lab or a similar one, maybe in my home lab or a similar one, maybe with an AMD epic CPU, that is a great with an AMD epic CPU, that is a great with an AMD epic CPU, that is a great choice to run a home lab because a choice to run a home lab because a choice to run a home lab because a powerful enterprise server platform powerful enterprise server platform powerful enterprise server platform gives you a huge upgrade potential. You gives you a huge upgrade potential. You gives you a huge upgrade potential. You have many more memory slots, more drive have many more memory slots, more drive have many more memory slots, more drive bays, more PCI lanes, and expansion bays, more PCI lanes, and expansion bays, more PCI lanes, and expansion slots than on normal consumer platforms. slots than on normal consumer platforms. slots than on normal consumer platforms. You mostly even get a secondary CPU You mostly even get a secondary CPU You mostly even get a secondary CPU socket. But also, if you don't want to socket. But also, if you don't want to socket. But also, if you don't want to use enterprise servers in your home lab use enterprise servers in your home lab use enterprise servers in your home lab because maybe they're too noisy or too because maybe they're too noisy or too because maybe they're too noisy or too big, you can still use desktop PCs. That big, you can still use desktop PCs. That big, you can still use desktop PCs. That can also be a viable option. For can also be a viable option. For can also be a viable option. For example, I also used desktop PCs a lot example, I also used desktop PCs a lot example, I also used desktop PCs a lot in my home lab in the past. I usually in my home lab in the past. I usually in my home lab in the past. I usually put them in a rack server case, but this put them in a rack server case, but this put them in a rack server case, but this was still a normal consumer platform. was still a normal consumer platform. was still a normal consumer platform. And I think this offers a great balance And I think this offers a great balance And I think this offers a great balance between a lower initial investment, more between a lower initial investment, more between a lower initial investment, more power efficiency, and still having some power efficiency, and still having some power efficiency, and still having some room for upgrades such as adding more or room for upgrades such as adding more or room for upgrades such as adding more or bigger memory modules, add more storage bigger memory modules, add more storage bigger memory modules, add more storage drives, or additional expansion cards. drives, or additional expansion cards. drives, or additional expansion cards. And I think increasing the resources of And I think increasing the resources of And I think increasing the resources of one host in your home lab infrastructure one host in your home lab infrastructure one host in your home lab infrastructure has one huge advantage over multiple has one huge advantage over multiple has one huge advantage over multiple machines. Because while a cluster may machines. Because while a cluster may machines. Because while a cluster may have more CPU and memory in total, a have more CPU and memory in total, a have more CPU and memory in total, a single workload, like a virtual machine single workload, like a virtual machine single workload, like a virtual machine for example, normally cannot combine the for example, normally cannot combine the for example, normally cannot combine the resources from several nodes. So, it resources from several nodes. So, it resources from several nodes. So, it still has to fit in one of them. And if still has to fit in one of them. And if still has to fit in one of them. And if the workload needs a lot of CPU and the workload needs a lot of CPU and the workload needs a lot of CPU and memory consolidated in one place, then memory consolidated in one place, then memory consolidated in one place, then one big server can actually be much one big server can actually be much one big server can actually be much easier. By the way, I think this also easier. By the way, I think this also easier. By the way, I think this also doesn't mean that you give up learning

doesn't mean that you give up learning doesn't mean that you give up learning and experimentation potential if you're and experimentation potential if you're and experimentation potential if you're building your home lab around a single building your home lab around a single building your home lab around a single server. That is an argument I hear all server. That is an argument I hear all server. That is an argument I hear all the time, but I believe you can still the time, but I believe you can still the time, but I believe you can still use virtualization and containerization use virtualization and containerization use virtualization and containerization to build many cool and exciting projects to build many cool and exciting projects to build many cool and exciting projects and setups. Just take a look at my and setups. Just take a look at my and setups. Just take a look at my Docker and Kubernetes videos that I have Docker and Kubernetes videos that I have Docker and Kubernetes videos that I have built completely on virtual machines. built completely on virtual machines. built completely on virtual machines. Or, if you're into networking, you can Or, if you're into networking, you can Or, if you're into networking, you can make your hypervisor VLAN aware and make your hypervisor VLAN aware and make your hypervisor VLAN aware and experiment with network segmentation, experiment with network segmentation, experiment with network segmentation, building complete network structures by building complete network structures by building complete network structures by just using your virtual machines, or just using your virtual machines, or just using your virtual machines, or install a firewall like Open Sense on a install a firewall like Open Sense on a install a firewall like Open Sense on a VM. There are many, many technologies VM. There are many, many technologies VM. There are many, many technologies and scenarios you can actually build and and scenarios you can actually build and and scenarios you can actually