Once the storage industry mastered the basics of RAID and could expect an external disk system to survive the failure of a disk drive without crashing, or without losing data, the next step was to protect customers from array failures.
This is achieved by replicating the data to another array, preferably in another data center. Replication was one of the first "value-added" features to appear on disk arrays and is available, often at extra cost, on all but the simplest arrays. The one catch for customers is that replication is homogenous--that is, you can only replicate data between arrays from the same vendor family. That's because array vendors have no incentive to support heterogeneous replication. If your shiny new ExaStor 7000 could replicate to anything, you might simply move your old disk array to your DR site instead of buying a second ExaStor for DR.
For a long time I've wondered why no one has simply leveraged iSCSI for heterogeneous replication. After all, an iSCSI array could have an internal iSCSI initiator that mounted a LUN on some other iSCSI array. It could then simply mirror its local LUN to the remote one.
Of course, it gets more complicated if you want asynchronous replication, and there's no simple way to implement WAN optimization, compression or any of the myriad features you can add when you control both ends of the link. That said, I still think there's value in support for some level of basic replication to dissimilar hardware.
[ Join us at Interop Las Vegas for access to 125+ IT sessions and 300+ exhibiting companies. Register today! ]
Whiptail's open-target replication manages to replicate data to any arbitrary storage system through the use of an application agent at the receiving end that runs on a Windows or Linux server. The application implements that same replication protocol that Whiptail's Accela arrays do, so it looks like another array device to the source array.
This architecture is a step up from what I imagined with my iSCSI replication model. First, Whiptail encrypts the replicated data in flight, which would otherwise require IPSEC or a VPN without the catcher application. In addition, Whiptail does snapshot-based point-in-time replication.
While this type of replication has a minimum practical RPO (recovery point objective) of 15 minutes or more, it has several advantages over real-time asynchronous replication. Because multiple writes to the same block over a snapshot period are aggregated, the point-in-time replication uses less bandwidth and is less sensitive to latency than real time replication.
The open target application stores its data in a raw logical volume, which can be on any block storage the Linux or Windows server can access, be it SAN or DAS. With this open target, users can have the performance of an all-solid-state array and the simplicity of array-based replication without having to invest in a second, expensive all-solid-state array for their DR site. The disk system at the DR site will, of course, be slower, but many organizations may find that acceptable when disaster strikes.