Lightwave - Enabling CDP: The new dimension in remote storage management

While government regulations and competitive pressures have made implementing continuous data protection and other storage services a business imperative, innovations in WDM and optical networking have made it cost effective.

A more keen awareness of the high, varied costs of network downtime; an increase in government regulations mandating data security and disaster recovery; and a determination to locate primary and recovery data centers farther apart are driving innovation in network server and storage capabilities, including continuous data protection (CDP). Advances in WDM and other technologies have enabled a wider array of enterprises to cost-effectively and securely implement more sophisticated storage networking capabilities, like CDP, over networks of larger scope and distances.

Lost revenues, staff idleness, and reputation damage can result when an enterprise experiences significant periods of application or network unavailability. A large financial enterprise alone could incur revenue losses of more than a million dollars per hour of network downtime. As enterprises grow increasingly aware of these business vulnerabilities, they seek a greater level of protection to ensure 24-hours-a-day/seven-days-a-week, high-performance, high-availability network operations.

At the same time, enterprise Information Technology (IT) managers and chief information officers (CIOs) are deluged on the regulatory front. The Sarbanes-Oxley Act, Graham-Leach-Bliley Act, the Health Insurance Portability and Accountability Act (HIPAA), California's Information Privacy Act, and other legislation have made it more critical than ever for enterprises to beef up their network contingency plans and security.

Moreover, recent disasters reveal how a single event can jeopardize an entire metropolitan area. Enterprises no longer are satisfied with deploying a backup facility merely in the next block; they seek to leverage sites hundreds and sometimes thousands of kilometers away.

Increased competition, more stringent regulations, and the need to locate data centers farther apart have driven the rapid development of new capabilities, such as IBM's Geographically Dispersed Parallel Sysplex, Server Time Protocol (GDPS STP), which enables enterprises to adopt a distributed model at extended distances for mainframe server clustering. CDP is intended to combine the characteristics of high availability and near-continuous operations to deliver high levels of service. It is based on geographical server clusters and data mirroring, including the latest 4- and 10-Gbit/sec Inter Switch Link (ISL) technologies.

Often incorporated into GDPS STP environments, CDP uses disk technology to capture and time-stamp data updates on a frequent basis. With CDP, the enterprise effectively develops a log of changes to its data from which it can recover operations. For example, an enterprise using CDP could revert to the most recent non-corrupted version of its data before a virus infected the network infrastructure.

CDP is especially valuable for applications that are mission-critical, in which data changes rapidly or in which corruption and other logical data issues frequently occur. In addition, CDP dramatically changes the way enterprises think about recovery point objectives (RPOs) and recovery time objectives (RTOs).

An application's RPO defines the amount of data that the enterprise deems acceptable to have lost once the application recovers to operational. The RTO refers to how quickly an application is restored after an outage. An enterprise assigns an RPO and/or RTO of seconds, minutes, hours, or days to a given application, depending on how frequently the data changes and how critical those changes are; the more frequent or mission-critical the changes, the shorter the RPO/RTO. The differentiating characteristic of CDP is the improvement it offers enterprises in the area of RPO/RTO and data restore.

Some CDP or "near-CDP" technologies take data snapshots at frequent, prescribed intervals. Other CDP products simultaneously write data to a disk and a second location; instead of being limited to the prescribed intervals, the enterprise can access data from any point in time. In either form, CDP enables recovery capabilities (and shorter RPOs/RTOs) far more advanced than the existing method of backing up data overnight or during regular periods of low user activity on the network.

Today, an enterprise can rely on Internet Protocol (IP), SONET/SDH, and WDM-enhanced optical networks to support storage applications and transport storage traffic among distributed data centers, and each are optimized for certain applications. Widely deployed and familiar to enterprise IT managers, IP is a cost-effective and simple choice for a remote tape vault, but it is not ideal for use in real-time geographic server clusters. And circuit-based SONET/SDH can be utilized for a disaster recovery solution in which ESCON, FICON, and Fibre Channel applications are mirrored among data centers if the enterprise has less demanding bandwidth requirements and RPO/RTO expectations. Only WDM networks, by contrast, provide the most robust capabilities for meeting the shortest RPOs/RTOs and supporting runtime-sensitive protocols like IBM's GDPS STP.

Consider a 60-terabye data recovery. It requires 45 days or more over a single OC-3 connection, but only 10 minutes with an 80-channel, 10-Gbit/sec, carrier-class DWDM system.