build and get hands-on experience by just running get hands-on experience by just running get hands-on experience by just running them virtualized on a single server. And them virtualized on a single server. And them virtualized on a single server. And you can even make parts of this setup you can even make parts of this setup you can even make parts of this setup fault tolerant. A storage is a great fault tolerant. A storage is a great fault tolerant. A storage is a great example. Like you can combine multiple example. Like you can combine multiple example. Like you can combine multiple drives into a hardware or software RAID, drives into a hardware or software RAID, drives into a hardware or software RAID, a separate storage server as a virtual a separate storage server as a virtual a separate storage server as a virtual machine, which I've, by the way, done machine, which I've, by the way, done machine, which I've, by the way, done multiple times and that always works multiple times and that always works multiple times and that always works great without any problems. And if great without any problems. And if great without any problems. And if you're using an enterprise server you're using an enterprise server you're using an enterprise server platform, you usually also have ECC platform, you usually also have ECC platform, you usually also have ECC memory, redundant power supplies, and memory, redundant power supplies, and memory, redundant power supplies, and multiple network interfaces all in one multiple network interfaces all in one multiple network interfaces all in one place. So, as you can see, a single place. So, as you can see, a single place. So, as you can see, a single server design is not automatically a server design is not automatically a server design is not automatically a cheap or limited homelab. You can really cheap or limited homelab. You can really cheap or limited homelab. You can really build almost everything on a single build almost everything on a single build almost everything on a single machine, and it can be a smart choice machine, and it can be a smart choice machine, and it can be a smart choice because it gives you one straightforward because it gives you one straightforward because it gives you one straightforward platform for all of your projects you platform for all of your projects you platform for all of your projects you actually care about, like running actually care about, like running actually care about, like running self-hosted services and building self-hosted services and building self-hosted services and building learning setups with virtual machines, learning setups with virtual machines, learning setups with virtual machines, containers, and virtual networks. So, containers, and virtual networks. So, containers, and virtual networks. So, you can actually spend more time on you can actually spend more time on you can actually spend more time on those interesting projects and less time

those interesting projects and less time those interesting projects and less time just keeping the homelab infrastructure just keeping the homelab infrastructure just keeping the homelab infrastructure itself running. And if you'd ask me now, itself running. And if you'd ask me now, itself running. And if you'd ask me now, after doing all of these huge projects after doing all of these huge projects after doing all of these huge projects in my homelab, like building the Proxmox in my homelab, like building the Proxmox in my homelab, like building the Proxmox cluster, getting three new drives, a new cluster, getting three new drives, a new cluster, getting three new drives, a new managed switch with 10 gigabits, or managed switch with 10 gigabits, or managed switch with 10 gigabits, or building a separate storage server, I building a separate storage server, I building a separate storage server, I sometimes wonder if I really need all of sometimes wonder if I really need all of sometimes wonder if I really need all of this. Maybe I could have saved myself a this. Maybe I could have saved myself a this. Maybe I could have saved myself a lot of money and headache by just lot of money and headache by just lot of money and headache by just building my homelab around one big building my homelab around one big building my homelab around one big server from the start. But there are server from the start. But there are server from the start. But there are still a few small things that I don't still a few small things that I don't still a few small things that I don't regret because here's the thing, it regret because here's the thing, it regret because here's the thing, it would have still been just one server. would have still been just one server. would have still been just one server. So, if anything happens to that machine, So, if anything happens to that machine, So, if anything happens to that machine, perhaps a hardware failure or simply a perhaps a hardware failure or simply a perhaps a hardware failure or simply a required reboot that you need to do, required reboot that you need to do, required reboot that you need to do, then every workload running on this then every workload running on this then every workload running on this machine is affected at the same time. In machine is affected at the same time. In machine is affected at the same time. In professional IT, you could also describe professional IT, you could also describe professional IT, you could also describe this as a single point of failure. And this as a single point of failure. And this as a single point of failure. And of course, just like I said, you can of course, just like I said, you can of course, just like I said, you can have many redundancies inside a single have many redundancies inside a single have many redundancies inside a single server as well, like a virtual backup server as well, like a virtual backup server as well, like a virtual backup server, redundant workloads on server, redundant workloads on server, redundant workloads on containers, but everything still shares containers, but everything still shares containers, but everything still shares the same motherboard, power supply, the same motherboard, power supply, the same motherboard, power supply, storage hardware, and hypervisor.

