Data replication and disaster recovery demands on storage nets

While the press concentrates on IP-based storage-area networks, it seems that every conversation with vendors or end users about IP storage eventually turns to a discussion of IP-based data replication and disaster recovery. Some vendors of IP storage management solutions even say that replication is driving most of their business.

IP-based replication is hardly a new idea, so why is there so much interest? The answer is found in the fortuitous convergence of evolving business requirements with a new generation of storage management systems. The result is that large and even midsize organizations can create advanced data protection architectures that would have been impossible or cost-prohibitive just a few years ago.

The events of the past few years have raised the awareness of regulators, auditors, IT professionals and even senior management to the need of providing business continuity in the face of extreme events. Real-time transaction systems that are available 24-7 require round-the-clock real-time data and systems protection. Additionally, business and legal demands now require up-to-the-minute protection of more data types than were previously required, including e-mail.

By contrast, traditional disaster recovery infrastructures are more reminiscent of the batch-processing era than of the Web generation.  Hot/warm/cold disaster recovery sites are populated with data by Chevy trucks full of backup tapes or host-based replication tools. Because of the overhead associated with these methods – particularly in terms of time required and people involved – data is generally updated only once a day, exposing organizations to significant data loss in the event of a major outage.

For much the same reason that IT has used snapshot technology to shorten the backup data protection interval, modern disaster recovery requires virtually real-time data replication strategies. Unfortunately, simple synchronous mirroring strategies aren’t appropriate because the latencies associated with long-distance IP communications would have a detrimental impact on application performance.  It would seem that disaster recovery demands asynchronous replication.

In an asynchronous replication environment, transactions are committed once a local copy (or copies) of the data is written to persistent media. It is then written on a delayed basis to the remote site. This enables high-speed transaction processing while minimizing potential data loss in the event of a disaster.

High-end storage arrays, such as the EMC Symmetrix, have begun to incorporate direct IP connectivity and asynchronous replication technology. For homogenous storage infrastructures (Symmetrix to Symmetrix, IBM’s Shark to another Shark) this may be the preferred solution.

However, as noted by Anne Skamarock in her newsletter about storage network intelligence (see editorial link below), there is another option: in-band storage management appliances (and eventually intelligent SAN switches), such as FalconStor’s IPStor and HP’s CASA, which provide more than virtualized storage pooling and provisioning. They also offer a variety of data movement services including Fibre Channel or IP-based data replication.

More on this next time.