Failure's No Option For Avokia apLive

“We want to virtualize the database layer [and] scale it across data centers,” said Alan McMillan, CEO of Toronto-based Avokia.

Avokia&'s apLive software aims to do just that by adding redundancy and backup by clustering, replicating and load balancing any number of identical databases that are locally or geographically dispersed.

“You can use our product to upgrade your database, pause your database [and] do work on it. The customer will still be running while the upgrades are being rolled out,” McMillan told CRN. “It&'s basically clustering or load balancing—but across WANs.”

McMillan, an IBM veteran, said his company is partnering with IBM but the technology is database-agnostic. He expects the biggest market to be existing IBM DB2 and Oracle shops and said MSPs will be a natural constituency for his product.

Chris Toushan, country manager for IT services company SunGard Availability Services in Canada, is an apLive fan.

“Enterprise software for large companies typically has single points of failure. We work to spread that out,” Toushan said. “Avokia builds resiliency into other environments. If a customer has one data center, we work with them to design and develop a geographically diverse infrastructure.”

SunGard&'s overall solution, which primarily will serve financial services, education and government markets, also will include hardware, telecommunications infrastructure, software and “Avokia as a component,” Toushan said.

The apLive software, which formally launched Feb. 6 at Demo 2006 in Phoenix, costs $27,000 per database CPU or $225,000 per application for a maximum of 32 database CPUs. The key is working across sites and WANs. Oracle&'s much-touted Real Application Clustering (RAC) only works within a LAN, not between dispersed servers, a fact Oracle acknowledged.

Oracle, Redwood Shores, Calif., says its Data Guard and Streams software provides active/passive failover between sites. However, according to Sohan DeMel, senior director of product management at Oracle, the second site “will take some finite time to ‘catch up&' with the first site.”

RAC itself is an active/passive solution, and although there is no theoretical limit to running RAC clusters long distance, performance will slow because of latency issues, DeMel said via e-mail.

Continuent, Emeryville, Calif., offers what it calls a database-neutral solution that also claims to eliminate single points of failure. Continuent offers its software in both open-source and commercial versions.