storage hardware, and hypervisor. storage hardware, and hypervisor. However, I also want to be fair here. We However, I also want to be fair here. We However, I also want to be fair here. We are still talking about a home lab, not are still talking about a home lab, not are still talking about a home lab, not about a company production environment, about a company production environment, about a company production environment, at least in my case. So, in my opinion, at least in my case. So, in my opinion, at least in my case. So, in my opinion, when people talk about high availability when people talk about high availability when people talk about high availability in home labs, I think you do not need to in home labs, I think you do not need to in home labs, I think you do not need to solve an availability problem you solve an availability problem you solve an availability problem you actually do not have. That means if you actually do not have. That means if you actually do not have. That means if you plan a reboot on a Sunday or your server plan a reboot on a Sunday or your server plan a reboot on a Sunday or your server is unavailable for a few hours, in a is unavailable for a few hours, in a is unavailable for a few hours, in a home lab that might be completely fine. home lab that might be completely fine. home lab that might be completely fine. For example, in my home lab, nothing For example, in my home lab, nothing For example, in my home lab, nothing really needs 24/7 availability, let's be really needs 24/7 availability, let's be really needs 24/7 availability, let's be honest. But of course, it can still be a honest. But of course, it can still be a honest. But of course, it can still be a lot of headache if you have to repair or lot of headache if you have to repair or lot of headache if you have to repair or replace components when everything is replace components when everything is replace components when everything is running on a single server, or if you running on a single server, or if you running on a single server, or if you need to do an update and reboot your need to do an update and reboot your need to do an update and reboot your host operating system, then you also host operating system, then you also host operating system, then you also need to reboot all of your virtual need to reboot all of your virtual need to reboot all of your virtual machines and restart your containers. machines and restart your containers. machines and restart your containers. That's of course always a good argument That's of course always a good argument That's of course always a good argument for building a home lab cluster. So, for building a home lab cluster. So, for building a home lab cluster. So, instead of running everything on one instead of running everything on one instead of running everything on one single machine, scaling out or do single machine, scaling out or do single machine, scaling out or do horizontal scaling, but here's the horizontal scaling, but here's the horizontal scaling, but here's the point, the reason why I built my Proxmox point, the reason why I built my Proxmox point, the reason why I built my Proxmox cluster in the first place was not cluster in the first place was not cluster in the first place was not because I necessarily needed high because I necessarily needed high because I necessarily needed high availability. As I said, it is still availability. As I said, it is still availability. As I said, it is still home lab. I mainly built it because I home lab. I mainly built it because I home lab. I mainly built it because I wanted separate physical nodes and get wanted separate physical nodes and get wanted separate physical nodes and get hands-on experience with technologies hands-on experience with technologies hands-on experience with technologies like quorum, shared storage, and doing like quorum, shared storage, and doing like quorum, shared storage, and doing live migration. Because you can live migration. Because you can live migration. Because you can virtualize many learning and virtualize many learning and virtualize many learning and experimental setups on one server, but experimental setups on one server, but experimental setups on one server, but underneath they still share the same underneath they still share the same underneath they still share the same physical hardware, and you're not physical hardware, and you're not physical hardware, and you're not actually seeing the full picture or not actually seeing the full picture or not actually seeing the full picture or not running in the same challenges and running in the same challenges and running in the same challenges and problems as with a cluster. This is problems as with a cluster. This is problems as with a cluster. This is really the key difference. With building

