Thursday, May 10, 2007

1 Step Back, 3 Steps Forward!

1 Step Back: Cancelled Servers at

  • March 2007 was a challenging month for BlogPatrol as we waited for SingleHop to finally deliver its servers -- and then lost dozens of hours of development time on a fruitless exercise trying to configure the servers to our needs.
  • I won’t go into the details of why things didn’t work out for us with SingleHop, but I do want to thank Drew Hurlburt and the technical team for trying their best to accommodate us.
  • It was a painful decision to cancel the deal with SingleHop because of the loss in time and money. Not to mention, BlogPatrol's reputation taking yet another ding because we made the wrong choice for us.
  • Ultimately, finding a new provider was the right thing to do because it keeps BlogPatrol aligned with the main reason for starting over: giving everyone involved with BlogPatrol every possible advantage for good service.
1st Step Forward: Online with New Servers at
  • I'm very confident that we've now got the right machines in the right configuration and environment.
  • We chose servers at that are in the same class as the ones we tried at SingleHop: Dual Intel Dual-Core 5130 Woodcrest, 8 GB RAM expandable to 16 GB, and a RAID 1 array of 4 x 320 GB SATA Drives.
  • Those of you who are keeping up with this blog know that the old server that we'll be getting rid of later this month is being leased from, which The Planet bought about a year ago.
  • In keeping with the tenet of starting over, however, I signed up as a completely new customer at The Planet. Nobody at The Planet has any idea that I also was a customer of EV1Servers.
2nd Step Forward: MySQL Enterprise in a Master/Slave
  • A huge part of the problem with our old system was that we were using a single instance of an outdated of version MySQL Community Server. Relying on a single instance meant that all the writes (BlogPatrol recording when someone visits your blog) and reads (when BlogPatrol displays real-time your stats reports) were happening on a single copy of the database. Imagine 1,000 people trying to read and/or sign a wedding guestbook at the same time and you'll get an idea as to why the system kept failing.
  • We've re-designed the database so that it now uses master (writes) / slave (reads) replication using MySQL Enterprise Server in a setup that uses as many virtual machines as needed. Now imagine yourself back at that same wedding -- except that there one or two guestbooks that people can use to sign in (write) and several dozen other copies of guestbooks for people to read. Everyone's a lot more comfortable!
  • I'm very pleased to announce that we've completed the installation of this new database design!
3rd Step Forward: Public Beta Testing of the new database setup
  • We've completed our internal testing of the new servers and database configuration
  • We're now in the process of inviting users who signed up to participate in BlogPatrol 2.0 to test the system too.
...May 19th: Your stats will be available to you again soon
  • We expect to turn on BlogPatrol's full functionality, including access to your stats reports, on or before May 19th.

If you like this article, then click on this link to "Digg It" and help promote Thanks!