Zero Data Loss Solutions for Data Center Consolidation White Paper October 2010
Summary Driven by a need for greater operational efficiency, data-center consolidation will continue to be a top priority for companies during this decade. Companies that have acquired or merged with other companies will look to reduce the number of data centers, consolidating into fewer, geographically dispersed super-centers. At the same time, companies that have allowed individual business units or departments to operate their own data centers will look to consolidate these under centralized management to ensure good corporate governance, to comply with laws and regulations, and to lower operational costs. Lower labor cost and business-friendly governments will drive many companies to locate super-centers in emerging economic regions, even when network infrastructure and power are substandard, and where civil and governmental turmoil increases the risk of a temporary disruption to operations. In these scenarios, enterprise data recording (EDR), a zero-data-loss storage solution, becomes increasingly valuable to corporations, simultaneously enabling the company to locate facilities in lower-cost areas, while protecting the company against data loss, through a broad range of man-made and natural disasters. Data Center Sprawl Growth of Data Centers through Acquisitions Over the past decade there has been a massive consolidation of companies across many industries, including banking, auto, technology, and consumer products. Supporting the operations of each of these companies were one or more data centers, and if these data centers were allowed to continue operations, the acquiring company would experience data-center sprawl. As an example, if Microsoft continued to operate the data centers of its acquired companies during the past decade, Microsoft would have more than 100 additional data centers.
Growth of Business Unit and Departmental Data Centers Even companies that have not been acquisitive may find themselves with a growing number of data centers. Departments, especially smaller ones, often became frustrated by the failure of centralized IT to prioritize their requirements. The dramatic reduction in server costs and the availability of appliance-like applications in a box, enabled many business units to acquire systems on their own and operate initial server deployments out of little more than a network closet. Drivers of Consolidation Economic and Operational Efficiency With almost every acquisition comes the promise of greater efficiency. Through consolidation and standardization of processes, companies promise to deliver more of the above-the-line revenue to bottom-line profit, by eliminating redundant processes, people, and infrastructure. Given the significant role that IT plays in many of today s data-centric businesses, one of the logical places to look to eliminate waste is the corporate data center. Legal and Regulatory Compliance Compliance with laws and regulations has become increasingly difficult for companies. One way in which companies have sought to deal with this difficulty is by centralizing authority across the company and assigning responsibilities to specialists. For example, it is common today to have specialists such as the Chief Security Office (CSO) and the Chief Compliance Officer (CCO), in addition to the Chief Information Officer (CIO). This centralization of responsibility enables the CCO, as an example, to ensure that IT processes across the entire organization are in compliance with laws and regulations. And this centralization of responsibility is simpler to enforce when operations are consolidated into fewer, larger locations.
Enablers of Consolidation Server Virtualization The dramatic improvements in server price and performance, when combined with server virtualization technology, such as VMware s ESX and Microsoft s HyperV, have enabled companies to consolidate multiple applications onto a single server. Virtualization has also helped IT departments become more responsive to infrastructure and application requests by departments and business units. Server and application deployment time can be dramatically reduced using server virtualization. Tools for centralized management of virtual servers are also enabling companies to reduce the personnel costs associated with IT infrastructure management. Replication and Failover Despite the improvements in price and performance, together with increased efficiency from centralized management, consolidation would still be considered too risky for many companies, had the industry not also achieved dramatic improvements in data-replication and application-failover capabilities. Today, leading companies leverage volume snapshot, open-file-backup, synchronous and asynchronous replication technologies, together with local clustering technology and wide-area-failover capabilities at the application and server layer, to ensure that data is not lost and that applications survive hardware failures and routine maintenance, while migrating or failing over to another server in a non-disruptive way. Types of Consolidation While there are many ways to approach data center consolidation, from an application and topology perspective, most companies will fall into one of the following, three scenarios: Two-Data-Center Synchronous, Two-Data-Center Asynchronous or Three-Data-Center Synchronous-Asynchronous.
