Blog

I recently had a server with some bandwidth limitations (tested using scp and rsync -P), where I was wondering if the problem was the data being transferred, or the server's link speed.

The simplest way to debug and verify TCP performance is to install iperf3 and run an iperf speed test between the server and my computer.

On the server, you run iperf3 -s, and on my computer, iperf3 -c [server ip].

But iperf3 requires port 5201 (by default) to be open on the server, and in many cases—especially if the server is inside a restricted environment and only accessible through SSH (e.g. through a bastion or limited to SSH connectivity only)—you won't be able to get that port accessible.

So in my case, I wanted to run iperf through an SSH tunnel. This isn't ideal, because you're testing the TCP performance through an encrypted connection. But in this case both the server and my computer are extremely new/fast, so I'm not too worried about the overhead lost to the connection encryption, and my main goal was to get a performance baseline.

We're at the stage in the Red Hat drama where everyone is consulting history, trying to figure out what parts are being repeated in 2023 after Red Hat effectively locked down the sources used to build RHEL clones.

One talk linked quite often was Fork Yeah! The Rise and Development of illumos, by Bryan Cantrill over a decade ago. Bryan was a software engineer at Sun, who went over to Oracle after the buyout, then left to join Joyent, and now resides as CTO of Oxide.

The talk focuses on Sun Microsystem's handling of Solaris and OpenSolaris, both before and after their Oracle acquisition, and the whole talk is worth a listen—so much context about the history of ZFS, Solaris, Illumos, dtrace, and even UNIX and Linux history are contained within.

But there was one section (around the 32:00 mark) where if you substitute "Red Hat" for "Sun," rhymes with this year's "open source company" drama:

I went back and looked at some of the mail trails about this and like, "oh, my God!"

For LTX 2023, I built this:

This build centers around the Time Card mini. Typically you'd install this PCI Express card inside another computer, but in my case, I just wanted to power the board in a semi-portable way, and so I plugged it into a CM4 IO Board.

The Time Card mini is a PCIe-based carrier board for the Raspberry Pi Compute Module 4, and by itself, it allows you to install a CM4 into a PC, and access the CM4's serial console via PCIe.

But the real power comes in 'sandwich' boards:

As someone who champions truth, yet knows truth is bent to espouse many ideas, I realize clever phrasing often turns irrational lies into strong beliefs—especially when passion takes over.

And we in the open source community are a passionate bunch.

But to clear the air a little bit—especially as I have seen some zingers going both directions (from Red Hat employees to the community, and vice-versa):

As promised in my video comparing SilverTip Lab's DIY Pocket NAS (express your interest here) to the ASUSTOR Flashstor 12 Pro, this blog post outlines how I built a 6-drive M.2 NAS with the Rock 5 model B.

The Rockchip RK3588 SoC on the Rock 5 packs an 8-core CPU (4x A76, 4x A55, in a 'big.LITTLE' configuration). This SoC powers a PCIe Gen 3 x4 M.2 slot on the back, which is used in this tiny 6-drive design to make a compact, but fast, all-flash NAS:

I'm testing Adlink's Ampere Altra Developer Platform. This machine has a 96-core Arm CPU, but now they sell a 128-core version. Apple also recently released the M2 Ultra Mac Pro, so the model I'm testing isn't the "fastest in the world" like I could boast a couple months ago... but it's close, and I actually doubled my performance from last time—I'll show how later.