The Linux Foundation

 
VAC Server Focus Group

From The Linux Foundation

Added by Gerrit - as a discussion point over the next few meetings to refine the topics and focal points of the Vendor Advisory Council.

Many of us on the LF-VAC have talked previously about areas in which Linux could/should be improved and that perhaps the Linux Foundation could help direct resources toward requirements on which there is consensus.

To that end, I'm submitting this candidate list for discussion. It comes from discussion with some of the IBMers who have been active in LF, OSDL, and FSG. I also reviewed notes from the last couple years of LUAC meetings and added most of those in as well.

  • Further SystemTap improvements
  • General Linux fit and finish; "out of the box experience"
  • Linux legal/IP issues -- possibly outside the scope of this discussion, but still very important.
  • Device drivers and kernel driver concurrency. Conflicting requirements for version/level for DDs and distributions required by ISV applications. The result usually works but is officially unsupported.
  • GCC backward compatibility.
  • Tools to assist in multi-threading. Many existing Linux applications were ports from Unix or Windows, so may not be threaded as well as on their original operating systems.
  • RDMA hooks for kernel
  • Life-cycle bug management: patches/upgrades not well handled (some even crash system or applications), very difficult to upgrade large collection of systems.
  • Linux is thought by many to use more power than its counterparts, even on servers.
  • TCP stack performance doesn't match peers (need a better Linux port?)
  • Multi-processor KVM and/or getting Xen in Linux kernel.
  • Crash Dump, including more selective/selectable dumping. Perhaps we need crash dumping to be enabled at install time, so users aren't told they have to reproduce the problem.
  • Microsoft interoperability: broad range of opinions here. Some want Linux to fit into Active Directory as a client. Others want Linux to replace an AD server. Still others just want single sign-on and don't care how. Some possibilities:
    • Samba/CIFS use factors - make easier to use for SMBs without large/trained IT staffing
    • Samba/CIFS Feature Gaps - remote home directory, roaming, offline nameservice, etc.
    • Multimedia Format support - a common open format similar to ODF - preferably OGG-based
    • Single Sign-On - particularly with older MS protocols, SMBs don't usually have LDAP running
    • Out-of-Crate detection and connectivitiy - to an existing MS network
  • Desktop Linux Inhibitors: application support, peripherals/device support, end-user training, Windows appls, linking with mobile devices.
  • SELinux (this one may be out of date): installation and adminstration of SELinux systems was much harder than standard and not well handled by existing tools/products.
  • Lack of good cross-hardware and cross-operating system tools for system administration.
  • ZFS ease of administration parity


Comments? Where is there consensus?

Thanks,

John Beauvais


[Article] [Discussion] [View source] [History]