Linux Scalability Project Status report for September and October, 2000

Linux Scalability Project

Status report for September and October, 2000

Center for Information Technology Integration
School of Information at the University of Michigan

The primary goal of this research is to improve the scalability and robustness of the Linux operating system to support greater network server workloads more reliably. We are specifically interested in single-system scalability, performance, and reliability of network server infrastructure products running on Linux, such as LDAP directory servers, IMAP electronic mail servers, and web servers, among others.

Summary

We're continuing to work with vendors such as Sun and IBM on their Linux scalability issues. We presented two papers at the Atlanta Linux Symposium. We're shifting our focus to concentrate on NFS performance and scalability. We're continuing to reach out to potential sponsors. Work continues on long-term projects.

Milestones

Marius Eriksen joins us on the Linux Scalability Project. Marius is a freshman computer science student who already has several years of experience working at linux.com on Linux scalability and performance issues. Welcome, Marius!
Stephen Molloy wraps up his scheduler work with a paper that will become a CITI tech report. Stephen will submit this paper to the Freenix track of the 2001 USENIX technical conference.
Marius and Mike Baker at linux.com have completed a Linux modification they call "snoopy." Snoopy hooks into the C library to monitor application behavior. Marius will submit a paper describing snoopy to the Freenix track of the 2001 USENIX technical conference. For more information about snoopy, see it's project listing on sourceforge.
Niels and Chuck presented two papers at the Atlanta Linux Symposium in mid-October. The two papers are available in the conference proceedings, or as CITI tech reports 00-1 and 00-7.
Chuck, Marius, and Stephen have studied the new TUX in-kernel web server for Linux. A CITI tech report describing our findings will appear in November.
In September and October, the NFSv4 project focused on implementing NFSv4 byte-range locking in the Linux 2.2.14 NFSv4 client and server. The implementation now passes all connectathon locking tests.
At the October 2000 NFSv4 bake-off, CITI's Linux NFSv4 implementation passed all basic and generic connectathon tests between a combination of servers and clients, including the Linux server and client, the Hummingbird NT client, the Network Appliance server, and the Solaris client. Several bugs were exposed during the connectathon tests.

Next steps for the project include bringing new graduate students on board, rebasing the Linux NFSv4 server to the Linux 2.4 kernel, and rebasing the Linux NFSv4 client and rpcsec_gss to the Linux 2.4 kernel.
iPlanet management has asked Chuck to focus on iPlanet product-specific projects, so he will be leaving the Linux Scalability Project in November. Because of the ongoing work with NFS on Linux, the LSP will concentrate mainly on NFS-related work going forward.
Dan Kegel (dank@alumni.caltech.edu) has created an independent benchmark to measure the performance and scalability of the poll() system call and /dev/poll. Dan's work has motivated him to design an even better version of /dev/poll for Linux 2.4 kernels. Dan's work is available here.
Work continues on improving mmap() performance. Chuck has designed a scalable in-kernel free area manager for process address spaces. He plans to replace Linux's AVL tree implementation for managing memory map metadata with something more efficient. We will submit a paper describing this work to the Freenix track of the 2001 USENIX technical conference.
We continue to pursue research relationships with Red Hat Software, Silicon Graphics Incorporated, VALinux, and Network Appliance.

If you have comments or suggestions, email linux-scalability @ citi.umich.edu