As organizations seek to have "always-on" services and access to data, plans have to give way to physics and financial realities.
In this video from his recent Storage Decisions presentation, analyst Jon Toigo highlighted some of the factors that can impact those efforts, including how latency has impacted application rehosting and high availability.
He noted that some organizations have conflated failover within an active-active cluster of a data center with "geo-clustering," which Toigo defined as using a WAN to span systems in different data centers.
He said that the nature of telecommunications systems limits how far apart data centers can be before latency can affect replication. Plus, signals don't necessarily travel in straight lines, adding to distance and potentially increasing latency.
"We understand active-active clustering in a LAN -- the boxes might be right next to each other. You're replicating stuff back and forth … you can see them both from the chair you're sitting in," said Toigo. "Try doing that over distance."
Toigo is CEO and managing principal of analyst firm Toigo Partners International, as well as chairman and co-founder of the Data Management Institute. He noted that the transition to third-party cloud services to provide support for some systems like email can run into trouble when vendors' services aren't available.
"Not all of the application these days lives on the application server. There are a whole bunch of configurations, a whole bunch of other parameters that also have to be backed up," Toigo said to his Storage Decisions audience.
"Is that acceptable? Would you trust mission-critical applications to something that the network is kind of flaky on? I wouldn't," he said.
Toigo noted that cost is also a factor in attempting high-availability for mission-critical systems and data. Wide-area network-based replication will require an "always-on" WAN service for active-active clustering, a backup network link, a remote "hot" site, plus the use of identical storage hardware for in-kind replication.
Also determining which applications need to be protected by HA, software needed to make this process work, training for staff and other efforts can also be costly, Toigo noted.
"One thing we don't have in great supply in disaster recovery is money," said Toigo. "We need a business-savvy recovery plan, something that basically delivers to management the protective services they're looking for at the lowest possible cost."