To each his own storage
Agencies need to build backup systems to match their unique storage challenges<@VM>Checklist: Backup software
As agencies adapt storage architectures to cope with double-digit growth in data, they also have more options about how to back up that data.
'Everybody and their mother is coming to market with a backup product,' said Lynn Heiberger, disaster recovery consultant and author of 'The Backup Book.' 'Whether it is a database vendor, storage vendor or an existing backup vendor, they all have a way to protect their systems and data.'
Jerry Wright, technical director of information technology support at the Federal Aviation Administration's Office of Airports (ARP) in Renton, Wash., recently switched the backup software his office was using so he could centrally manage file backup at more than 30 locations in FAA's nine regions.
'It seemed like each region had its own backup scheme,' he said. 'When the administrators in one region were gone, someone else had to try to run their backup software.'
To simplify matters, ARP installed 8T disk arrays from Dell at sites on the East and West coasts last fall and added EMC's Avamar backup software.
Half the ARP locations back up to each array, and the arrays replicate one another so both sites contain a full backup. It is all centrally controlled and automated, freeing local administrators' time for other pressing needs.
'It is not a cheap system ' we spent a couple hundred thousand ' but it is a cost saver,' Wright said. 'There is a saving in time: no tapes to buy, no upgrades to do. And the rack itself has plenty of expansion room.'
Every storage administrator faces a unique backup problem with its own specific solution.
ARP backs up files directly to disk each night, but the U.S. Geological Survey uses Double- Take software for real-time replication at the bit level to SCSI-attached storage. That storage is attached to a CommVault Systems media server, which then runs daily tape backup using a LTO-3, 120-tape, three-drive library from Spectra Logic.
Support for virtualization was a major consideration in selecting this solution, said Erik Hernandez, a systems administrator at USGS in Fort Collins, Colo. 'We are constantly looking at emerging technologies in the areas of storage, virtualization, networking and server infrastructure,' he said. 'When choosing backup software, we want to make sure it can support technologies we are using or thinking about using in the future without having to worry about changing our backup software for each upgrade.'
Backup requirements vary even within a single organization. At NASA's Jet Propulsion Laboratory in Pasadena, Calif., systems architect Tomik Abrahamian supports the Earth Sciences Division and Deep Space flight projects.
He switched from Legato to Arkeia Software's Network Backup to keep backup costs under control on the Earth Sciences Division's 120T Redundant Array of Independent Disks boxes.
'What is so interesting is that the flight projects' backup needs are considerably different than [those] of the Earth Sciences,' he said. 'So backup requirements are highly dependent on the application and environment, especially moving forward when all of the data cannot be backed up because of the sheer volume.'No single solution
Heiberger said it is a mistake to look for a single product to meet all of one's needs. Different types of applications have different recovery-point and recovery-time objectives, and different software is required to meet those objectives. Traditional nightly backup software takes care of catastrophic outages.
'Then you can snapshoot every hour, and that quickly reduces the recovery- point objective from a day to an hour, as well as addressing recovery-time objectives,' she said. 'But snapshots only protect you against data loss or corruption, not hardware failure.
For that, you need mirroring, which means paying for a redundant set of hardware.'
Although disk-based snapshots and replication make it easier to restore data, it isn't easy to completely eliminate tape, said Robert Stevenson, managing director at TheInfoPro, a New York-based consulting firm.
'You need a good classification of how your applications are in terms of their tolerance of risk,' he said. 'Many won't tolerate an environment without tape.'
But even if the data needs to be backed up to tape, it is still good to use disks as an intermediate step.
'If everyone is trying to get things backed up off-hours on a Sunday night, that is a very heavy workload to handle,' Stevenson said.
'Bringing more virtual media servers and virtual tapes into the fold on a Sunday and then staging the data off to real tape throughout the week gives you a way of managing the workload without having to buy the necessary total raw tape drives.'
Heiberger said organizations likely will need to use different types of software to meet different business needs, but she cautioned against going overboard.
'The conundrum for any administrator in this day and age is not to overkill with their backup strategies,' she said. 'If you back it up to tape, back it up to the [storage-area network] and create archive logs, you have created three things that all do the same thing with different recovery-time and recovery-point objectives.'Photo album
USGS runs hourly snapshots on the primary Dell EqualLogic iSCSI storage device. They store two weeks' worth of snapshots, but these never go to tape. Instead, the CommVault software handles the daily, monthly, quarterly and annual tape backups.
Data deduplication, which eliminates redundant data, is becoming widely available to further offset the costs of backing up storage. ARP uses the block-level deduplication that comes with Avamar, so its daily backups include only those blocks that have changed since the previous day. Deduplication can substantially reduce backup costs. Heiberger said one client saw an overall reduction of 20 to 1 with deduplication, although the ratio varied from one application to another. On Microsoft SQL Server databases, which had multiple backups, the ratio was 30 to 1; on Oracle databases with a single backup, the rate was 3 to 1.
'You change your price point from $10,000 per terabyte down to $500 per terabyte because you average a 20X dedup rate,' she said, 'but it ranges all over the board depending on the quantity of backups you are keeping.'
Another factor to look for is encryption, which vendors are including in their backup products. Compliance with Federal Information Processing Standard 140-2 encryption requirements was one reason USGS selected CommVault's backup software.Ahead of the game
'New technologies in many cases surpass the security standards that the federal government has approved or has made the standard, which can be a long process,' said Jeff Schafer, lead IT specialist at USGS in Fort Collins. 'We like to make sure we not only meet the federal standards as a minimum but implement technologies that surpass them in an effort to keep us ahead of the ballgame and keep us secure.'
Keeping ahead of the ballgame is a common theme of storage administrators. With no end to storage growth in sight, you need to plan for expansion when devising your backup strategy.
'While coming up with a backup strategy, consider what you would do if the data volume went up by an order of magnitude in two years and whether your current approach will be able to handle it,' Abrahamian said.
● How much data do you currently have to back up? What is the data growth rate?
● What types of data need to be backed up?
● What are the recovery-point and recovery-time objectives for each type of data?
● What storage architectures does the backup software support ' tape, virtual tape libraries, disk to disk?
● What level of data deduplication does the software offer? Is the deduplication being done with online or near-line storage?
● How is it priced? Flat rate? Per device? Per terabyte?
● Do you need additional technologies, such as snapshots or replication? How does the backup software integrate with these?
● What type of encryption does the software provide?
● What services does the vendor offer for installation and support?
● How are the backups managed? Can a single instance support multiple data domains with separate administrators?
● How many locations do you need to back up? Do you need to back up from workstations or laptop PCs?
● Will this backup solution require additional hardware?