Salesforce CEO calls site outages unavoidable
Disruptions are expected to become less frequent with new data-redundancy technology, CEO Marc Benioff says.
The Dec. 20, 2005, outage cut many companies off from critical data for hours on a busy, pre-holiday business day. It also called into question how well Salesforce, which stores customer and sales records for thousands of businesses, is holding up under rapid growth.
Benioff, who has commented little about the incident publicly, said in an interview at a media and customer event here that outages are an inevitable part of computing and that they happen very rarely at Salesforce.
"We don't want outages and we're doing everything we can not to have them, but we'll occasionally have them," he said. "That's part of computing...nothing runs at 100 percent availability."
Salesforce, based in San Francisco, claims an availability or "uptime" rate of between 99 percent and 100 percent. Yet a handful of customers that complained to CNET about the Dec. 20 glitch said smaller, less disruptive outages occur more frequently than they anticipated.
To combat such concerns, Salesforce has invested $50 million in new computing infrastructure designed to stay up and running in the event of a natural disaster. Benioff said during a speech Tuesday that the company will complete its conversion to its new systems, which include "immediate failover" capability, next month.
"It's a new level of performance, reliability and scalability for the company," he said during the speech. "Everything had to be rewritten. It was many, many steps we had to take and a huge amount of work over a year."
In addition to replacing nearly all its hardware and software and building two new data centers, Salesforce is adding database "mirroring" technology that will significantly boost its data redundancy capabilities, Benioff said. The mirroring system creates a duplicate database in a separate location and synchronizes the data instantaneously. In the event that one database is destroyed or disabled, the other one takes over.
But even mirroring would not have helped Salesforce avoid last month's snafu, which was caused by a database software bug, Benioff said.
"This will protect against huge natural disasters; we'll have zero downtime in a natural disaster," he said. "But it's not insurance against a bug."