OSCON 2006 Notes
Posted by Jim Jagielski on Monday, July 31. 2006 in Open Source
Well, I'm back from OSCON and the trip itself wasn't as bad as I had anticipated. I flew NorthWest out there (and back) and on the outgoing flights there were delays out the wazoo. But I didn't miss my connection and was able to get some reading done, so it wasn't all that bad. What *was* bad is that I missed a dinner meeting with Raven Zachary of The 451 Group, but was able to meet him late that night for drinks at the DoubleTree bar.
I spent my time mostly split between the Covalent booth and the ASF booth. It was very nice seeing so many people again. One bad thing about living on the East Coast is that so many people I know live on the other side, so I only get to hang with them not as often as I'd like. I'd never move, but if I did, that would be one big reason to do so. A bunch of us grabbed dinner in downtown Portland and then headed back to the convention center. Of course, we took the wrong line, so we needed to back track and grab the right one. And we're the people you trust to write code. Most likely, we were all still distracted by the young woman playing pool at the restaurant/bar.
I also met with Rachel Chalmers (also of the 451 Group) and Steve O'Grady (from RedMonk), for the more traditional analyst meetings. My talk went OK, although I had a bunch more content than time, so I had to breeze past like about 5-6 slides.
As for OSCON itself, I really enjoyed the sessions I attended, and the show itself seemed to draw a nice size crowd. It's a darn shame that the wireless connectivity was as bad as it was. Helpful hint: if you want to provide wireless, do it right and contact Cliff Skolnick of Toaster.Net, et.al.
As expected, I regretted not being able to stay longer. Next year, I'll need to plan some more time out there, and buy an extra ticket so I can take Eileen out with me and spend some time just sightseeing.
More...
OSCON 2006
Posted by Jim Jagielski on Wednesday, July 19. 2006 in Open Source
Next week I'll be suffering from cramped legs and a sore back as I'll be flying cross country to attend and speak at OSCON. Flying is no longer a fun event, at least for me. In fact, it's hardly tolerable anymore, what with the long lines and the packed flights. Plus, as I've mentioned before, my productivity goes *way* down when traveling, which I find just as painful.
Still, events as cool as OSCON make it worthwhile, so despite me whining about getting there, I'm really looking forward to attending the conference. I have a short session on Wednesday (4:30pm) regarding LAMP and I'll be manning the ASF booth and the Covalent booth when I can.
But I still wish that there were at least *some* good open source related conferences in Baltimore... We really do have a nice Convention Center.
More...
More proxy work on the httpd trunk
Posted by Jim Jagielski on Friday, July 14. 2006 in ASF
A few days ago I added "hot standby" to the mod_proxy_balancer code in httpd-trunk. This allows for a balancer worker to not be used at all unless all the other workers are unusable (stopped, disabled or in error). Not only does this add useful functionality but it kicked off discussion on other load balance methods (like a cluster set value, kind of like mod_jk's "distance" attribute) as well as how the Proxy module uses the httpd scoreboard.
Apache's scoreboard is a central shared memory structure accessible by all the Apache processes and threads. From the start it's had a very specific function, but it's just so darn useful that there's a real temptation to use it as a central dumping point for all kinds of shared data, which I try to discourage as much as I can.
So there's discussion now about how to better abstract out the scoreboard, not only to make it easier for modules to "extend" it, without monkeying with it, but also to allow other storage providers to be used. One can see how a memcached-compatible scoreboard would be very useful for load balancing. In essence, I think we're looking at a suite of scoreboards, each for specific purposes, which default to the apr_shm implementation as a baseline. Should be quite useful.
Anyway, I'm looking at other improvements to the proxy balancer, and hope to propose backporting to 2.2.x after sufficient testing and feedback. If you have suggestions or comments, don't hesitate to send them along.
More...
Page 1 of 1, totaling 3 entries