You999
Using tape for backing up data that doesn’t need to be accessed at a moments notice is a great solution so long as your data sets are large enough to justify the high cost of the drives themselves. If you are backing up about 200Tb or less it’s cheaper to build a hard drive array but above that the cost per terabyte for tapes surpasses hard drives and makes a lot of sense. Although I hear managing tape libraries are a pain with how frequently they jam while swapping tapes.
I’d say replace your current server with a larger one that can pull double duty as a NAS and VM host. There are plenty of SFF nas cases on the market now.
The classic choice would be fractal design’s node 304 which fits six 3.5" drive in an ITX form factor. There’s also Silverstone’s CS381 which while being larger can fit eight hotswappable 3.5" drives and a micro atx motherboard.
Even if you go ITX you don’t have to feel limited by the lack of PCIe slots. Since m.2 uses the PCIe protocol it’s very easy to adapt it to your needs such as to an additional PCIe 4x slot. There are even m.2 10GBe cards in both intel and Realtek flavors.
Side question, what coral TPU do you have because it was my understanding that they use m.2, mini pcie, or USB and not the full size pcie slot?
Coming from someone who very much endorses amtrak and passenger rail, I think someone is abusing our tax dollars to make six figure jobs for their friends or family. Just look how much money is going into repeatedly studying if the stampede pass would be a viable passenger rail line.
Is that after the R&D expenses?
I’ve spent the last two weeks on getting a k3s cluster working and I’ve had nothing but problems but it has been a great catalysts for learning new tools like ansible and load balancers. I finally got the cluster working last night. If anyone else is having wierd issues with the cluster timing out ETCD needs fast storage. Moving my VMs from my spinning rust to a cheap SSD fixed all my problems.
I know you are saying Google doesn’t have to worry about redundancy to simplify the math but I think that makes it completely useless.
Redundancy is not just about having another copy incase of data loss but more importantly for enterprises redundancy allows for more throughput. If each video was on a single hard drive the site would not be able to function as even the fastest multi actuator hard drive can only do 524 MB/s in a perfect vacuum.
From my understanding, the old acpi only allowed for the kernel to request three levels of voltage/frequency on zen2 based CPUs. The new P-state driver allows for finer grain control of the voltage/frequency which may lead to a lower power draw and faster clocks for a given task.