really the key difference. With building really the key difference. With building a home lab cluster, maintaining the a home lab cluster, maintaining the a home lab cluster, maintaining the infrastructure itself becomes a project. infrastructure itself becomes a project. infrastructure itself becomes a project. So, you need to set up Proxmox HA, maybe So, you need to set up Proxmox HA, maybe So, you need to set up Proxmox HA, maybe set up shared storage with Ceph, doing set up shared storage with Ceph, doing set up shared storage with Ceph, doing migrations, monitoring, and dealing with migrations, monitoring, and dealing with migrations, monitoring, and dealing with all of the problems and failures. And all of the problems and failures. And all of the problems and failures. And also the reason why I recently rebuilt also the reason why I recently rebuilt also the reason why I recently rebuilt my original two-node Proxmox cluster. my original two-node Proxmox cluster. my original two-node Proxmox cluster. For those of you who have not followed For those of you who have not followed For those of you who have not followed this project, I originally had a small Q this project, I originally had a small Q this project, I originally had a small Q device as a third vote machine on a device as a third vote machine on a device as a third vote machine on a small Zimaboard. So, in my Proxmox small Zimaboard. So, in my Proxmox small Zimaboard. So, in my Proxmox cluster, I just needed two physical cluster, I just needed two physical cluster, I just needed two physical servers. But because I wanted to learn servers. But because I wanted to learn servers. But because I wanted to learn how to run and maintain a real how to run and maintain a real how to run and maintain a real three-node Proxmox cluster, I removed it three-node Proxmox cluster, I removed it three-node Proxmox cluster, I removed it and now I have three independent and now I have three independent and now I have three independent physical servers that allowed me to physical servers that allowed me to physical servers that allowed me to build a Ceph storage cluster as well. build a Ceph storage cluster as well. build a Ceph storage cluster as well. And this was honestly such a cool and And this was honestly such a cool and And this was honestly such a cool and exciting project because Ceph works exciting project because Ceph works exciting project because Ceph works completely different from maintaining completely different from maintaining completely different from maintaining local storage I had used before. Because local storage I had used before. Because local storage I had used before. Because instead of keeping the complete VM disk instead of keeping the complete VM disk instead of keeping the complete VM disk on one single drive, Ceph breaks the on one single drive, Ceph breaks the on one single drive, Ceph breaks the data into objects and distributes the data into objects and distributes the data into objects and distributes the replicas across the cluster. So, in my replicas across the cluster. So, in my replicas across the cluster. So, in my setup, each NVMe drive is one OSD, and setup, each NVMe drive is one OSD, and setup, each NVMe drive is one OSD, and Ceph tries to keep three replicas of Ceph tries to keep three replicas of Ceph tries to keep three replicas of every object across the three physical every object across the three physical every object across the three physical nodes. And this is where the biggest nodes. And this is where the biggest nodes. And this is where the biggest advantage of running a cluster becomes advantage of running a cluster becomes advantage of running a cluster becomes visible. Because now I have multiple visible. Because now I have multiple visible. Because now I have multiple points of failure, and the VM disk is no points of failure, and the VM disk is no points of failure, and the VM disk is no longer tied to one single machine. I can longer tied to one single machine. I can longer tied to one single machine. I can easily do live migration, move a running easily do live migration, move a running easily do live migration, move a running VM from one node to another. Proxmox VM from one node to another. Proxmox VM from one node to another. Proxmox only needs to transfer the active memory

only needs to transfer the active memory only needs to transfer the active memory and the runtime state, and a sudden node and the runtime state, and a sudden node and the runtime state, and a sudden node failure works differently as well. Like failure works differently as well. Like failure works differently as well. Like if anything happens to one of my home if anything happens to one of my home if anything happens to one of my home lab servers, maybe an internal NVMe is lab servers, maybe an internal NVMe is lab servers, maybe an internal NVMe is gone or another hardware failure occurs, gone or another hardware failure occurs, gone or another hardware failure occurs, the virtual machines on that machine the virtual machines on that machine the virtual machines on that machine would stop working, yes, but if the would stop working, yes, but if the would stop working, yes, but if the cluster still has quorum, so the cluster still has quorum, so the cluster still has quorum, so the majority of servers are still online and majority of servers are still online and majority of servers are still online and available, Proxmox can restart the most available, Proxmox can restart the most available, Proxmox can restart the most critical virtual machines on another one critical virtual machines on another one critical virtual machines on another one of these nodes. And my homelab workloads of these nodes. And my homelab workloads of these nodes. And my homelab workloads can keep running completely while I can keep running completely