4 Nov 18:06
DLSLUG Notes, 1-Nov-2007: Ed Haynes of Wind River and real-time Linux
From: Ted Roche <tedroche@...>
Subject: DLSLUG Notes, 1-Nov-2007: Ed Haynes of Wind River and real-time Linux
Newsgroups: gmane.org.user-groups.linux.gnhlug
Date: 2007-11-04 17:07:13 GMT
Subject: DLSLUG Notes, 1-Nov-2007: Ed Haynes of Wind River and real-time Linux
Newsgroups: gmane.org.user-groups.linux.gnhlug
Date: 2007-11-04 17:07:13 GMT
A dozen people attended the November meeting of the Dartmouth-Lake Sunapee Linux User Group, held as usual on the first Thursday of the month on the Dartmouth College campus. This month, we were located in the lower level of the new Haldeman building, room 028. (Make sure you check the announcements to find out which building and room each month's meeting is in!) Ed Haynes was a former minesweeper crew member, a former Nortel employee and is currently a Wind River employee. Ed's a Technical Account Manager there, and was accompanied by Mike Gravel, an Account Manager, to answer non-technical/sales/strategy questions. Ed's been working with a proprietary kernel module from Wind River that loads into a standard Linux System and intercepts all of the interrupts, passing on to the kernel those for the kernel, and responding to those that require real-time responses. Essentially, the module in effect installs a real-time kernel that runs the Linux Kernel as the lowest-priority, non-real-time task, but responds to real-time tasks appropriately. Ed brought along several slides that defined real-time as it is used and misused in the computer industry, and talked about the concepts of response time, jitter, and the issues of pre-emption, the politics of Linux kernel development, and some of the recent Linux advances. Ed had a vivid demo. Using the Eclipse interface (we've seen in previous Wind River presentations that Wind River has contributed back several patches and new modules to the Eclipse project), Ed connected his laptop to a Via-based-processor motherboard to run the same code both with and without the real-time modules and could show how responsiveness might not change, but the predictability of expecting a result within a period of time could be made far more confidently. (Ed also mentioned that he picked up his prototype via newegg.com, for only a couple hundred dollars, and it gave him a reasonable match for a system he was targeting, with the advantage of a CD ROM and built case.) It was a great presentation and lead to some great on-topic and drifting-off-topic discussions. We got into a good discussion on how to have a remote shell available for administrative work when the server has gotten tied up with other tasks. Bill McG related a problem a client had with a database server getting hammered with requests, making a terminal unresponsive. Bill Stearns suggested checking out ionice to boost the responsiveness of a remote session, and we talked about what processes needed to be nice'd in order to administer a tied-up machine remotely, whether using a serial console or an ssh-and-screen technique. I thought this is something that ought to be a best practice to set up, if it could be done reliably and securely. Thanks to Bill McGonigle for organizing and promoting the meeting, for Ed Haynes and Mike Gravel of Wind River for making the trip to Hanover and a great presentation, and to all attendees for their attention and participation. -- -- Ted Roche Ted Roche & Associates, LLC http://www.tedroche.com
RSS Feed