Archive for the 'caching' Category

FeedBlendr Updates Live

Monday, May 21st, 2007

So those updates that I mentioned I was working on… they’re live now!

As mentioned in that post, this was mostly a bunch of core upgrades that you really won’t see any major differences from, but it makes life a lot easier for me going forward. It’s looking like it might also help with some of the caching problems we’ve been having (since I had to completely overhaul the caching system to factor in my new globalized caching code).

I’ll be keeping an eye on things to make sure the new code settles in nicely, but please let me know here or via email if you notice anything strange.

And that post about Amazon EC2 and S3 is coming soon, honestly :-) I’ll try to get it written up tomorrow, but I want to give it the attention it deserves.

Major Core Upgrades

Wednesday, May 16th, 2007

I’m currently working on some significant core upgrades to FeedBlendr that will also allow the service to start scaling towards some of the other tools that I’m going to be rolling out. These changes affect pretty much everything internally, and will allow me to add a few new features, and to build out the other tools faster>

  • New, standardized Caching framework (that will hopefully not cause more problems!)
  • Improved database connection handling (this will allow me to more easily use slave DBs etc when required)
  • Standardized and centralized “auto-discovery” class for locating feeds
  • Improved and generalized HTTP class, customized to handle everything I need and nothing I don’t
  • Totally separate “format conversion” layer for doing output in Serialized PHP, JS, JSON etc – some cool things can happen here!

I’m really excited about these changes (and some others that I’m still working on). This is the next step in ramping up towards the rest of the Feedville family. Keep an eye on this blog for more info!

Oh — and in response to my previous post about moving to Amazon Elastic Compute Cloud and Simple Storage Service, I got a few questions about my experiences etc, so I’ll be writing up another quick post summarizing it all soon.

Amazon EC2+S3, Here We Come

Tuesday, May 8th, 2007

I’ll explain in detail once I’m moved, but if you’re having problems with FeedBlendr, it’s probably because I’m in the process of moving servers (for real this time).

UPDATE 2007-05-13: As promised, here’s some more information about this Amazon business.

After running FeedBlendr on DreamHost for a year and bit, and having a few problems along the way with causing too much load, they finally pulled the pin on me. They’ve been very good about things, and worked with me to help identify and solve the problems I was causing (I was on a shared server, so my usage was affecting other customers). Basically, in the end there wasn’t a “fix” per se, because I just had too many people requesting blends too many times a day (over a MILLION times a month!), and had to do something about it.

In comes Amazon. For those of you who don’t know, Amazon has started getting into the Web Services world, and 2 of their offerings are of particular interest to me (us!):

  1. Elastic Compute Cloud (EC2): A system whereby I can request a new “copy” of a complete server on-demand, and use it as part of a cluster of machines to power FeedBlendr.
  2. Simple Storage Solution (S3) Unlimited, fully-redundant (in the good way) online storage, allowing me to keep copies of things out there in “Amazon-space” where my new EC2 servers can get at it.

I won’t bore you with all the details (although feel free to get in touch if you’re interested), but FeedBlendr is now running on 2 Amazon EC2 “instances” in a balanced manner (requests go to both machines), so hopefully performance is a lot better, and things will be more reliable. You may also have noticed that I fixed some caching bugs, so blends load faster, and should be more stable. I’ve also bumped up the minimum age for blends slightly, so you may notice now that blends can get slightly older before they will get refreshed. This is mainly because people were just requesting their blends too often, causing my servers to have to rebuild a complete blend every 5 minutes, just because a single feed changed. I’ll be looking at better ways to address this, but in the mean time, please let me know if this causes any problems.

Here’s to looking forward, and seeing FeedBlendr continue to improve and serve your feed-blending needs better!

Updates Never-ending!

Sunday, April 22nd, 2007

Another week, another round of updates and tweaks to the core FeedBlendr service. This week brings:

  • Fixed an RDF-parsing problem so that now you’ll get full content of a post if it’s available (was particularly affecting some FeedBurner feeds)
  • Added the ability to force re-check a feed URL if it fails (just click the red dot that appears next to it to force check it)
  • Fixed a few small things that were preventing feeds from rendering properly in FireFox, so both Atom and RSS versions should work now. Haven’t looked at IE7’s internal feed rendering yet, so I don’t know about that
  • Improved the internal caching models to speed things up all over the place
  • Switched over to a system that allows parallel outgoing requests for a pretty significant performance increase (especially on larger blends)
  • LOADS of small bug fixes and improvements throughout the core code

Growth continues, and I haven’t been kicked off DreamHost just yet (fingers crossed). I’m actually looking in Amazon EC2 to see if that will provide a reasonable option, so hopefully I’ll figure that out in the near future. I’ll keep you posted.

Keep on blending!

UPDATE: Looks like there was a bug in some of the caching that was preventing Blendr from picking up any new feed entries, but I believe it’s fixed now. Thanks to Orlin, Steve and benjy for pointing this out!

More Blendr Updates

Sunday, April 15th, 2007

I’ve just finished uploading and testing a new round of updates for FeedBlendr, please let me know if you spot anything funky!

  • Added a “What Are Feeds?” page (link) to the homepage.
  • Improved RSS output (stripped some unnecessary attributes, output <category> tags properly, and a couple others).
  • Changed all the URLs that you use to request blends and subscribe pages etc. I’ve added redirects to handle all the old formats and redirect them to the new formats, but please take note of the new formats as outlined on a blend’s detail page. This change makes all the request URLs much more RESTful.
  • Added the “Source” in the output for mobile reading.
  • Got rid of the separate subscription buttons and added a single, “AddThis” button which links to all sorts of readers, both online and actual clients.
  • Added caching to JavaScript and JSON output, so hopefully they’ll work a little faster. You may see slightly longer delays with updates of content, but actual page-loads should be a bit quicker for you now.
  • Added the option to get headlines-only from JavaScript and JSON outputs. Read about it in the Developers section

You can email me using this address if you notice anything strange, or would like to suggest a new feature.