 |
According to a new study from Forrester Research, the rate of adoption for a continuous availability approach is on the rise. A major reason for that uptick is that at many organizations, most disaster recovery architectures and processes look the same today as they did as far back as 20 years ago. This is somewhat befuddling when you consider availability demand – and risk – has increased significantly in that time. Many of the conversations I regularly have with customers support these findings. As a result, IT executives in greater numbers are looking for information about continuous availability, and they are curious about how EMC can help them achieve a continuous availability approach with off-the-shelf products. Here’s a sampling of the recent questions and answers I have exchanged with IT leaders: Question: What is Continuous Availability? Continuous Availability (CA) is a service level created by infrastructure and application architectures that have no single-point-of-failures (SPOFs) at the component, application, or site levels. In a basic CA architecture, all components, application servers, and sites have complete redundancy. As a matter of practicality, applications are deployed with redundant capabilities in at least two data centers. A transaction can be completely processed in either of the two data centers and data coherency is maintained between them. If anything fails (component, server, or site), processing continues and the applications that exploit CA capability continue to run regardless of the failure. Question: How can an application run in Continuous Availability mode? Applications can be deployed in CA mode with commercially available off-the-shelf technology. At the infrastructure layer, data centers need to be setup as Active/Active sites. That is two sites with completely redundant infrastructure- network feeds, network infrastructure, security, LAN, SAN, DNS, and server farms – along with a mechanism such as VPLEX to maintain data coherency between sites. At the application layer, redundant server farms are deployed in each of the sites. Finally a transaction distribution system is deployed to route transactions to the sites and to recognize a site offline condition and route transactions to the other site. Transaction distribution systems are most often deployed as hardware load balancers, although some software mechanisms are used in some sites Question: How does VPLEX Metro enable Continuous Availability? VPLEX Metro in of itself does not create a Continuous Available environment, it proves continuous availability at the storage layer. VPLEX Metro, however, does provide the critical foundation for Continuous Availability by presenting readable/writable virtual volumes in two geographically separated data centers. The volumes can be read from and written to simultaneously by servers in each site. Logically you can think of a VPLEX Metro volume as a RAID-1 mirror set; one side of the set, or “mirror-a” is in one site, and the other side of the mirror, “mirror-b” is in another site. VPLEX’s coherency mechanism ensures that a write to “mirror-a” is synchronously written to the underlying disk in both site-a and site-b; likewise, a write to “mirror-b” is synchronously written to the underlying disk in both site-b and site-a. Question: How does an application or database maintain data coherency across a transaction that involves a series of reads and writes to the same sectors of data? In an environment with multiple readers and writers to the same data, locking mechanisms need to be employed to keep data protected between read and write cycles. If you think about a typical database update, data is first read; it is updated and then rewritten back to the same spot on disk. In a scenario with multiple writers, a lock is used to prevent a writer from inserting data in between the read and write cycle in an update transaction. Several commercial clustered databases such as Oracle RAC have been certified by both EMC and Oracle to operate over VPLEX. Question: How can I leverage Continuous Availability in my environment? There are four basic components in a Continuous Availability Architecture: Active/Active data centers; applications that are built to run in CA mode; a data locking mechanism at the data layer; along with a data coherency mechanism at the storage layer. Question: What is an Active/Active data center? Active/Active data centers are two or more data centers, each provisioned to run production applications with all of the necessary infrastructure to run independently. The data centers are connected by a high-speed data center interconnection and usually have a common name space, authentication mechanisms, and address space. Transactions are typically distributed between the data centers by a global load balancing mechanism that understands which application is running where. The key is that between the two data centers, there is no single point of failure. Question: How are applications built to run in a Continuous Availability mode? In general, applications can be adapted to run in full CA mode by stretching web and app farms along with the database clusters between two sites. The application can be Active/Active if both sites are live and processing transactions for the same application at the same time and the four basic components mentioned above are deployed, A near CA mode can be achieved with 2-site HA failover using server clustering mechanisms such as vSphere HA, MSCS, PowerHA, MC Service Guard, and Veritas Clustering.
Conclusion Organizations have been deploying Active/Active data centers for several years along with a semblance of Active/Active applications. The recent availability of clustered databases over VPLEX Metro and clustered servers though enables true continuous availability up and down the stack and puts it in the hands of the average data center using commercially available off-the-shelf technologies. What once was a promise of the future is a reality today. |
| Update your feed preferences | |
