Business data is the lifeblood of a modern organization. IT’s responsibility is to ensure data is continuously available and can be quickly recovered in case the active copy is damaged or unavailable. That means backing up data belonging to AIX, Linux, UNIX, Windows and, increasingly, data belonging to applications that run under Linux on System z.
An important message emerging in the IBM New Enterprise Data Center (NEDC) initiative is that a data center is no longer defined by the bounds of four walls. IT organizations must recognize their data center team now has responsibility for an interconnected set of critical components that must work together, but which are often geographically separated. No matter where data resides, the requirement remains the same: protect that data by moving a copy from each disparate server to disk or tape, somewhere it will be safe and available for a quick recovery. The challenge is finding the right combination of hardware and software to do that reliably, efficiently, securely, quickly, and economically, while causing the least amount of disruption and preferably with an automated and unattended process.
Centralizing and consolidating the backup process is key to meeting this challenge. Instead of needing to manage backups on each individual and often geographically dispersed server, your staff can back up large numbers of servers— from a single server, as a single job, saving time and aggravation.
All enterprise data protection solutions face limitations (e.g., client operating system disk access, network transport speed, backup server operating system disk/tape access) that can affect their ability to perform consistently. Mainframe systems have been centralizing and consolidating backup for many years. Today, there are significant advantages to centralizing enterprise open systems backup and administration by employing a backup server running on an IBM System z.
Specifically, the strengths of the IBM mainframe are its well-proven reliability, consistent operational integrity, superior reporting, and unassailable security.
Data protection applications running on a System z take advantage of these well-documented strengths to provide a flexible, efficient, high-performing backup/ restore solution. Additionally, employing the mainframe for open systems data protection helps an organization leverage the investments it has already made in a disciplined, well-trained staff as well as existing disk and tape resources.
The common advantages of z/OS-resident enterprise data protection solutions are that they provide a predictable, reliable, secure backup and recovery. There are several ways to provision a z/OS environment to achieve these cost-efficiency and elapsed time savings.
Three Different Transport Mechanisms
The System z environment supports three transport mechanisms for an enterprise data protection solution to move data from an open system disk to a mainframe backup server:
• On-network transporting data across a TCP/IP communication network
• Off-network channel gateway data transport
• Off-network storage controller-based data transport.
On-Network Approach
Traditionally, the on-network System z approach is a client/server model that transports both control information (metadata) and backup data over a conventional TCP communications network (see Figure 1). The primary advantage of backup with an on-network enterprise data protection solution is that typically the network is already there. Using TCP/IP networks for large bulk data transfer, however, has some notable disadvantages.
Compared to traditional disk-to-tape backup, longer elapse time and lower performance may be the first noticeable disadvantage. The typical speed of most communications networks today is 1Gbit/second with some network segments still running at only 100Mbits/ second. The maximum aggregate data transfer capacity of a 1Gbit network is approximately 100MB/second. If the quantity of backup data is relatively small, and there’s no critical or little other network traffic, network transfer speeds may be sufficient. Backing up larger amounts of data, however, will require a transport that can support the high aggregate data rates resulting from concurrent parallel backups.
Considering the maximum theoretical network speed, the minimum time to transfer 1TB of data over an idle 1Gbit network is about three hours. Backup, however, as anyone familiar with on-network data transfer will acknowledge, runs at considerably less than the maximum theoretical network speed. It’s always constrained by the aggregate network capacity, even after eliminating other choke points (i.e., disk and tape transfer speed). Employing data reduction and de-duplication techniques to reduce the amount of data is especially important to on-network solutions, but these techniques benefit off-network solutions, too.
Another drawback to TCP/IP data transfer is its high CPU resource requirement. Studies show that the z/OS CPU capacity required to achieve a TCP/IP communication rate of 100 MB/second is more than 100 MIPS (http://regions.cmg.org/regions/phcmg/ May08BYeager.pdf). Heavy use of communications networks for large, bulk data transfer will likely also have an adverse impact on other communications network traffic workloads. The other workloads, which most likely are higher priority business communications, will without exception, experience sluggish, inconsistent, or otherwise unacceptable network response times.
Virtual On-Network Data Transport
The disadvantages of using TCP/IP networks for large bulk data traffic apply to all operating system platforms. However, what if there was a way you could use an existing TCP/IP backup solution without putting the data on the physical network? That’s now possible with HyperSockets, a unique Linux on System z data transport technology not available to non-mainframe (i.e., open systems), on-network backup solutions.
HyperSockets is a System z virtualization technology for Linux on System z that provides a low-cost, high-performance virtual networking transport for high-speed data transfer via a System z hardware memory bus (see Figure 2). Though it will still employ the TCP/IP protocol, a virtual on-network Hyper- Sockets data protection solution can move larger amounts of data to a System z enterprise backup server over cross-memory, HyperSockets bus connections more efficiently than a physical TCP communications network.
Off-Network Solutions
What if you could transport data to the System z backup server without putting it on the physical network at all? Off-network solutions can efficiently move large quantities of enterprise backup data from multiple, geographically distributed open systems servers to a z/OS backup server over Fibre Channel Protocol (FCP) and FICON channels (see Figure 3).
