Disaster Recovery, Meet The Cloud

Disaster recovery-as-a-service (DRaaS), in my opinion, is one of the most exciting areas I look at. To me, using the cloud for disaster recovery (DR) purposes makes perfect sense: the cloud is an on-demand resource that you pay for as you need it (i.e., during a disaster or testing). Up until now, there haven't been many solutions out there that truly offered DRaaS--replicating physical or virtual servers to the cloud and the ability to failover production to the cloud provider's environment (you can read more about my definition of DRaaS in my recent TechRadar report), but so far today, we've seen TWO new DRaaS platforms announced from VMware and SunGard! Here's a quick roundup of what was announced today:

  • VMware. VMware announced at VMworld that they will be making their popular Site Recovery Manager (SRM), a DR automation tool, available as a service through hosting and cloud partners. At launch, participating partners are FusionStorm, Hosting.com, iland, and Veristor.  Benefits: Built into the VMware platform. Limitations: VMware specific.
  • SunGard. SunGard is announcing its Recover2Cloud services which leverages asynchronous replication or backup to send data to the SunGard cloud. Recovery time objectives range from 15 minutes to 24 hours depending on the service-level and method of protection. Benefits: works with physical OR virtual servers. Limitations: recovery is not as automated as with SRM, but SunGard manages everything for you.

Why infrastructure & operations folks should care: DRaaS has the potential to make your DR implementations less expensive and more automated. Enterprises should be taking a close look at their open-systems business-critical (i.e., Tier II or mid-tier) applications to see if they would be a good candidate for DRaaS. DRaaS most likely won't work for all of your applications, but it can be a powerful tool in your continuity portfolio. Be aware that like fixed-site DR services, most DRaaS providers will likely oversubscribe their customers so recovery will be on a first-come, first-served basis for widespread disasters. The benefit to you, of course, is that drives your costs down.

What are your thoughts? Would you use DRaaS? Why or why not?

Comments

Cloud-based recovery is still

Cloud-based recovery is still going to be challenging for any applications/systems that have legacy characteristics (e.g. mainframe elements, multiple platforms, non-virtualized components). Most Fortune 1000 companies are not 100% virtualized and still have quite a few "one-off" platforms, systems and storage that are deemed critical.

I completely agree

Hi Damian, I definitely agree with you that most companies are NOT 100% virtualized, nor is DRaaS a solution that will solve all DR needs at most companies. What I do suggest is that DRaaS can serve as a strategy for some tiers of recovery (not all), mainly for open-systems applications that are not too complex and can be virtualized.

DraaS does not require virtualized production servers

Mr. Walch is incorrectly equating the use of DRaaS with having to have a completely virtualized production environment. Not so. DRaaS services exist that can protect a mix of physical and virtual servers. And while virtualization offers enormous benefits in the cloud recovery environment, dedicated physical servers can also be part of the mix. Ditto his assertion that one-off systems preclude the use of DRaaS. A variety of protection & replication technologies can be used to protect all systems to a DRaaS target. Yes, there are DRaaS providers that only handle Windows-on-VMware environments. But that is a business choice that those providers have made, not an inherent limitation of the DRaaS model.

Would not use DRaaS because based on obsolete architecture

RAID Cloud Architecture is design of 6 smaller datacenters rather then the usual primary and DR datacenter plan. The datacenters are connected via layer 2 VPN to create a single virtual datacenter and servers are stripped across all six datacenters to provide high level, low cost (16% vs. 100%) redundancy. These 6 datacenters (colocation) are grouped into 3 datacenter cluster pairs west coast, center US, east coast. and SAN replication between DC pairs for additional redundancy. The individual DCs are small footprint 3 rack app. 250 vServer capacity designed with components for max. cost efficiency at this size. This is a modular design so start with one DC and deploy more as needed. small in size so can build in < 1 week. rapid scalability. More DCs can be added or cycled through to expand form 300 to 500 to 700 vServer capacity. NO DR because worst case is loss of capacity not outage
@parkercloud

Comment

Cloud based recovery for disaster requires considerable amount of management so it can be a challenging task for disaster management companies. The key to success can be by meeting RTO guidelines and service level agreements.