WDM creates "virtual channels" across optical fiber networks, each transporting its own application of a given protocol, such as Ethernet 10/100/1000/10G, ESCON, FICON, Fibre Channel (1, 2, 4 and 10G), Coupling Link, Sysplex Timer/Server Time Protocol (STP), ATM, and SONET/SDH. At one end of an optical link, the traffic is multiplexed and, at the other, de-multiplexed. There is no performance degradation for the applications.

"A lot of the disaster recovery networks that we are being asked to look at today wouldn't be economically practical or technologically feasible without WDM," explains Dr. Casimer DeCusatis, distinguished engineer of IBM's System and Technology Group. "The cost of deploying WDM over dark fiber today is about the same as a typical OC-3 link. Plus, the enterprise gets protocol independence, a high degree of fault tolerance, and plenty of channel and bandwidth capacity for deploying additional services," he adds.

Factors such as bandwidth, service, and cost requirements will weigh into an enterprise's decision to deploy highest-capacity DWDM, lowest-cost Coarse WDM (CWDM), or hybrid C/DWDM. An enterprise could deploy CWDM for support of a limited number of services along point-to-point links or in rings with link spans not exceeding 50 miles. To support more services over longer distances and to provide for maximum system scalability, DWDM is the ideal choice. Hybrid CWDM/DWDM platforms are valuable because they enable an enterprise to deploy only a few channels today and affordably add channels as business needs dictate. Initial installation cost is less than that of a full DWDM platform.

In fact, many enterprises will choose to employ more than one of the transport options and strategically apply different storage media to different infrastructures. Cost issues will weigh in these decisions, as well as the access options available at each enterprise location, the particular storage applications to be supported, and the distances to be covered.

For example, the supportable distance for transporting synchronous storage applications such as GDPS without the need for intermediate amplifiers or repeaters today is 100 kilometers. Asynchronous storage applications such as CDP, meanwhile, can be carried across network links reaching thousands of kilometers.

Most enterprises must cost-effectively and strategically mix and match applications and connection media. A New York-based financial service provider, for example, might connect its Manhattan headquarters with a secondary site in New Jersey via DWDM-enhanced dark fiber in a GDPS STP geographical clustering environment that protects its millions of transactions per second. GDPS STP, a synchronous mirroring application, guarantees no lost data anywhere in the "sysplex." An asynchronous CDP application might be transported to a third site, via Fibre Channel over IP, more than 1,000 kilometers away from the metropolitan area. In this type of three-site mirror service, an enterprise positions itself for rapid, comprehensive recovery from the failure of a single data center and bankruptcy avoidance in the event of regional network collapse.

Eliminating intermediate amplifier/repeater sites--or mid-span huts--along an optical-fiber link not only reduces an enterprise's cost and management complexity, but also minimizes security risks.

This is an important point for enterprises, particularly in light of the new regulatory pressures they face. Some of the new regulations stipulate that even the individual executives (in addition to the enterprise itself) can be penalized stiffly for security violations. IT managers and CIOs today must demonstrate reasonable and acceptable due diligence that data is protected from rogue access in the data center or while in transit for mirroring from one site to another. Of course, the risk here is not limited to government penalty. When information threat reveals the vulnerability of its network infrastructure, an enterprise can sustain a devastating blow to customer confidence and loyalty.

In addition to the elimination of mid-span huts, enterprises have found physical-layer intrusion detection and in-flight data encryption to be valuable methods of securing CDP and other storage applications. In physical-layer intrusion detection, a WDM optical networking platform would be configured to take a pre-programmed automated course of action (shutting down service on a primary fiber and switching to a backup, for example) in the event that a threshold of signal degradation is passed. During in-flight data encryption, storage data is encoded before traversing the SAN.

Innovation in storage capabilities has accelerated as enterprises face heightened regulatory and competitive pressures and become more aware of their businesses' vulnerability to disasters and network downtime. Enterprises have benefited from the emergence of powerful applications such as CDP and GDPS STP, which, in their most advanced implementations, enable a company to revert to data from any point in time.

Improvements in data transport have given enterprises unprecedented options in cost-effectively and securely connecting primary and recovery data centers and supporting increasingly aggressive RPOs and RTOs. Consequently, the community of enterprises deploying sophisticated storage networking capabilities--from disaster recovery, to business continuity, to CDP--is growing more and more diverse.

Todd Bundy is the director of Enterprise Business Development and Alliances at ADVA Optical Networking.

Geographically Dispersed Parallel Sysplex (GDPS) is a registered trademark of IBM.

Add RSS Feed