Platinum Sponsors

  • Intel
  • IBM

Gold Sponsors

  • NetApp

Silver Sponsors

  • HP
  • Google
  • MontaVista
  • Sandisk


  • Portland State University
  • Linux Foundation

Press Partners

  • Linux Journal
  • Linux Weekly News
  • Linux Pro Magazine

Sponsorship opportunities

For more information on sponsorship opportunities, please contact Angela Brown. Linux Plumbers Conf sponsorship packages.

Linux Server Management

These days, running Linux on a server doesn't mean quite what it did ten years ago–form factors have changed, and systems are expected to keep on chugging even though chunks of hardware can appear and disappear at any time. In this track we aim to discuss three things: Running Linux on devices in addition to hosts; exposing hardware location details in a running system; and how Linux interacts with system firmware (at least on x86).

Servers are no longer just a big box in a specially constructed room; Linux is now embedded on target devices: at first there was support for USB gadgets and SCSI targets, and now people are putting Linux on PCIE devices. Ideally we ought to convene everyone in a room to figure out how these setups are going to function and to see if there exist any ideas or functionality that could be reused or shared between device stacks.

On the management front, Linux is expected to perform continuously in the face of hot hardware reconfiguration; though the hotplug mechanisms are much improved over what they were a few years ago, there are usability problems that remain. Without an in-depth knowledge of the hardware device tree, mapping physical devices (PCI cards, disks, RAM, etc) to their in-kernel representations is an incomplete story. Hardware and firmware some times come from the factory with physical descriptions of the devices to which they connect, but there currently isn't a coherent way to export all that data to administrators.

Finally, there exists a fair amount of system functionality hidden in the system BIOS; we will discuss how those interfaces have been introduced and made more generic over time to prevent the encroachment of firmware-specific control warts, and how we might try to steer firmware interface and functionality development to work with Linux.


This track has space for a total of 2 hours and 30 minutes of content. There will be three short talks, which leaves us enough time for a short break and a wrap-up discussion at the end. Each talk should aim to fit in twenty-five minutes including Q&A, though there will be ample time for hallway discussions outside of the track. Following the three scheduled talks, there will be a 45 minute discussion section where people can talk about their concerns in a large forum, pose questions to the three presenters, or break off into smaller groups to hack up some code.

Please use these talks as a tool to discuss where something is inefficient, inelegant or wrong. Don't just propose a solution; also state the problem. Likewise, don't just complain; let's hear your thoughts about addressing those complaints. That said, if you've built some sort of device that the world has never seen before, some time could be spent airing that device publicly to gather peoples' reactions.

This is the tentative schedule for the track. The track will be from 10am to 12:30pm on Thursday, September 18th.

Time Session
0:00 Ground Rules
0:05 Arnd Bergmann
0:35 Andrew Chiang
1:05 Break
1:20 Len Brown
1:50 Discussion/Round Table
2:20 Next Steps
2:30 End

Accepted Talks

I have accepted these talks for the server track.

  • Infrastructure for PCIe endpoint devices - Arnd Bergmann - An increasing number of Linux based machines features PCI express ports that can be used as an endpoint…
  • Physical PCI Slots - Alexander Chiang - Using firmware data (DMI/ACPI/etc) to translate kernel device names to physical devices in PCI slots…
  • Hiding ACPI - Len Brown - Burying ACPI data/functionality in generic K/U “firmware” interfaces…

Other Talks

A few more talks that I argued for inclusion in Track 3 (miscellaneous):


Here are some ideas scraped off the front page that may or may not fit into this track:

  • Server manageability, systemtap, portability, IPMI and configuration management
  • Checking on long-haul daemons to make sure they're not slowly self-destructing
  • Virtualization, containers, namespaces
    • Putting more than 1 app on a server and keeping things tidy
    • Putting more than 1 VM on a system and keeping it up even when the host goes senile
    • How does the virtualization story change when we have obscene numbers of VMs?
  • Linux annoyances: What are they, and how can we eliminate them?
  • Putting the Linux kernel in non-traditional places
  • Embedded issues
  • Networking: going to 10Gb and beyond.
    • What are the necessary API and infrastructure changes?
    • Existing poll and RDMA interfaces are rarely used
    • Realtime and latency?

That list is sorted roughly in the order of importance. Do not be discouraged if your area of interest is not on that list.