Continuous Processing: Failure Is Not an Option
Stratus' new offering highlights the shortcomings of high availability.
- By Dan Kusnetzky
Stratus Technologies just announced that it's partnering with Solarflare to address the needs of companies in the financial services market, and applications executing on Linux systems.
Stratus says that the eighth generation of its ftServer, running Red Hat's Red Hat Enterprise Linux (RHEL) 7 combined with the third generation of its ftScalable Storage, provides extremely high levels of availability and continuous processing. The company also explained that its new systems are built using Intel’s latest E5-2600 v3 processors, and claimed a performance gain of more than 75 percent.
The company says that these systems would be a perfect fit for online trading applications in financial services, Voice over IP (VoIP) in telecommunications, and other applications where repetitive processing within a known constrained timeframe are critical.
Stratus Technologies is one of the few remaining players in the continuous processing market. How does that differ from high availability, one might ask? Continuous processing systems completely hide nearly all hardware failures by having totally redundant system designs. This is a different approach than deploying clusters of computers combined with monitoring and application movement technology.
Lots of '9's
Since continuous processing systems are designed so that every system component is duplicated and system firmware has been designed to both detects failures and press other components into service, failover times are typically measured in nanoseconds or milliseconds, offering uptimes ranging from 6 "9s" (99.9999%) to 7 "9s" (99.99999%). This is far better than the failover times of minutes to hours offered by clustered solutions, with uptimes of 4 "9s" (99.99%) to 5 "9s" (99.999%).
Competitors offering clustered high availability solutions composed of off-the-shelf systems combined with clustering, VM movement or other technology have been able to frame the industry conversation to focus on lower initial cost, rather than discuss the costs and potential costs of failure.
When Is 'Good Enough' Not?
As their products got easier to deploy and use, many customers were persuaded that a continuous processing solution wasn't the best choice. After all, they might say, why pay for additional 9s of uptime if the clustered solution is "good enough?"
In response, Stratus has focused on different market segments over the years, including:
- Government systems supporting first responders
- Cloud service providers offering very high service level promises
- Telecommunications service
- Health care systems
- Weather or geophysical disasters
- Financial services
Stratus sought out the niches in which customers couldn't tolerate any visible failure. These customers were typically providing services in situations in which people's lives or financial disaster was at stake.
It was increasingly difficult, however, for Stratus to get its message of continuous processing out over the noise created by many vendors offering "high availability" solutions that could deal with the majority of applications at what appeared to be lower initial prices.
Dan's Take: Worth Another Look
I've followed Stratus through many incarnations of their hardware and software products, and also through different owners of the company and company organization. I've always been impressed by the technology they've offered and thought that business and IT decision makers would be well advised to consider the company's technology for applications that can't be allowed to fail.
I hope this partnership and renewed focus on financial services and Linux applications will be the charm.
Daniel Kusnetzky, a reformed software engineer and product manager, founded Kusnetzky Group LLC in 2006. He's literally written the book on virtualization and often comments on cloud computing, mobility and systems software. He has been a business unit manager at a hardware company and head of corporate marketing and strategy at a software company.