Storage essentials (by Merlin Ran)

Editor's Notes

• #4: Direct-attached storage architecture has distinct limitations. Only one server can use the storage attached to it. Storage that goes unused becomes a “stranded” or underused resource. If an application is saturating one server, you cannot deploy another server to share the burden because the application data is only available on the direct-attached server.
• #5: Note to Presenter: View in Slide Show mode for animation. Deploying a SAN lets you share storage. By consolidating with a SAN, you can reduce your total cost of ownership and eliminate needless duplication and resynchronization of files. SANs provide greater storage usage through consolidation. You can easily allocate storage to servers as needed. And, as we’ll see on the next slide, a SAN makes backup easier and more efficient.
• #16: RAID 3?
• #25: Note to Presenter: View in Slide Show mode for animation. The key to determining the best replication solution for a particular use case is a thorough understanding of service levels. Specifically, you need to balance business requirements for performance, functionality, availability, and economics against capacity needs, bandwidth requirements, and overall total cost of ownership. Every solution has benefits and risks, and you probably want to balance all of these factors in order to decide on the best solution—whether it be for daily backup and recovery or remote data replication for application consistency and disaster recovery and restart. In the end, the need typically is around assuring application consistency and recovery. Application consistency and recovery is important for sustaining business operations—and, not surprisingly, are the top replication and disaster recovery challenges EMC hears from both customers and analysts.
• #26: When EMC or its partners talk about remote replication, they usually mean between storage at two locations. The source and target are physically separated to reduce the risks associated with co-location. Remote replicated systems could be across a campus, across a town, or across the globe. Their physical distance and technology selected can affect how quickly you recover from a disruption and how much data is lost. Organizations normally set requirements for how much lost data and how much time to come back online is acceptable. Recovery point objective (RPO) is the point in time to which critical data must be restored following an interruption before its loss severely impacts the organization. Note to Presenter: Click now in Slide Show mode for animation. Recovery time objective (RTO) is the amount of time that it takes to recover the data and restart business services—including critical applications central to business operations—from recovered data before the absence of the data or applications severely impacts the organization. The bottom line is that replication is appropriate for all types of data, and the RPO and RTO you target are going to affect your implementation.
• #27: Not surprisingly, different replication technologies yield different RPO and RTO results. And, though zero RPOs and RPOs may seem ideal, each approach and its corresponding technologies has its place depending on the business need. Let’s take a look at a few of these typical service levels for different replication technologies. Tape vaulting (or offsite tape storage) takes the most time since physical tape needs to be retrieved from a storage facility and tape fed back into the infrastructure. Tape vaulting may be appropriate for small businesses or for remote or branch offices of larger organizations where data may be retained for analytical purposes or for archiving requirements to meet compliance requirements. Full backups are another common practice, but with exponential data growth, it has become a challenge as organizations grapple with how to recovery quickly. Recovery from tape can take days; fortunately, recovery from disk can be done in much fewer than 24 hours—often around 8 to 10 hours or better. Continuous asynchronous disk mirroring provides near zero RPO but still may not be appropriate for those situations with the most stringent recovery requirements. Synchronous disk mirroring replication, which provides zero RPO, may be preferred for business-critical application and data protection in larger enterprises, but is only as good as the last copy. Continuous protection provides the benefits of zero RPO but through point-in-time bookmarks—meaning as good as the last of many copies chosen as the recovery point. Many small-to-midsize companies rely on tape for both replication and disaster recovery—or better yet, on disk-based backup. However, many organizations need the small RPO and RTO delivered by asynchronous and synchronous replication technologies—or continuous protection—to be competitive.