Friday, October 22, 2010

AboutUs Weblog

AboutUs Weblog

Our Office Plant Covert Op

Posted: 21 Oct 2010 04:30 PM PDT

The fabulous Nyco organized an office plant love and tenderness party! We are not showing before pictures, only after pictures. We hope the plants forgive us and prosper.

IMG_0077e

Raising Our Game: New AboutUs Data Center

Posted: 21 Oct 2010 10:45 AM PDT

Moving your entire production site to a facility located 175 miles away – while aiming for zero down time – is a challenging enterprise. In late August we did exactly that, migrating AboutUs.org from a hosting provider in Seattle to one a mere mile from our Portland office.

There are a few ways to manage a complex move with zero down time. We chose to build a parallel site with new equipment.

The simple route would have been to make an identical copy of the site we were running in Seattle, and install it in Portland. Instead of taking that easy path, we decided to seize the day and take advantage of a few opportunities the migration presented to us. We updated or upgraded almost every piece of software, and switched from a CentOS Linux platform to Ubuntu Linux Server. Our site has always run entirely on open source software.

The old site had a complicated infrastructure, including Memcache, Solr, RabbitMQ and various other pieces that we were able to eliminate in the move. We now run on 12 new servers from Silicon Mechanics: a MySQL database master; three MySQL database slaves; three Ruby on Rails application servers; two Sphinx search servers; and three Xen virtual machine hosts.

All of these new servers needed to be set up identically; we knew we were going to be adding hardware over time, and wanted to diminish the cost of managing all of this equipment. I am the only full-time systems administrator managing 30 systems, supporting the main site application, and supporting the development team.

To this end, the second component of our move was to unify our systems and software configuration management. This guarantees that we can replicate any server at any time from scratch, and can get new systems up and running in a matter of minutes. We are a Ruby and Ubuntu Linux shop and thus I chose Chef, from OpsCode, which is open source, implemented largely in Ruby and has good support for Ubuntu. This has been an invaluable tool to guarantee the repeatability of our configurations and systems.

Chef also helped a lot when we needed to build a number of Xen virtualization servers running on top of Ubuntu, which does not support Xen. We now have a Chef recipe that will run on a clean Ubuntu server and turn it into a Xen host on demand. Chef has allowed us to manage the new site in a way that never would have been possible in the old Seattle infrastructure.

While these were some of the more interesting aspects of our move, the main goals were:

  • to locate the physical servers where I can put my hands on them directly
  • to improve site performance
  • to increase reliability of the network layer

We are now happily located in the Lightpoint data center in downtown Portland, three miles from my house and one mile from our office. Rather than the single uplink from our Seattle location, we are now served by redundant links from Level 3 and Time Warner (TW Telecom).

Critical to the migration is the fact that the new location is just four milliseconds over fiber from the Seattle location. This allowed us to continue running from the Seattle database master until the entire site had been moved to Portland.

Lightpoint’s excellent network service has run without a hiccup since our relocation, something we were not used to in our previous facility. Server-side site performance has improved by 50 percent since July, due to a combination of better hardware, better connectivity, better configuration, a slimmer application stack, and strong support from the AboutUs development team toward improving the site application itself. Best of all, I am no longer fighting nightly fires that impact our visitor experience.

We managed the move with zero down time, and even managed to cut over to the Portland infrastructure a few days early when our Seattle provider was hit with a denial of service attack. We are continuing to look for opportunities to build on the gains we banked with the move.

Fix Broken Links to Save Link Juice

Posted: 21 Oct 2010 10:00 AM PDT

You’re redesigning your website to make it more usable for people and more easily found by search engines. Those are worthy goals. But if you’re changing any URLs on your website — the web page addresses — you’d better make sure you track down every link to every important new page, and change it. Otherwise your site visitors will get the dreaded 404 error message: Page not found.

AboutUs guest author Michael Cottam tells you how to conserve your hard-won link juice in his newest article about classic search-engine optimization (SEO) errors that even experienced web designers can make.

Michael is a Portland SEO consultant who’s advised clients in many different industries. His articles covering some technical aspects of SEO add to the wealth of advice about online marketing for business owners in the Learn section of AboutUs.org.

Are there any online marketing topics you’d like us to cover? Email me at Aliza@AboutUs.org and tell me what you’d like to learn.
Posted by at

No comments:

Post a Comment

Subscribe to: Post Comments (Atom)