By Heidi Biggar
It used to be that if you wanted to replicate data you had to do it at the array level using proprietary products such as EMC's SRDF, Hitachi Data Systems' XRC, or IBM's PPRC. But users' needs have changed, and vendors have responded.
"There's nothing good about array-based replication, except for the fact that you've got [big vendors] supporting and servicing it," says Arun Taneja, founder and consulting analyst at the Taneja Group consulting firm. "It's expensive, proprietary, and difficult to implement."
In contrast, fabric-based replication is much easier to implement, significantly less expensive (though not as inexpensive as host-based products from vendors such as Legato and NSI), and more flexible (in terms of array support) than array-based alternatives. For these reasons and others, analysts say we can expect to see a dramatic change in the way users replicate data over the next 12 to 18 months.
Currently, EMC's SRDF has more than a 70% share of the total replication market, according to the Taneja Group.
While analysts don't expect users to throw out their array-based replication platforms, they do expect users to gravitate to fabric- or host-based alternatives, or a combination of the two. "You should have your head examined if you're still doing array-based replication a year from now," says Taneja.
Fabric-based replication products are currently available from a variety of vendors, including Candera, DataCore, FalconStor, Kashya, Topio, and Troika (when combined with StoreAge's software).
Of these companies, Kashya and Topio can loosely be paired (both focus solely on replication and are pure software players), as can Candera and Troika (both offer storage services platforms that run a variety of applications, including replication); and DataCore and FalconStor (both offer a suite of applications that can run in the network or on storage services platforms or intelligent switches).
However, there are also some significant differences among these pairings, notably where the applications run (remote and/or local site), what platforms the applications run on (the software vendor's or third-party vendors'), and how easy it is to port the applications to other vendors' platforms.
While current interest in fabric-based replication is coming from non-switch-based approaches, switch-based replication is also expected to garner significant user attention over the coming months as products become available from vendors such as Brocade, Cisco, MaXXan, McData, and their software partners. For example, FalconStor has ported its IPStor 4.0 software, which provides a variety of storage services, including replication, to MaXXan's MXV320 Intelligent Application Switch.
"We're seeing interest from existing StoreAge customers who are already doing replication but are looking to expand their coverage to multiple SANs, and from users who have never done replication at all," says Steve O'Brian, VP of marketing at Troika Networks.
Earlier this year, Troika and StoreAge teamed up to deliver a fabric-based replication product. Troika expects to begin shipping its Accelera Network Storage Services (NSS) platform running StoreAge's Storage Virtualization Manager (SVM) software this month. O'Brian says that Troika is also pursuing opportunities to embed its NSS platform in either a blade or chip within an intelligent switch or even within disk arrays.
Similarly, Topio says it is seeing interest from users who have never deployed replication before, as well as from users who are already running array-based synchronous replication products, such as SRDF, and users who want to replicate over longer distances.
Whatever the scenario, "there is still a lot of education that needs to be done," says Eric Burgener, VP of marketing at Topio. "Users either assume replication can't be used in many environments due to cost and bandwidth constraints, or they aren't aware of the potential data integrity issues associated with many asynchronous replication products today."
Burgener recommends that, before buying any replication product, users get a good understanding of how a particular product maintains write-order fidelity (i.e., data consistency) and what, if any, impact the product will have on recoverability and performance.
It's a Catch-22: Asynchronous replication products get around the distance limitation issue associated with synchronous replication, but they can also introduce integrity problems if you're not careful, says Burgener. He claims that Topio's Data Protection Suite avoids this potential problem by using a system of global clocks that are able to keep track of multiple devices running a particular application or application(s). A similar capability is also available from Kashya. Both Topio and Kashya sell dedicated replication platforms.
"Because they are dedicated to replication, Kashya and Topio offer some 'bells and whistles' features [i.e., consistency] that you can't get from some of the other vendors," says Taneja. Both products allow users to keep volumes that were worked on at different times and on different servers in sync during the replication process, he explains.
While fabric-based replication may not be the cheapest way to replicate data from one site to another (host-based replication is), it's significantly cheaper than array-based replication. For example, Topio's replication platform is priced from $50,000.