Server platforms: experiment with your expensive hardware too!
Rusty R. Hall | Sun 24 Jan 1:30 p.m.–2:15 p.m.
Presented by
-
Jeremy Kerr
@jeremymeep
http://jk.ozlabs.org/
Jeremy Kerr is a freelance Linux and open source systems developer, working on the kernel, drivers, firmware and plumbing that (metaphorically) make systems tick. He's also worked on a few clock drivers, which (literally) make them tick too.
Jeremy is the technical lead of Code Construct, a small consultancy developing in the open source and embedded-systems area. Previously, he has worked for IBM's Linux technology center on their POWER platforms, most notably on the Summit, Sierra and Roadrunner supercomputer systems.
In addition to server platform development, Jeremy has contributed to a range of other open source projects, including petitboot (a Linux-based bootloader), K42 (a research operating system), patchwork (a web-based patch-tracking system) and nfsim (the netfilter simulation environment).
Jeremy's favourite piece of software is only eight instructions long, and only works on an ISA that no longer exists. Ask him about it for a disproportionately verbose explanation of what it does.
Jeremy Kerr
@jeremymeep
http://jk.ozlabs.org/
Abstract
We love to disassemble, rework, reconfigure and improve our consumer devices,
but tend to leave the more "serious" server platforms alone, with their warranties
intact. Most servers include at least one embedded Linux system for platform
control, which we can modify to suit almost any purpose. This talk brings the
concept of hackable hardware into the server space.
Recent server platforms have a huge scope for customisation, and most are built
on an open source core. With just a little background on conventional server
hardware designs, and the customisable components they're built from, we have
the exciting opportunity to adapt most server platforms to our own purpose -
whether it be tweaking the management interfaces to better suit a deployment
environment, implementing custom functionality for specific workload models, or
full-scale replacement of platform firmware.
This technical talk covers a background on common server design conventions,
plus the specific hardware, software, firmware and tooling that you'll likely
encounter in this space. This will provide a background for experimenting with
your own devices, and take warranty-voiding to a whole new (enterprise!) level.
We love to disassemble, rework, reconfigure and improve our consumer devices, but tend to leave the more "serious" server platforms alone, with their warranties intact. Most servers include at least one embedded Linux system for platform control, which we can modify to suit almost any purpose. This talk brings the concept of hackable hardware into the server space. Recent server platforms have a huge scope for customisation, and most are built on an open source core. With just a little background on conventional server hardware designs, and the customisable components they're built from, we have the exciting opportunity to adapt most server platforms to our own purpose - whether it be tweaking the management interfaces to better suit a deployment environment, implementing custom functionality for specific workload models, or full-scale replacement of platform firmware. This technical talk covers a background on common server design conventions, plus the specific hardware, software, firmware and tooling that you'll likely encounter in this space. This will provide a background for experimenting with your own devices, and take warranty-voiding to a whole new (enterprise!) level.