while I can keep running completely while I recover or replace a failed server or recover or replace a failed server or recover or replace a failed server or rebooting and updating, and I do not rebooting and updating, and I do not rebooting and updating, and I do not even have to restore anything from a even have to restore anything from a even have to restore anything from a backup. In my opinion, this is a huge backup. In my opinion, this is a huge backup. In my opinion, this is a huge plus. So, the cluster is not only an plus. So, the cluster is not only an plus. So, the cluster is not only an amazing learning project where I can get amazing learning project where I can get amazing learning project where I can get hands-on experience, it also gives you hands-on experience, it also gives you hands-on experience, it also gives you real operational value because you can real operational value because you can real operational value because you can maintain or even replace one node maintain or even replace one node maintain or even replace one node without taking the complete platform without taking the complete platform without taking the complete platform offline. So, even if I eventually decide offline. So, even if I eventually decide offline. So, even if I eventually decide to shut down the Proxmox cluster and put to shut down the Proxmox cluster and put to shut down the Proxmox cluster and put everything on one giant server, which everything on one giant server, which everything on one giant server, which might be the most straightforward option might be the most straightforward option might be the most straightforward option for a homelab, I think building the for a homelab, I think building the for a homelab, I think building the cluster was not a complete waste of cluster was not a complete waste of cluster was not a complete waste of time. And besides the learning time. And besides the learning time. And besides the learning experience and availability benefits, a experience and availability benefits, a experience and availability benefits, a cluster also gives you a very flexible cluster also gives you a very flexible cluster also gives you a very flexible way to grow your homelab. Like if you way to grow your homelab. Like if you way to grow your homelab. Like if you need more compute or storage later, need more compute or storage later, need more compute or storage later, where you would hit a hard limit with where you would hit a hard limit with where you would hit a hard limit with one physical server, instead with a one physical server, instead with a one physical server, instead with a cluster, you do not have to replace or cluster, you do not have to replace or cluster, you do not have to replace or upgrade anything. You can just add more upgrade anything. You can just add more upgrade anything. You can just add more nodes and expand your homelab nodes and expand your homelab nodes and expand your homelab step-by-step. This is, by the way, also step-by-step. This is, by the way, also step-by-step. This is, by the way, also the reason why I use small mini PCs for

the reason why I use small mini PCs for the reason why I use small mini PCs for my Proxmox cluster because I think they my Proxmox cluster because I think they my Proxmox cluster because I think they make a lot of sense for this type of make a lot of sense for this type of make a lot of sense for this type of setup. These systems can give you a lot setup. These systems can give you a lot setup. These systems can give you a lot of compute performance while using of compute performance while using of compute performance while using relatively low power. So, even if you're relatively low power. So, even if you're relatively low power. So, even if you're a beginner and starting your homelab, a beginner and starting your homelab, a beginner and starting your homelab, you can just start small and efficient you can just start small and efficient you can just start small and efficient with just a two-node Proxmox cluster, with just a two-node Proxmox cluster, with just a two-node Proxmox cluster, run a smaller Q device somewhere else, run a smaller Q device somewhere else, run a smaller Q device somewhere else, and then, when you want to grow your and then, when you want to grow your and then, when you want to grow your homelab, you can just add one or more homelab, you can just add one or more homelab, you can just add one or more physical mini PCs. That is, in my physical mini PCs. That is, in my physical mini PCs. That is, in my opinion, the best strategy you if you opinion, the best strategy you if you opinion, the best strategy you if you want to build your homelab around a want to build your homelab around a want to build your homelab around a cluster architecture. But, of course, cluster architecture. But, of course, cluster architecture. But, of course, running a cluster architecture in a running a cluster architecture in a running a cluster architecture in a homelab also comes at a price because homelab also comes at a price because homelab also comes at a price because every additional node also means more every additional node also means more every additional node also means more hardware, more networking, more power, hardware, more networking, more power, hardware, more networking, more power, and more things