By Jack Fegreus
Recent surveys of backup practices at mid-sized businesses show a surprising lack of concern for performance or time constraints. Nonetheless, these administrators have considerable concern for backup procedures, or the lack of such procedures. Their worries center on complexity, which hinders implementation. To solve that problem IBM developed Tivoli Continuous Data Protection (CDP) for Files software, which is specifically targeted at end-user computers such as laptops and workstations.
In a world obsessed with litigation and regulations, leaving data on desktop systems vulnerable is not a viable option. Unfortunately, there are many who believe that running an anti-virus program is all that is necessary on the desktop. Reality dictates, however, that data-destroying virus attacks are rare events compared to losing data because of inadvertent deletion or alteration of a file by a user, data corruption, or normal device failure. Anti-virus software does nothing to help those scenarios. Virus scanning software does, however, provide an excellent paradigm for providing the next generation of backup or what’s dubbed “continuous data protection” (CDP).
In concept and in implementation, Tivoli CDP for Files is what business consultants call a “strategic look-alike” for anti-virus software. The model is simple: Low-cost software is installed on every desktop or laptop computer. This software involves minimal configuration, and after the initial setup the software just runs in the background.
To examine how well IBM handles this model of set-and-forget software, we installed Tivoli CDP for Files on a high-end Pentium 4 workstation with hardware-based Ultra320 SCSI RAID, Gigabit Ethernet connectivity, and Windows XP PRO. On this system, the Windows My Documents folder is mapped to its own 25GB partition, which is used to store data files for currently active projects. A NAS volume is used for disk-to-disk (D2D) backups. This volume is imported from a Windows 2003 Server, which is used to export SAN-based arrays to desktop systems.
The first hurdle for IBM in meeting this model is cost. IBM cleared that hurdle by pricing the software at $35 per system. To make that happen, Big Blue teamed up with Digital River to deliver Tivoli CDP for Files over oneNetwork (http://onenetwork.digitalriver.com/) to end users who never dreamed of using down-loadable IBM software.
More importantly, at $35 per system, Tivoli CDP for Files provides an affordable D2D solution for every business scenario, from SOHOs to global enterprises. In SOHO scenarios, a high-capacity, USB-attached Serial ATA (SATA) disk drive will likely be the target for the backup data. In larger business environments, a NAS or iSCSI system that can be easily accessed by a server for a more traditional backup solution will likely be the destination device. In large enterprises, Tivoli CDP for Files can directly target a server running Tivoli Storage Manager without any need for TSM agent software.
The next two hurdles for IBM were easy installation and configuration followed by continuous unobtrusive performance in the background. With the initial goal in mind to back up the entire population of vulnerable desktops and laptops in an enterprise, the software had to be able to scale into the tens of thousands of clients. That kind of performance put it out of the client/server category. How could such a system scale? The answer was simple: Each client was on its own. As with a virus-scanning package, policies are set on a file basis locally (although those policies can be cloned and distributed to all of the systems).
As for installation and configuration, there is little to do. While this is sure to please SOHO users, the dearth of options is likely to leave a systems manager in a large IT shop wondering if anything this simple can actually work.
For “continuous protection,” it is necessary to identify a local and/or remote (LAN-based) disk target. Next, the files or directories to be protected must be identified. That should be all that’s needed for continuous backup; however, there is one undocumented problem that requires one extra step.
Within Tivoli CDP for Files there is a restore interface that is easy to use, but it is only required if using Tivoli Storage Manager as the target repository.
It will also be necessary to set up “scheduled protection” for these same files and directories. In our lab, we set up scheduled protection (which can be daily, weekly, or monthly) on an internal ATA drive that is used to locally archive seldom-used data, such as downloaded software CD distributions for testing. At issue for the files under continuous protection are files that have been modified and saved within applications in a way that is masked from the Tivoli CDP software. To resolve that problem, a backup of all files on a scheduled interval-we chose daily-is all that is necessary. There is neither a need nor a facility to schedule the backup any more precisely than daily. On the appropriate day, the application will choose a quiet backup time.
In addition to what should be backed up, an advanced configuration screen allows the user to set up exception rules for what not to back up. For example, directories such as caches, system volume information, and recycle bins can be excluded from the backup operations. In addition, performance parameters that put a cap on network bandwidth and the maximum size of files that will be backed up can be set. Finally, limits can be set to restrict the amount of space that can be consumed on the target drive and also on the maximum number of versions to save for each file.
At this point Tivoli CDP for Files fades into the background. As new files are created and added to directories that are under continuous protection, duplicate files are added on the target device. This is done asynchronously within a few seconds following the creation of the original. As a result, local fast-access devices, such as the RAID arrays on our test system, are not slowed down by a file repository that is built on slower drives with higher capacity. In addition, should the target drive be inaccessible, the data will be queued and writing will be attempted again later.
Tivoli CDP for Files keeps all the protected file copies in their native format in a directory tree that parallels the original tree. As a result, the copies are always directly accessible by the application with which the originals were created. Nonetheless, problems do occur within certain popular applications when editing files.
For example, when a file is edited and saved within Word using Office XP, those changes are not written to a new file in the backup archive. Similarly, changes to a file using Adobe’s Photoshop CS2 will not be reflected; however, in the case of Photoshop the creation date of the protected original copy will be changed to reflect the time that the file was modified. These issues are easily remedied by setting up a daily scheduled backup. When that backup occurs, a new copy is created in the protected directory that reflects the current state of the file. As a result, at the start of each day, all files in the host and target directories are in synch. At any time of the day, the user can go to a file that is in the initial state for the start of that day or the initial state in which it was created on that day.
While Tivoli CDP for Files provides tools and views to see and restore the saved copies, those tools are not needed to manipulate the copied files in any manner. This is in contrast to traditional backup software that re-packages the protected files into archives managed like a database. This simplicity is critical for small businesses, where the number of sites that have no formal backup procedures in place is striking. For example, in an end-user survey conducted by Imation, 19% of the sites reported having no formal backup procedures in place, and at sites with 51 to 100 employees, one of four sites did not have formal backup procedures in place.
Clearly, Tivoli Continuous Data Protection for Files is not an “old-school” backup package (although it can be easily integrated with one). Combining simplicity with CDP may enable IBM to carve out new market opportunities.;
openBench Labs Scenario
Continuous D2D backup
WHAT WE TESTED
IBM Tivoli Continuous Data Protection for Files
HOW WE TESTED
- Dell Dimension 4550 workstation
- Windows XP Pro
- UltraSCSI 320 RAID drives
- Gigabit Ethernet
- HP ProLiant DL580 G3 Server
- Windows 2003 Server
- QLogic SANblade QLA2340 HBA
- Gigabit Ethernet
- Very simple interface with minimal settings
- The inability to track all file changes made within all applications requires setting both continuous and scheduled file protection on the same set of files for full coverage.
- The number of file versions to maintain, the maximum size of the backup repository, along with file exclusion policies for backup, can be set by the user.
- Copied files are placed in a directory tree in their native application format and can be accessed through the standard Windows point-and-click interface.
- Setting both continuous and daily backups along with a limit of two versions on a set of files creates a backup repository that contains restorable images of all of the files at the start of both the previous and the current workdays.