1. 27
  1.  

  2. 6

    I use an NUC as my primary workstation. My needs are quite modest, and it works well for me. I run Xen on it, and at any given time, I tend to have a couple VMs running too. All of that said, I won’t buy new Intel gear.

    1. 2

      The NUCs are really cool. I’m using one for running various services from home rather than development, but it works great. They get amazing prices on eBay since there are companies that use them as very basic desktops and sell them when upgrading/moving.

      I also used avahi for access first, but then realised phones don’t do mdns. Ended up using ZeroTier network everywhere and just assigning real route53 entries for that network. It added the benefit of seamless external access.

      1. 1

        The discussion of CPU frequency governors got me wondering: Is it at all common for Linux servers in a data center to be running at less than peak performance because they’re using the powersave governor rather than the performance governor, and the admin wasn’t aware? I’m running Debian on a dedicated server, and it was using the powersave governor by default.

        1. 1

          I am using powersave governor and I am aware of it, there isn’t significant performance different between those two modes. And powersave seems more suitable when you have to pay the electricity bill

          1. 1

            And powersave seems more suitable when you have to pay the electricity bill

            Have you measured the difference? IIRC there was some paper a while back (that I’m now having trouble locating…) that basically concluded that something like ondemand was more energy efficient in the long term because it caused the CPU to spend less time grinding away on work and was able to return to a lower power state faster than then using powersave. With powersave, the CPU is active for longer when doing some work (because the performance is drasticaly reduced), and spent less time in lower power states.

        2. 1

          I’m considering doing a similar thing to this but with a VM. Although, now that I have my PGP keys on a yubikey I’ll need to figure out how to get those passed through for commit signing and SSH. Anyone have any ideas on how that could work?

          1. 2

            If you don’t need those keys on the host, the simplest solution is to forward the whole yubikey USB device to the VM.

            1. 1

              Yes, you can forward the GPG Agent socket to the remote system over SSH. It works seamlessly even where touch confirmation on the Yubikey is concerned.

            2. 1

              I have a similar setup except I use a corporate virtual machine. For a shell connection I use EternalTerminal with tmux in control mode and iTerm2. This lets me create native terminal tabs that are actually remote tmux tabs. EternalTerminal ensures that the connection never breaks even if my IP address changes (e.g. if I move from office back to home or I am on a wonky mobile connection).

              1. 1

                What is it with this fear of wearing out SSDs suddenly coming back up to haunt us? This wasn’t a topic of concern ever since the death of netbooks around 10 years ago. There’s no horror stories about failing SSDs and in my own circle there are tons of 2012 first gen retina MBPs around still in heavy use and I have seen multiple failures, but none SSD related.

                I get the use of NUCs for other reasons (Docker FS performance being one thing), but wearing out SSDs? Really?

                1. 1

                  I guess it depends on the work load, if your writing hundreds of GB a day you will wear out an SSD a lot sooner than most people. If that SSD is replaceable then its not that big an issue, as far as I know the newer macs have their SSD chips soldered on the motherboard so if your SSD goes bad you need a new motherboard, which normally means a new computer as the cost is near the same.