you actually need to and more things you actually need to and more things you actually need to maintain. So, on the other hand, a maintain. So, on the other hand, a maintain. So, on the other hand, a cluster quickly becomes more complex cluster quickly becomes more complex cluster quickly becomes more complex than running a single server, and there than running a single server, and there than running a single server, and there are many other devices and components are many other devices and components are many other devices and components that depend on this. I can just tell you that depend on this. I can just tell you that depend on this. I can just tell you from my own project experience, even from my own project experience, even from my own project experience, even though I use small mini PCs for the though I use small mini PCs for the though I use small mini PCs for the cluster, which are cheaper and more cluster, which are cheaper and more cluster, which are cheaper and more power efficient than one huge server, of power efficient than one huge server, of power efficient than one huge server, of course, I still need several of them course, I still need several of them course, I still need several of them running at the same time, plus the running at the same time, plus the running at the same time, plus the network back end and enough disk network back end and enough disk network back end and enough disk capacity. So, both the initial price and capacity. So, both the initial price and capacity. So, both the initial price and the ongoing operational cost can add up the ongoing operational cost can add up the ongoing operational cost can add up quickly if you don't pay attention. For quickly if you don't pay attention. For quickly if you don't pay attention. For my setup, for example, I said I needed my setup, for example, I said I needed my setup, for example, I said I needed to buy three completely new 2 TB NVMe to buy three completely new 2 TB NVMe to buy three completely new 2 TB NVMe drives, which are not cheap today, and drives, which are not cheap today, and drives, which are not cheap today, and with the three replicas, I only get with the three replicas, I only get with the three replicas, I only get roughly 1/3 of the raw capacity as a roughly 1/3 of the raw capacity as a roughly 1/3 of the raw capacity as a usable VM storage. And of course, usable VM storage. And of course, usable VM storage. And of course, redundancy is not a backup. I still need redundancy is not a backup. I still need redundancy is not a backup. I still need an independent backup server with larger an independent backup server with larger an independent backup server with larger hard drives that I need to keep running

hard drives that I need to keep running hard drives that I need to keep running so I can recover virtual machines from so I can recover virtual machines from so I can recover virtual machines from different points in time. So, that's why different points in time. So, that's why different points in time. So, that's why I'm also running a separate Proxmox I'm also running a separate Proxmox I'm also running a separate Proxmox backup server, but that's not all. You backup server, but that's not all. You backup server, but that's not all. You also need the networking requirements also need the networking requirements also need the networking requirements that can also be challenging and that can also be challenging and that can also be challenging and expensive. For example, for running a expensive. For example, for running a expensive. For example, for running a solid three-node Proxmox cluster with solid three-node Proxmox cluster with solid three-node Proxmox cluster with Ceph, you need 10 gigabit networking, Ceph, you need 10 gigabit networking, Ceph, you need 10 gigabit networking, and even there I made some compromise and even there I made some compromise and even there I made some compromise because I did not have another fast because I did not have another fast because I did not have another fast network interface on every single node network interface on every single node network interface on every single node for a completely separate Ceph back-end for a completely separate Ceph back-end for a completely separate Ceph back-end network, which honestly would be ideal. network, which honestly would be ideal. network, which honestly would be ideal. And then there are even all the services And then there are even all the services And then there are even all the services around it, Proxmox forum, Ceph monitors, around it, Proxmox forum, Ceph monitors, around it, Proxmox forum, Ceph monitors, managers, OSDs, placement groups, managers, OSDs, placement groups, managers, OSDs, placement groups, recovery states, and health warnings. recovery states, and health warnings. recovery states, and health warnings. So, every layer adds more complexity, So, every layer adds more complexity, So, every layer adds more complexity, which also means troubleshooting becomes which also means troubleshooting becomes which also means troubleshooting becomes harder. Sure, a cluster is an exciting harder. Sure, a cluster is an exciting harder. Sure, a cluster is an exciting project. It gives you more options, more project. It gives you more options, more project. It gives you more options, more flexibility, but it