>My Linux VMs are still Ubuntu 14 and getting quite old. I have only installed 32Bit versions, as I prefer to keep mail(Thunderbird), surfing(Firefox), some scripting (Node.Js, Python) and Office(WinXP) in separate VMs - my paranoid streak coupled with only 12 GB Laptop RAM.
>
>I might try out a few new distris: Mint32 feels slow and moans about missing graphics drivers, Zorin Lite looks like a mixture of XP, Win7 and some Linux - which is actually not bad, feels "alive" compared to Mint - but fails currently in installing guest additions, blocking out _clipboard - both on .ISO and apt-get install...
>
>When Win7 is unsupported, there will be a Win10 32 bit VM for all new Office versions and other Windows stuff. But I will try to prevent continous snooping by trying out current Zorin, even if it is 64bit, as they claim superior WINE support to try Office97, Office XP, Totalcommander and Fox there as alternative to my internet-isolated XP or Win10 "workhorse 32-Bit", perhaps Robolinux as well, as they aim for Win Users also with better VM integration no idea,what is supposed to be better done there. Not so sure about ReactOS, rumoured to be unstable.
>
>Perhaps PuppyLinux from the [tiny, install on USB] set as I want my Surf and Mail VMs be sleek and low on fat - and then read up on a few Arch distros, as they don't aim for the full load, but only install what you wish for.... Any hints in that area?
>
>Any further hints on distributions running well for single applications with 1 core, 768-1280MB Ram, 8-25 GB of SSD .Vhd as system parameters for the guest VM ? I know some small distros can run on less than 4GB disc, but copy speed of USB3 makes 10-20GB palatable, if the RAM footprint is low but current Firefox and Thunderbird work well.
>
>And Al, on SSD: in HD fixed disc size was clearly better for .VHD/.VDI. On SSD wear is more of a problem - fixed disc size (=fixed cells of SSD?) still the best practice ? All others invited to give their opinion, if they think about such odd HW topics...

In your position, one thing I'd look deeply into is how well your hypervisor of choice handles overprovisioning.

If you run a group of guest VMs where each guest continuously uses all of its provisioned CPU, RAM and disk then you can't overprovision - there's no "slack". Or, if you absolutely, positively require that each guest respond and/or perform predictably at all times, then you might need or want to allocate resources statically.

I've found most real-world processes, and the computers they run on (physical or virtual) to be very "bursty". If that's your case, and you can tolerate some unpredictability in the performance of your guests, you might be a candidate for overprovisioning.

Where am I going with this? Maybe you're not doing that now - perhaps you've precisely divided your resources amongst the guests you run. Maybe you've decided you have a fixed budget of 1 core, ca. 1GB RAM and minimal disk for a utility Linux VM. But does that really have to be the case? With overprovisioning, maybe you could double those values. When the VM is in the background, the hypervisor pages most or all of its RAM out to disk so the active ones can use it. Then when you switch your focus to it, the hypervisor backgrounds other VM(s) and brings back your full allocation for the utility VM. With extra resources it probably performs better and maybe more reliably. But more importantly, the larger budget gives you a wider range of distros to choose from.

I haven't looked into this in great detail. I know overprovisioning is a thing, and that different hypervisors offer different levels of support. On a quick Google I found a Dell document discussing it in the context of VMWare/vSphere in what appears to be a datacenter environment. I imagine a lot of the advanced concepts aren't applicable to single hosts but it's an interesting read overall. Google [Best Practices for Oversubscription of CPU, Memory and Storage in vSphere Virtual Environments] which should take you to a Dell PDF (sorry, couldn't figure out how to get a direct link).

As for SSDs: I'm not sure that wear per se is too much of a concern, even consumer SSDs are seeing big improvements in endurance: https://www.zdnet.com/article/samsung-860-evo-v-nand-ssd-offers-huge-endurance-improvements/ . As I understand it, it's more important that the host OS properly support TRIM so performance is maintained. This may not be the case with RAID, if for example you're using RAID1 for redundancy.

As you point out, high-speed external drives such as USB3, Thunderbolt 3 and the upcoming USB4 can offer expandable storage that in some cases can compare to internal fixed storage. There are almost too many options to evaluate.

In this age of ransomware etc I'm starting to think ease and performance of backups, to rotatable external media is important when choosing storage. For raw backup, copying a small number of large files (e.g. .VHDx) is ideal for performance. However, solutions that can do smart delta or rsync type backups are very attractive. I had some experience using Altaro on Hyper-V, daily backups of just what got changed were quick. https://go.altaro.com/vm-backup/altaro-vm-backup-gppc