LibreBMC and Kestrel: two separate BMC tastes that taste separate separately

After our article earlier this week on LibreBMC and my concern as to what it meant for Raptor's own Kestrel "soft BMC" project, Raptor's Tim Pearson contacted me and advised that Kestrel is a standalone Raptor product on its own timetable that's still very much in development. In fact, the attached screenshot shows evolution from just last month with On-Chip Controller support now in the firmware, where it is able to read the temperature sensors and has scaffolding for monitoring fan RPM. Although Raptor makes it clear that Kestrel is and always will be open, and is developed on OpenPOWER systems using open tooling, it is mostly Raptor in-house code with the only non-Raptor bits being Zephyr (the OS), LiteX (the FPGA designer), Microwatt (the CPU core) and some minor components like I2C.

The other noteworthy thing is that Kestrel will indeed be offered as an aftermarket plug-and-play upgrade for existing Blackbird and Talos II systems, no soldering required. This is excellent news, because while LibreBMC is a very encouraging development and has wide-ranging implications beyond OpenPOWER systems, its basis on OpenBMC means a heavier installation that continues to be less well suited to a workstation environment (in terms of interface and sheer startup time amongst other things). While Antmicro's planned LibreBMC card may also work in Raptor systems, it's really meant for OpenPOWER servers as well as other server-class machines with onboard BMCs that aren't necessarily Power-based. Whether we like it or not, while there's a lot of nerd cred around OpenPOWER workstations, such systems presently represent a minority of the POWER9 installed base (let alone of all BMC-managed hardware). It thus makes sense that Raptor, currently the only manufacturer of OpenPOWER workstation machines, would be in the best position to create an open BMC that also is better tailored to improving the desktop user experience. As I've written here before, the envisioned improvements to fan control and user interface are very welcome, but Kestrel's promised 10-second power-to-IPL is a huge win for the desktop when ASPEED BMCs running LibreBMC right now take over a minute or more. LibreBMC's OpenPOWER core will certainly improve performance but I doubt it will be to that extent.

But I also suspect more good news than just an aftermarket upgrade is afoot. I pointed out in the LibreBMC article how much one of the physical form factors in particular looks an awful lot like an single-board computer, because really when you have a system-on-a-chip and on-board peripherals, you kind of have a "PowerPi" already. Raptor isn't saying, but Kestrel, like LibreBMC, will necessarily have its own SoC at its heart and thus could easily be the basis of another OpenPOWER SBC project, possibly one that might be even better suited for that environment. Plus, being Raptor, you know all the components will be open and blob-free (especially now that they've eliminated the blobs for the onboard Broadcom Ethernet, leaving just the Microsemi PM8068 as the last blob firmware component and only if you buy it as a BTO option). It will be worth watching to see when and how all this comes to market. Either way, getting your choice of BMC alternatives is a really great thing especially when the BMC is so critical to the system as a whole.