also gives you many flexibility, but it also gives you many flexibility, but it also gives you many more places to look. And sure, the more places to look. And sure, the more places to look. And sure, the learning experience might justify learning experience might justify learning experience might justify building this project, but does it building this project, but does it building this project, but does it actually mean you benefit from keeping actually mean you benefit from keeping actually mean you benefit from keeping this infrastructure running forever? I this infrastructure running forever? I this infrastructure running forever? I mean, at some point you have done the mean, at some point you have done the mean, at some point you have done the project and you have learned how to project and you have learned how to project and you have learned how to build and maintain it. At that point it build and maintain it. At that point it build and maintain it. At that point it just becomes a burden and takes time just becomes a burden and takes time just becomes a burden and takes time away from you that you could actually away from you that you could actually away from you that you could actually invest into building other projects. So invest into building other projects. So invest into building other projects. So if I summarize both setups in a direct if I summarize both setups in a direct if I summarize both setups in a direct comparison with a general recommendation comparison with a general recommendation comparison with a general recommendation for you guys, I think both setups have for you guys, I think both setups have for you guys, I think both setups have their pros and cons and I think it their pros and cons and I think it their pros and cons and I think it depends a lot on what you actually depends a lot on what you actually depends a lot on what you actually prioritize more in your home lab. So if

prioritize more in your home lab. So if prioritize more in your home lab. So if you want the simplest way to grow, maybe you want the simplest way to grow, maybe you want the simplest way to grow, maybe you have specific workloads that need a you have specific workloads that need a you have specific workloads that need a lot of CPU cores and memory together on lot of CPU cores and memory together on lot of CPU cores and memory together on the same host and your learning goals the same host and your learning goals the same host and your learning goals are more around virtual machines, are more around virtual machines, are more around virtual machines, containers, maybe virtual networking and containers, maybe virtual networking and containers, maybe virtual networking and more the software side of things and you more the software side of things and you more the software side of things and you are just okay with having internal are just okay with having internal are just okay with having internal redundancies, backups and occasionally redundancies, backups and occasionally redundancies, backups and occasionally longer maintenance windows are okay longer maintenance windows are okay longer maintenance windows are okay because you want to spend your time more because you want to spend your time more because you want to spend your time more on projects than actually maintaining on projects than actually maintaining on projects than actually maintaining the infrastructure itself. Then I think the infrastructure itself. Then I think the infrastructure itself. Then I think it is a better strategy to build your it is a better strategy to build your it is a better strategy to build your home lab around one huge giant server home lab around one huge giant server home lab around one huge giant server and just run everything on this one big and just run everything on this one big and just run everything on this one big machine. But instead, if you want to machine. But instead, if you want to machine. But instead, if you want to grow your home lab step-by-step by just grow your home lab step-by-step by just grow your home lab step-by-step by just adding more nodes at the time when you adding more nodes at the time when you adding more nodes at the time when you need them and you want independent need them and you want independent need them and you want independent physical failure domains and storage physical failure domains and storage physical failure domains and storage across the nodes because maybe you want across the nodes because maybe you want across the nodes because maybe you want to live migrate workloads during to live migrate workloads during to live migrate workloads during maintenance and you're okay with maintenance and you're okay with maintenance and you're okay with investing a lot more money into spare investing a lot more money into spare investing a lot more money into spare capacity, additional disks, fast capacity, additional disks, fast capacity, additional disks, fast networking and harder troubleshooting networking and harder troubleshooting networking and harder troubleshooting and maintaining, then of course it is and maintaining, then of course it is and maintaining, then of course it is better to build your home lab on a better to build your home lab on a better to build your home lab on a cluster architecture, maybe with a cluster architecture, maybe with a cluster