Right now, I’ve recently happened upon an offer that the Planet is doing for new servers, basically for just roughly $40 more a month we can have newer servers that feature better processing power (the old servers were Celerons, the new ones are Dual-core Xeons) and larger hard drive space (both servers only had 80 GB SATA drives, the new ones have 250 GBs SATAs).
I’m still working out the pricing and other issues with the sales people at the planet, to see what kind of deal we could be given. (I’m working out whether we can be credited for the 1 GB of RAM we purchased back in January, so that we can apply that towards an additional 1 GB upgrade on the newer servers, should we get them.)
So instead of the total server cost being ~ $260 a month, it would be roughly $300, which isn’t a big increase in price, but a definite increase in performance.
Here are the techincal specs: http://www.theplanet.com/dual-core_xeon_3040_conroe/
I expect to have something later on tonight or, at the latest, sometime tomorrow afternoon.
Update (4/3/07): There is a slight issue with the $250 1 GB of ram we purchased mid-January. Apparently, this is not transferrable and cannot be credited to my account. So there is a small connundrum with upgrading both servers. The best solution, in my view, is to merely upgrade Apollo and transfer the 1 GB stick of RAM to Athena, as both Apollo and Athena are the same machine type. Or, barring that, convert the current Apollo server to the mysql server and rename it Athena, and buy only one Conroe server to replace the old Apollo server.
This entry was posted on Tuesday, April 3rd, 2007 at 12:32 am and is filed under Drifting Debris. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
Awesome.
On the one hand, I worry about covering the $300 per month, especially during the “offseason”.
However, hopefully our new wikis will be able to help pull their weight, and we’ll need some more horsepower to run multiple sites (especially considering that we were barely able to keep up, if that, at the peak usage of a single wiki).
I guess worst case we can always drop down to a lower tier if funds/advertising revenue/traffic dictates.
Go for the extra CPU, you surely need it!
I would also recommend, not that you need more to do, but that you redirect the DNS on battlestarwiki.org to some server that just returns a server down notice (perhaps with link to the blog). Just put a short TTL on the DNS record, so that when you are back up, you can switch to the new server with no wait.
Being totally down is not great form — folks have links to the battlestar wiki and it’s better if they show a notice rather than a failure. Perhaps do it as a temporary redirect so spiders are not bothered.
I’d say go for it!
The extra bandwidth you buy for those extra $40 per month results in more pageviews (more people can visit simultaneously, and people view more pages as they load faster), which in turn boosts ad revenue. I don’t know how much money you make on the ads, but they could (at least partly) compensate those $40.
I posted an maintenance page, which also doubles as a 404 error page.
As for that extra $40 dollars a month, that is covered by Amazon.com / iTunes commissions.
I haven’t studied behaviour lately, as to which is better to present spiders. A 404 may convince link-checkers etc. that there are dead links and report them to webmasters. I think the spiders of search engines are smart enough to require seeing this (or site down) for longer than a few days to get upset.
I think the best approach is still a Redirect-Temporary to another page (perhaps on this site) explaining things. Spiders understand temp redirct and will re-fetch when things are back to life. Users see something to explain what’s up. With luck you’re alive before too long anyway.
Yeah, go for the performance, the wiki definately needed it.
I always thought that it was more the MySQL server that needed to be upgraded rather then the http server.
Then again, that is based solely on just browsing the site.
We’ve been keeping an eye on both servers, which involves Shane and I basically keeping an eye on the process list via SSH while the server gets slammed fifty thousand ways to Colonial Day
and reviewing the statistics after that.
What it boils down to is that Athena, as she is right now, is fine. That issue we solved.
Our issue has always been the processing power of Apache / Squid on Apollo, which has painfully been apparent a few days immediately following the first airing of an episode.
I had the same initial thought regarding the upgrade (upgrade the machine without the extra RAM, swap the older machine to the SQL machine, possibly swap the names).
And if it turns out the new machine needs more RAM it looks like we should just go month to month, since it appears that “lifetime” is more relative than we realized and the odds are pretty good that the apache server will need periodic upgrades as deals like this make themselves available.
Sounds great, but why change the old Apollo to Athena? Just curious.
If you mean functionally, it’s because the SQL box doesn’t need as much horsepower as the machine that actually serves all the web traffic.
If you mean why switch names… “Athena” was intended to be the repository of knowledge (Goddess of Wisdom and what not), so that name may travel to whichever server happens to serve that role.
Hell Yeah
Macs would be better than that other stuff, its great were getting so big
As a Mac tech, I’d say sure, we’d get good throughput, but it’s still cheaper to build your own boxes and ensure you have diverse support options from several vendors and not one. If anyone wanted to host mirrors on their Macs, I’m sure Joe wouldn’t say no.
I’m game for more web power; going with the Conroe and swapping the IDs of Apollo and Athena would be fine with me.