BSDCan2011 - Final (with audio).5

BSDCan 2011
The Technical BSD Conference

Lars Noldan
Day Talks - 2 - 2011-05-14
Room DMS 1160
Start time 10:00
Duration 01:00
ID 216
Event type Lecture
Track System Administration
Language used for presentation English

Speed Daemons

High Performance FreeBSD Clusters for Fortune 500 Companies (and Their Admirers)

This case study will focus on how advanced networking in FreeBSD was leveraged to create a high availability, high performance web hosting cluster for a Fortune 500 company

This case study focuses on a Client's request for a website with no single points of failure, an SLA that requires 99.95% uptime, and Disaster Recovery capabilities. Fulfilling these requirements meant configuring ten servers across two datacenters. Facilitating high availability and high performance required extensive use of FreeBSD networking tools.

The project uses the ZFS file system to help solve key data synchronization and shipping problems. During configuration a few weaknesses in FreeBSD were identified. Most notably, problems with iSCSI and ZFS ultimately lead to a moving the storage servers to OpenSolaris.

We will discuss the methods used to keep the Disaster Recovery datacenter in sync with the production datacenter, and the methods employed to keep search data as well as web content up to date in near real time. This involved writing a custom ZFS snap shot system referred to in the office as the "Fortville Snapshot Distribution," named after the town our office is in, Fortville, IN.