architecture, maybe with a Proxmox cluster. So I hope I could help Proxmox cluster. So I hope I could help Proxmox cluster. So I hope I could help you a little with that decision-making you a little with that decision-making you a little with that decision-making based on my experience, of course. I've based on my experience, of course. I've based on my experience, of course. I've really made videos on all of these really made videos on all of these really made videos on all of these projects. You will find them in the projects. You will find them in the projects. You will find them in the description down below. So if you want description down below. So if you want description down below. So if you want to learn more about my Proxmox cluster to learn more about my Proxmox cluster to learn more about my Proxmox cluster or maybe about running a big enterprise or maybe about running a big enterprise or maybe about running a big enterprise server, I will link you this so you can server, I will link you this so you can server, I will link you this so you can check it out. I just want to say at the check it out. I just want to say at the check it out. I just want to say at the end what I have learned from making all end what I have learned from making all end what I have learned from making all of these videos and building all these of these videos and building all these of these videos and building all these projects that you should not start

projects that you should not start projects that you should not start building something just because it looks building something just because it looks building something just because it looks impressive in the server rack. Start impressive in the server rack. Start impressive in the server rack. Start with a real use case, write down your with a real use case, write down your with a real use case, write down your actual goals and the projects you want actual goals and the projects you want actual goals and the projects you want to learn, and sometimes less is really to learn, and sometimes less is really to learn, and sometimes less is really more. And building a complete home lab more. And building a complete home lab more. And building a complete home lab around one capable server might simply around one capable server might simply around one capable server might simply be the smarter choice for you. Now, for be the smarter choice for you. Now, for be the smarter choice for you. Now, for my home lab, if you ask me, I probably my home lab, if you ask me, I probably my home lab, if you ask me, I probably will keep just everything running will keep just everything running will keep just everything running because I enjoy making these videos and because I enjoy making these videos and because I enjoy making these videos and building all these projects for you building all these projects for you building all these projects for you guys. So, let's be honest, I probably guys. So, let's be honest, I probably guys. So, let's be honest, I probably won't stop buying more expensive gear won't stop buying more expensive gear won't stop buying more expensive gear and wasting time and money on building and wasting time and money on building and wasting time and money on building more clusters. But of course, that's more clusters. But of course, that's more clusters. But of course, that's just me. So, I know I want to hear from just me. So, I know I want to hear from just me. So, I know I want to hear from you guys. So, are you running everything you guys. So, are you running everything you guys. So, are you running everything on one large home lab server, or are you on one large home lab server, or are you on one large home lab server, or are you building a cluster with multiple smaller building a cluster with multiple smaller building a cluster with multiple smaller machines, or maybe you have a completely machines, or maybe you have a completely machines, or maybe you have a completely different strategy? Maybe you're running different strategy? Maybe you're running different strategy? Maybe you're running your home lab in the cloud, I don't your home lab in the cloud, I don't your home lab in the cloud, I don't know. So, please tell me your story in know. So, please tell me your story in know. So, please tell me your story in the comments down below, and let's talk the comments down below, and let's talk the comments down below, and let's talk about this on my Discord server. I think about this on my Discord server. I think about this on my Discord server. I think that would be cool. As always, thank you that would be cool. As always, thank you that would be cool. As always, thank you so much for watching. A big thanks goes so much for watching. A big thanks goes so much for watching. A big thanks goes out to all of my supporters in our out to all of my supporters in our out to all of my supporters in our community. You guys are really you make community. You guys are really you make community. You guys are really you make all these free tutorials and videos all these free tutorials and videos all these free tutorials and videos possible. And of course, I'm going to possible. And of course, I'm going to possible. And of course, I'm going to catch you in the next one. Take care.

catch you in the next one. Take care. catch you in the next one. Take care. Bye-bye.