
I run quite a bit on my home network, these are largely docker containers in a VM. Some examples are Home Assistant to control many aspects of my home and Mealie to hold delicious recipes. I run around 80 docker containers and a a few VMs. Each of these machines run just under $400 and are extremely fast considering.

Source: Amazon
All these run off of a couple of mini pcs running Proxmox. Proxmox is a very easy to use virtual machine system that has native clustering and failover. I run two mini pcs (AMD 5700U w/ 32G ram) along with a third device for a final quorum vote on a Raspberry Pi. The third device acts as a moderator to prevent both machines claiming control at the same time.
This system works fantastically, if I pull the plug on either of the devices that run VMs, the other will take control and start up all the machines in under 3 minutes.
I have been running close on ram, and I wanted to increase it. I saw no point in replacing the ram and buying DDR4 which will cost just as much as DDR5 with the ai craze. I decided to upgrade the two primary machines which will give me a lot more ram and 3-4 times more performance.
The primary reason I run mini pcs is because they are so energy efficient and it's just really cool to have all almost 100 machines running on a machine I can fit in my pocket.
The system I chose to replace it is the Minisforum MS-A2, a mini workstation that has an extremely fast CPU, can hold 3 NVME drives, 96GB of ram, and dual 10Gbit SFP+.

The total system cost is just over $1,300. Considerably more expensive than my previous mini pcs, but the 10Gbit SFP+ is a big seller for me and the much faster CPU.
While I had zero performance issues despite running a crazy amount of docker containers, this will be a massive upgrade in terms of CPU performance at a significant power increase.

The big advantage of my previous setup as each machine generally used 10W-30W, most of the time under 15W.

This is a chart of the energy consumption over the 2-3 days prior to taking it offline. Here is a chart of the secondary node.

While this machine isn't actively running any virtual machines, it does keep the ZFS file system in sync with the primary device so it can fail over quickly and without a restore.
If you are curious, these charts are from my Home Assistant instance that I have smart outlets on most of my hardware and can pull data from any day or second in the last 4 years.
So what about my new systems? Unfortunately I could only get one, the ram is very hard to get and it will take about a week to get another DDR5 96GB set sent over. For now I am using my EVO-X2 which is very similar performance (about 95% as fast) and even has 128G of shareable ram. I was using this for AI, but I have new hardware coming in that is much much faster.

I accidentally got a very short DAC cable that won't allow me to use 10gbit until the new one comes in. I did get some high end thermal paste to repaste the CPU as these machines can have thermal issues at full speed.
Installation is not as easy I would like but still only took a few minutes.

Installing the NVME drive was the easiest part. I didn't have to remove a single screw outside of the nvme set screw. The case actually slides off with a single button press. This is a Gen 4 2TB nvme. I don't use a ton of space, only about 120-130g of space. Anything super large is generally stored on my NAS.

Installing ram involved removing 3 screws and slotting them in.

Finally I removed the heat sink which was another 3 screws and cleaned it off. I only removed the thermal paste on the CPU and applied Frost X45 thermal paste, considered one of the best thermal paste available right now. 15.2w/m-k is considerable higher than typical 8.5w/m-k thermal paste and even the higher end 12w/m-k thermal paste. Because this device is in such a small case and is nearly as fast as a $800 AMD 9950X, I wanted to have the best paste you can get without going liquid metal.

After everything came together, I installed Proxmox and rebuilt a new three node cluster.
I have a second machine on order and should have it in a week hopefully.
While i had no complaints of the old system, it ran surprisingly well, there is a noticeable response time improvement when doing backup/restore and loading apps. It is all micro seconds, but it is noticeable.
Now I can check out my recipes and decide what I want for dinner.

I will likely use one of the old machines as a new backup workstation if anything goes wrong with my primary and sell or give away the second.
upgrading home lab is always fun
Those mini PCs sound like a solid upgrade for running containers smoothly. 💪
Now you just need an uber eats function for the home lab with a robot that automatically cooks the meal for you.
It hurts us that Sun really liked you.
He shared his Autism woes with you and you used it against him.
We will destroy you sir. You will know unimaginable pain.
You may as well speak openly.
We know this isn't what it seems.
And we know what it says.
And who it talks to.
Hubs bought an Asus recently, he refuses to buy any MIC hardware if he can avoid it. That, along with his Raspberry Pis keeps him happy and occupied
Very impressive. Those coconut cheesecake bars look interesting. If you send me some samples I will provide my gustatory estimation, which will be of little, even negligible, utility, but will defo generate good karma.
Thanks!
You can have the recipe if you want.
I bet I could enjoy even the mess I would make of it. I have come up with the best banana bread recipe I've ever tasted. I could even trade for it with my secret recipe for that.
Enjoy
My Banana bread:
3 bananas
2 eggs
1 stick of butter, melted
1 tsp baking soda
1 tsp baking powder (aluminum free)
1 tsp vanilla extract
pinch of salt and cinnamon
1 cup flour
1 cup sugar
1 cup whey protein (I prefer vanilla flavor, but there's other flavors)
1 cup chopped walnuts
1 cup raisins
~6-9 dried plums (prunes) to top the batter in the pan.
Bake at 350 for an hour or so, until a toothpick inserted into the loaf comes out clean.
Substituting a cup of whey protein for flour as above reduces carbs while greatly increasing protein content, and also increases the moistness of the bread, resisting degelatinization of the starches and going stale.
Enjoy!
Edit: geez! I forgot the sugar. Fixed now.
I just like banana, not all the other stuff. I like walnuts but not in my brownies, banana bread, or cookies.
Well, it's good that you get to make what you like best, I reckon. I've got a pumpkin on the porch Imma try to do something with soon, probably my next attempt will be pumpkin bread.
You also might like subbing in 50% whey protein for flour, as it does make the bread moister.
I'm a bit confused by the pic of the bread above, as it appears to show chopped walnuts in the sliced bread that aren't included in the recipe.
As I usually enter recipes way before or after I used them, I will frequently just grab an image off the Internet to fill it in so I have a nice image. I don't usually take pictures when I cook things.
Not sure there is a better banana recipe than mine
Tastes differ, and we make what we like. It's likely you'll prefer yours to mine, but I like mine, and everybody that's tried it liked it too. Maybe you'll like it too. I won't be wounded if you prefer yours, because you should prefer things made the way you like them.
That is pretty cool. So were you able to just add the new machine to the cluster and it got everything it needs for when you decommission the old ones? I love those mini PC's I have been using them more and more myself.
I created a new cluster, it was super easy.
And there. Look.
I switched it off.
Cool huh?
This is me right now:
Cherrie already turned on you.