In-Depth
Revivio Takes on Mirror-Splits
Time Addressable Storage writes block-level changes in real time, letting you go back in time
If you have ever wanted to provide continuity of operations for your organization the way that the big financial institutions do—namely, by mirroring data between arrays and providing near-instantaneous data restore if an array (or facility) goes south—but were discouraged by the hefty price tag for the solutions offered by the Big Iron vendors, then this column may be of interest.
Last week, a Lexington, MA company called Revivio that we have been watching closely for some time finally took the covers off of its new technology, called Time Addressable Storage. The idea behind the technology is simple enough: drop Revivio’s 25 processor server appliance into a fabric and it will begin mirroring data—and journaling changes to data blocks—on your primary storage on the fly.
The first part of the Revivio offering is a simple mirror: if you lose your primary storage, you can fail over to the Revivio-controlled copy. Where Revivio goes one better than traditional mirroring is with it’s Time Addressable Storage feature. Let me explain.
The well-understood bugaboo of mirroring is that errors and data corruption events are copied at the same high rate of speed between storage devices as are valid data updates. So, your mirror copy may be just as corrupt as your primary data once a virus or program error or disgruntled user has had its way with you. The solution to this problem offered by leading mirroring configuration providers like IBM, EMC, and HDS has been hardware intensive: the mirror split. In short, these vendors tell you to configure their arrays so that there are several discrete sets of disks used to make several discrete point-in-time copies.
In one shop I recently visited, the company had divided its arrays into eight sets of disks: one primary and seven “mirror splits,” each holding a copy of the primary data as of 11:59PM for each day of the week. So, if your data became corrupted on Tuesday, you could restore from Monday’s mirror split losing a maximum of 24 hours of transactions.
Revivio’s Time Addressable Storage takes a different tack. By maintaining one complete mirror, then journaling block-level changes at intervals of time to be determined by the user, Revivio enables you to dial back to that point in time just before the data corruption event occurred and restore your data to a usable form quickly, easily, and with a minimum of data loss. Pretty cool stuff.
Of course, restoring data by “rewinding” block level changes does not necessarily place storage back into a usable form. You may need to try reestablishing data at several different points in time in order to return data to an application-friendly condition (much like snapshot-based restores), but Revivio’s tools let you do so with a minimum of hassle.
Mike Rowan, Revivio’s CTO, tells me that the technology is robust and fast. It realizes his design goal of returning all of those disks currently allocated to split mirroring back into productive use. Moreover, it may be possible to use legacy hardware for Revivio’s mirror and block journal, prolonging the useful life of equipment already purchased, or even to use less-expensive arrays for data mirroring.
If you have any doubts about the Revivio solution, you may want to keep one set of mirror-splits in use for a period of time and use the Revivio-controlled mirror as a “tertiary” mirror. That way, primary data is copied to secondary disk in a traditional and dependable way, and the secondary is copied to Revivio’s tertiary as another process. Once you are satisfied that Revivio performs like it says in the brochure, you can return the secondary disk mirror to productive use and depend on Revivio exclusively going forward.
For companies already invested in multi-hop mirror-split configurations, Revivio can help you buy back some of the capacity you paid too much for to use solely for data replication. And if you have always wanted to mirror but found the price tag for the Cadillac solutions offered by the brand name vendors too daunting, Revivio may be just the ticket.
A final word: erroneous comparisons are being made in the trade press between Revivio’s Time Addressable Storage, Content Addressable Storage from EMC, and Commonality Factoring technology from Avamar Technologies in Irvine, CA. In fact, each of these offerings attacks different storage problems. EMC seeks to track data across its lifecycle as it migrates from more expensive to less expensive platforms. Avamar seeks to reduce the volume of data that needs to be replicated and protected by using a unique placeholder substitute for data that is replicated across many different servers and storage repositories.
I like to think of Revivio’s technology like a “way back machine” in the old Bullwinkle cartoon series. Time Addressable Storage provides a way to dial back your data to a particular point in time and go forward from there: editing out the impact of data corruption events from the new timeline.
It's worth a look.
About the Author
Jon William Toigo is chairman of The Data Management Institute, the CEO of data management consulting and research firm Toigo Partners International, as well as a contributing editor to Enterprise Systems and its Storage Strategies columnist. Mr. Toigo is the author of 14 books, including Disaster Recovery Planning, 3rd Edition, and The Holy Grail of Network Storage Management, both from Prentice Hall.