Data center-to-data center and data center-to-cloud optimization are fast-growing areas, says Don MacVittie, technical marketing manager for F5, which has WAN optimization modules that plug into its Big-IP chassis. But software-only WAN optimization seems to be the trend.
Certeon CTO Donato Buccella says the company's focus on software running on commodity hardware sets its products up for deployment in the cloud, where traditional hardware appliances just don't fit. "There is no place to put those pizza boxes," he says. And in branches, putting software on server hardware with other applications simplifies deployments, he says. Multiple applications running on virtual machines in the same server can simplify branch-office infrastructure, he says. Virtual optimizers are more flexible than the hardware appliances that are limited by the memory and processing power of the hardware. Adjusting the resources allotted to the optimizing virtual machine can boost its capabilities, he says.
While most vendors agree that virtual versions of their products are important, the basic functions the products perform to optimize traffic are the vendors' bread and butter. Silver Peak's CTO and founder David Hughes divides WAN optimization techniques into three buckets: network memory, network acceleration and loss. Memory includes deduplication and compression of traffic, acceleration minimizes the effects of latency by keeping WAN pipes full, and loss deals with preventing retransmission of lost packets through forward error correction. Acceleration also includes application-specific optimization for chatty protocols such as CIFS, Hughes says. But application optimization isn't something Silver Peak focuses on.
Riverbed does, though, and boasts a long list of applications and protocols for which it has written specific optimization code. By understanding an application, the WAN optimization software can anticipate what it will call for next and prefetch, and can proxy responses locally to keep chatty back-and-forth off the WAN wire.
Even competitors who don't focus on application optimization acknowledge that it works, but point to possible challenges. Application optimization requires frequent updating as vendors issue new versions of their software. "It's extremely hard and extremely complex," Buccella says. "It's about staying on top of everything."
The downside is that applications must be optimized one at a time, and as new versions come out, that optimization has to be updated. "Trying to solve the problem one application at a time really doesn't scale," Hughes says. "It's not whether application-specific optimization works, it's whether you could ever keep up."
Sign up for Computerworld eNewsletters.