Two Data Center Synchronous The two data center, synchronous topology is often used when application data loss is viewed as exposing the company or the company s customers to significant risk. Given the impact of typical synchronous replication technology on application performance, most companies that leverage this topology will locate the data centers within 40km of each other. This provides acceptable application performance for update-intensive transaction processing applications, while providing some measure of protection against catastrophic failure from the loss of two data centers. It will not, however, ensure application availability and recoverability given certain disaster scenarios, including man-made disasters, such as war and terrorism, or certain physical disasters, such as Category 5 hurricanes and massive earthquakes. In addition, this topology will not enable a company to take advantage of lower-cost labor and operational costs found in many developing countries. As a result, many companies that opt for this topology are regional companies that would not operationally survive a regional disaster anyway and which are not covered by regulations requiring a more-robust disaster recovery capability. Two Data Center Asynchronous The two data center, asynchronous topology is most often used when the potential for loss of recent updates to application data is viewed by the company and its customers as less critical than the ability to survive a regional disaster and restore operations. Leveraging asynchronous replication technology, the company can locate a second data center well outside any foreseeable regional disaster zone and take advantage of lower-cost labor in developing countries. Because the replication technology is asynchronous, some data will assuredly be lost in a disaster. The amount of data lost will depend on the frequency of transmission of the updates to the remote site, which can vary from minutes to hours. Companies that opt for this approach typically maintain business operations in more than one region, and may also have business reasons, including local government regulations or incentives, for locating one data center in a particular region.
Three Data Center The three data center, asynchronous-synchronous topology combines the benefits, and, unfortunately, most of the costs of the previous topologies. In this topology, write operations are synchronously replicated to a second, local data center, again, typically located within 40km of the primary data center. A third data center, typically located outside of the local region, receives asynchronous updates and, assuming the proper choice of asynchronous replication technology, enables an application-consistent recovery of data. In this topology, a disaster that is localized to a single data center will not cause data loss or a disruption to operations. A regional disaster that affects the two synchronous sites will result in data loss, but will not prevent a resumption of IT operations. Risks with Consolidation When all applications and IT infrastructure reside in a single data center on shared servers and storage systems, companies can achieve substantial efficiency. These efficiencies come through optimization of server and storage utilization, centralized management, and reduced power and cooling costs. However, this approach also exposes the company to enormous, catastrophic risk, as the company, in a disaster, may lose critical corporate data across the full range of applications. Companies faced with data center topology choices must balance very real budgetary constraints against the risk and probability of data loss. In general, the more consolidated applications are, the more risk of a catastrophic site disaster, and thus, the greater the need for a two-data-center or three-data-center topology. In addition, the lower the tolerance for data loss, the greater the need for a higher-cost, three-data-center topology. Until recently, companies chose three-data-center topologies, when laws and regulations required it, and two-data-center topologies, when they did not. Even with companies that chose a three-data-center topology, it was, frankly, too expensive for most companies to afford across the full range of applications. So ultimately companies adopted a combination of replication approaches depending upon the risk profile of the application, the regulations, and the value of the data.
An Affordable Zero-Data-Loss Solution for Consolidated Data Centers With the advent of Enterprise Data Recording (EDR), companies can now leverage the combined advantages of two and three-data-center topologies without incurring the expense of a three-data-center topology. EDR stores recent write operations in a hardened storage system, capable of surviving the most extreme disaster scenarios, much like those encountered by an airplane Flight Data Recorder. These disasters include fire, flood, smoke, building collapse, and earthquakes. The first commercially available EDR is Axxana s Phoenix System RP, which was codeveloped with EMC and integrates with EMC s RecoverPoint. RecoverPoint provides the application-consistent, asynchronous replication to a remote site, leveraging low-cost asynchronous infrastructure. At the same time, the Phoenix System RP stores recent right operations in a disaster-proof enclosure that, in the event of a disaster, can automatically transmit encrypted updates to the remote site leveraging the cellular network. Companies utilizing the Phoenix System RP can choose to locate the system in the primary data center or nearby in a second, substantially smaller secure site, thus avoiding the cost of a full, local synchronous data center. Axxana s implementation of EDR ensures that companies will not lose data, that the secondary site can be located virtually anywhere in the world, and that the cost will be little more than is required using long-distance, asynchronous replication today. By using the Phoenix System RP, companies can gain the full benefits of data center consolidation, while affordably protecting all critical corporate data across the full range of applications.