Virtual Server Technology: Consolidate Your Disaster Recovery Replica Servers

Achieve significant cost savings through consolidation of disaster recovery servers by combining the award-winning disaster protection of CA XOsoft™ Replication (formerly CA XOsoft WANSync) with the latest virtual server technologies from Microsoft and EMC. Consolidating disaster recovery servers on virtual machines can yield both reduced hardware costs and significantly lessened management burden.

For more information, please download either of our white papers on virtual server technology, Consolidation of Disaster Recovery Replica Servers using Virtual Servers and/or VMware Infrastructure and CA XOsoft's High Availability (formerly CA XOsoft WANSyncHA).

Costs of Business Continuance

At the same time, two dynamics are driving up the total cost of maintaining an adequate business continuance capability.

The first dynamic is that the ever-growing numbers of IT-based business processes that are candidates for disaster recovery protection. For example, communication and collaboration systems like email, IP telephony, instant messaging and web portals are becoming increasingly central to revenue-generating business operations, even as increasing numbers of core business processes are becoming IT-based and driving growth in mission-critical database and web servers. Further, the need for redundancy to achieve performance goals, provide fault tolerance, and enable business continuance in the face of disasters is leading to significantly increased numbers of servers that must be managed.

The second dynamic arises from the fact that the complexity of management increases nonlinearly with the total number of systems. Each new server represents not just one additional unit of management work, but introduces from a few to many new interactions within the environment that must be tracked, understood, and managed.

To make things worse, much of this proliferation of systems is not only complex, but is wasteful as well. The typical utilization of an Intel-based server is in the range of 5% -15%. Even accounting for peak demand times, a huge percentage of server hardware, server software, and worst of all, server maintenance and management effort is wasted. A December 2004 Gartner Research report has stated that through 2007, "organizations with more than 200 servers will waste between $500,000 and $720,000 annually supporting underutilized application/server combinations."

Disaster Recovery and Server Consolidation

Server consolidation has become an accepted methodology for both increasing capacity utilization and decreasing management cost. In a December 2003 survey at its annual Data Center Conference, Gartner found that well over 90% of participants planned to undertake a server consolidation effort; two thirds already had projects underway. Server consolidation can provide direct reductions in hardware costs, but can also generate significant management cost savings and yield additional benefits in improved ability to manage security and availability.

Consolidation of systems for DR and business continuance purposes is something of a special case. On the one hand, to provide effective business continuance capabilities while keeping management complexity under control, DR replica systems must match production systems one to one. On the other hand, as systems intended solely to provide backup for production systems, utilization is typically extremely low. While it is not difficult to justify complete duplication of a few key mission-critical servers, applications and data, it is far less obvious when DR protection of the next tier of applications requires duplicating dozens or even hundreds of servers.

Replication and Virtual Server Technology

Virtualization solutions like Microsoft's Virtual Server 2005 and EMC's VMWare introduce an abstraction layer that insulate the BIOS, operating systems and applications on Intel architectures from the physical hardware so that multiple virtual machines, with heterogeneous operating systems, can all run simultaneously on the same physical machine. Each virtual machine, or VM, sees its own consistent, normalized set of virtual hardware (e.g., RAM, CPU, NIC, etc.) upon which an operating system and applications are loaded.

The use of VM technology to consolidate Intel-based servers has exploded over the last couple years. The reason is simple. According to a 2003 Gartner report, "Enterprises that do not leverage virtualization technologies will spend 25% more annually for hardware, software, labor and space for x86-based servers." In fact, given the particular requirements of DR servers, virtualization technology can easily reduce TCO by 50% or more through a combination of hardware savings and savings in a variety of IT tasks like upgrading and monitoring servers.

Implementing the Solution

From the standpoint of CA XOsoft Replication and CA XOsoft™ High Availability (formerly CA XOsoft WANSyncHA), virtual servers are no different than physical servers. As long as the servers are running supported versions of the applications and operating systems, CA XOsoft software installs and runs identically on virtual and physical servers. CA XOsoft Replication and CA XOsoft High Availability have been thoroughly tested both on Virtual Server 2005 and on both the GSX and ESX versions of VMWare.

Example Scenarios

The following figures illustrate the power of combining virtual machine technology and high-availability software like CA XOsoft High Availability. Each of the figures shows a production site on the left, in San Francisco, and a secondary DR site across the country in New York, with production data being replicated over the WAN. The New York site could be a dedicated DR facility or simply another office of the same company.

[enlarge]

Figure 1. Basic DR virtualization. Two servers at the production facility, one running Microsoft SQL Server and a second one running Oracle, are replicated to two virtual servers running on a single physical server at the DR site. The OS and application configurations on the production and virtual machines are identical, but the use of virtual server technology has saved the extra physical hardware.

[enlarge]

Figure 2. DR virtualization of clusters: The production facility is running its Exchange application on a two-server Microsoft cluster as well as a standalone file server. It is of course possible to replicate the cluster to a standalone server, however, if the DR facility is expected to support production usage for an extended period in the event of a disaster, providing cluster capability there as well can make sense. It is also possible to run multiple cluster nodes as virtual servers on a single physical one, but this would eliminate most of the value of running a cluster on the DR site at all. So, in this case, two physical servers are used to contain the two cluster nodes at the DR site, and we are still able to consolidate a DR file server on one of them for some savings.

[enlarge]

Figure 3. The full power of virtualization. Here we illustrate the real power of DR server consolidation using virtual server technology, simply by scaling up exactly the same usages that are shown in the previous two examples. In this case, the production site is running 11 different applications on 14 physical servers, with three of the particularly critical applications running on clusters to increase availability. With powerful servers at the DR site, it is reasonable to run 4-5 applications per server, particularly since each of the servers is running one passive node of a cluster during normal operation. With the proper distribution of applications, we are able to provide DR capability for 14 physical production servers on only 3 physical DR servers and still offer clustering on the DR site with each cluster pair split across physically separate servers.