An off-network solution, which moves backup data directly over the enterprise FICON/FCP Storage Area Network (SAN) infrastructure, is especially attractive because data is directly shipped to z/OS in large chunks for efficient storage on mainframe disk and tape. The advantages of removing backup or restore data transfer from the communication network are:
• Reduced overall network traffic
• Megabyte FICON channel access to open system data rather than megabit network performance
• Concurrent parallelism of the massively scalable System z I/O channel system (up to 336 FICON channels).
Not transporting backup data on the communications network makes for a solution that’s extremely reliable, efficient, and can employ additional FCP/FICON channels. This allows for more parallelism, which results in higher aggregate data transfer rates that can effectively reduce overall backup time requirements. Reducing CPU resource use is an additional benefit of taking backup data traffic off the TCP/IP network. Here are two specific off-network solutions:
Channel Gateway Data Transport
Current versions of enterprise data protection solutions already include support for a channel gateway appliance that lets them transparently employ a superior off-network transport mechanism for moving data to a z/OS mainframe- resident backup server.
Channel gateway data transport technology lets an enterprise data protection solution use an organization’s existing FICON/FCP SAN infrastructure to deliver the same benefits of a consolidated/ centralized enterprise data protection solution with no network data transfer, resulting in higher throughput and lower CPU consumption.
Typically, a single channel gateway appliance can sustain aggregate transfer rates of 250MB/second. This is 150 percent faster than a 1Gbit communication network and provides ample bandwidth for multiple, concurrent backup streams. Additional channel gateway appliance units can be added to provide more bandwidth in 250-plus MB/second increments. Increasing bandwidth supports additional multiple concurrent backups, increasing parallelism and reducing the aggregate elapsed time dedicated to backup activities.
Storage Controller-Based Data Transfer
Storage controller-based data protection solutions employ storage system technology features that allow z/OS access to AIX, Linux, UNIX, Windows and Linux on System z data by providing an enterprise data protection application with extremely efficient direct access across FICON channels to FCP open system volumes and the data they contain (see Figure 4).
Storage controller-based data transfer presents an opportunity for z/OS-resident data protection solutions to bypass many of the constraints a TCP/IP client/server model imposes. Instead of TCP/IP communication network protocols, an enterprise data protection application employing storage controller-based data transfer technology can use direct, high-speed, cross-platform FICON channel access to back up FCP open system disk to z/OS mainframe tape or disk. Direct access to open system disk volumes means the same principles that make physical volume backup and logical file restore the hallmark of z/OS mainframe data protection can apply to open system backup restore and provide for direct application of non-disruptive backup capabilities.
The key benefit of direct FICON (up to 4Gbps or 400MB/sec) access is ultra high-speed channel performance. This directly translates to high-speed backup and restore of open systems data with mainframe reliability and RACF security control.
Combining the same storage controller technology that allows heterogeneous, cross-platform access for direct System z physical backup with a client/ server model creates the opportunity for an all-encompassing enterprise storage data protection solution. A z/OS enterprise data protection solution, providing incremental backup, logical file restore and central administration combined with policy-based management facilities for local, direct-attach disk storage, can extend logical file granularity and efficient, high-speed data protection to geographically distant, system-attached storage as well as Network-Attached Storage (NAS).
Instead of transmitting data across the network, in a storage controller-based data protection solution, an enterprise data protection client can transport data by writing and reading backup data across a Fibre SAN connection to a dually accessible local backup disk, in a storage system controller attached by a FICON channel to the System z. The z/OS enterprise backup server reads the backup data from the local backup disk, across the FICON channel, and records it on z/OS-managed tape or disk. The backup client running on the open system platform maintains a command and control conversation with the z/OS-resident backup server on the mainframe by exchanging metadata across a TCP/IP communications network link.
Using a z/VM host system to virtualize z/OS access to the open systems volumes employing direct z/VM Fibre Channel protocol access to open system disk is another promising variation of a storage controller-based data protection solution; that doesn’t require special storage controller technology.
Conclusion
As enterprises look to meet the challenge of providing data protection and business continuity for their rapidly growing open system storage, many are finding significant advantages to centralizing data protection for enterprise open systems on System z.
All enterprise data protection solutions face limitations. On-network solutions transporting data with a TCP/IP protocol are likely to hit some type of transport limit, such as network speed or network capacity, and are less able to deliver consistent outcomes.
Employing off-network System z technology to consolidate enterprise open systems data protection and business continuity (backup/restore and administration) yields an extremely reliable and efficient solution. A solution that allows for more parallelism, resulting in higher aggregate data transfer rates that can repeatedly significantly shorten open systems backup times.
There are additional significant advantages to centralizing enterprise open system backup and administration with a backup server running on the System z. Organizations choosing an IBM System z-resident z/OS data protection (backup/restore and administration) solution to protect their open systems environments may also benefit from special IBM System z New Application License Charges (zNALC). zNALC offer a reduction in price for the z/OS operating system software running a qualified new workload application as well as similar money-saving offerings on System z10 Business Class processors in conjunction with IBM TS1130 tape and DS8000 disk systems.
