Ways to Avoid a Data-Storage Disaster
Total Page：16
File Type：pdf, Size：1020Kb
TOOLBOX WAYS TO AVOID A DATA-STORAGE DISASTER Hard-drive failures are inevitable, but data loss doesn’t have to be. ILLUSTRATION BY THE PROJECT TWINS THE PROJECT BY ILLUSTRATION BY JEFFREY M. PERKEL was to “clean up the data and get organized”, advantages, and scientists must discover what she says. Instead, she deleted her entire pro- works best for them on the basis of the nature racy Teal was a graduate student when ject. And unlike the safety net offered by the and volume of their data, storage-resource she executed what should have been a Windows and Macintosh operating-system availability and concerns about data privacy. routine command in her Unix termi- trash cans, there’s no way of recovering from In Teal’s case, automation saved the day. Tnal: rm −rf *. That command instructs the executing rm. Unless you have a backup. The server on which she was working was computer to delete everything in the current In the digital world, backing up data is regularly backed up to tape, and the “very directory recursively, including all subdirec- essential, whether those data are smart- friendly and helpful IT folks” on her depart- tories. There was just one problem — she was phone selfies or massive genome-sequencing ment’s life-sciences computing helpdesk were in the wrong directory. data sets. Storage media are fragile, and they able to recover her files. But the situation was At the time, Teal was studying computa- inevitably fail — or are lost, stolen or damaged. particularly embarrassing, she says, because tional linguistics as part of a master’s degree Backup options range from USB memory Teal — who is now executive director at The in biology at the University of California, sticks and cloud-based data-storage services Carpentries, a non-profit organization in San Los Angeles. She had spent months develop- to huge institutional magnetic-tape servers, Francisco, California, that runs workshops ing and running simulation software, and was with researchers typically exploiting more than on scientific computing — had previously at last ready to begin her analysis. The first step one. But not all such strategies have the same worked for the information technology ©2019 Spri nger Nature Li mited. All ri ghts reserved. 4 APRIL 2019 | VOL 568 | NATURE | 131 TOOLBOX (IT) team. It was “like the lifeguard having for data backup, but only one is approved for gave in tip 1, Cobb says: “So, 3-2-1 backup, and to be rescued”, she says. use with sensitive data. Your department’s then restore [some key files]. And test it on a Here are 11 tips that could make potential IT team can offer advice. “Being out of com- different computer, in a different room, on a data-loss disasters a little less painful. pliance for data protection can be very serious. different device — because if the worst-case You could face financial penalties, or lose the scenario happens, you won’t have your device.” 1. Apply the 3-2-1 rule. The rule of thumb ability to conduct research,” Wickes says. to follow when making data backups, says 9. Expect the unexpected. Life happens. Michael Cobb, director of engineering at 5. Automate backing up. When making back- Cobb — who lost all of his personal posses- DriveSavers, a data-recovery firm in Novato, ups, automation is key. Kelly Smith, a cardiac sions in a wildfire in 2017 — had a client who California, is ‘3-2-1’: “It’s three copies, [on] geneticist at the University of Queensland stored a rack of 96 hard disks underneath a two different media, one off-site.” You might, in Brisbane, Australia, has access to a shared fire-control sprinkler. One day, the sprinkler for instance, maintain copies on your per- network drive that is copied to tape. She used popped, and the disks were inundated with sonal computer, an external hard disk and to move her files to the drive manually, but water. “None of that data was backed up,” he the cloud-based file-synchronization service only monthly; in the event that the drive failed, says. Leslie Vosshall, a neurobiologist at the Dropbox (US$12.50 per user each month, for newer files could be lost. An automated cloud- Rockefeller University in New York City, almost 3 or more users and 3 terabytes of storage). based backup system called Druva inSync, lost her mosquito-genome sequencing data in “This is a rubric to take inspiration from, not from data-protection firm Druva in Sunnyvale, 2012 when her basement servers were flooded a law,” notes Elizabeth Wickes, an information California, now obviates that concern. “It’s one in the wake of Hurricane Sandy. Such events are scientist at the University of Illinois at Urbana– less thing I have to worry about,” she says. unavoidable but can often be anticipated — so Champaign — precious data might require “You have to not think about it,” explains search hard for vulnerabilities. About a year- extra precautions. Teal. “Because when you’re most stressed and-a-half ago, Cobb’s office was shaken by is when things go down, and when you’ve a small earthquake — hardly a surprise in 2. Talk to the specialists. Your institution forgotten backups for the past three months.” California. A picture of former US president, employs people to think about data full-time, and one-time client, Gerald Ford fell off the so talk to them, advises Juliane Schneider, 6. Protect raw data. All data are precious, wall and hit his laptop “just right”, shattering who leads data curation at Harvard Catalyst but raw data are irreplaceable: the only way to the screen. “After that, I was like, ‘I better move in Boston, Massachusetts. Your research- recreate them is to run the experiment again. things around so I am better prepared’.” computing centre might offer free or low-cost These must therefore be backed up — and institutional backup systems; your librarian kept as read-only files. Wickes once had to kill 10. Keep a backup offline. Internet-con- can help you to craft a data-management a project because she opened a crucial file in nected backup devices are convenient: the strategy; and your grants office can advise you Microsoft Excel, which automatically format- data are instantly available. But those devices on funding-agency requirements, including ted a column, changing the values and ruining are also instantly vulnerable to user error and how, and for how long, data must be main- the underlying data set. So, protect your raw malicious software (malware). Craig Rager, tained. “They want to help you keep your data, says Martinez, “no matter what”. chief technology officer at Data Mechanix, data — especially if you have a grant,” she says. a data-recovery firm in Irvine, California, 7. Make backing up achievable. A data- says that many of his clients have suffered 3. Manage your data. Reliable backups management plan must be easy to follow for ransomware attacks, in which a virus encrypts require clever data management. Referenc- new members of the lab, as well as for postdocs a computer’s hard disk, making it unusable. A ing the organizing method devised by Marie who are pulling an all-nighter. “You might say, backup drive, whether attached to the com- Kondo, a popular Japanese lifestyle consultant ‘Oh, this is a perfect system.’ Okay, now, are you puter directly or through a network, can also and author of The Life-Changing Magic of going to do it at 3 a.m., after you’ve been work- be hit in such an attack, he notes. “Because you Tidying (2014), Ciera Martinez, a data scien- ing for 24 hours on something? Are you going can never eliminate this threat 100%, the only tist at the University of California, Berkeley, to do it when you’re in the middle of fighting thing you can really do is have a device that advises asking of each file: ‘Does this need to with a code problem?” Wickes says. Discuss the you back up, which is then taken offline or not be stored?’ Adds Teal, with a laugh, “You can’t strategy as a team, and make sure that it’s work- accessible to your network,” for instance, by just keep the data that bring you joy!” able. Then, just as you would for your −80 °C being powered off. Establish conventions on file naming and freezer, simulate what would happen if disas- organization — for instance, that each project ter strikes: what data would you lose, and how 11. Plan ahead. Ultimately, your data need to gets its own folder; that data and code go quickly could you recover? “At a minimum, that be available in the future. So, think about “future into dedicated subdirectories; and that each as a thought experiment would be valuable,” you”, says Teal. Consider the media on which project folder includes a file that documents says Teal. your data are saved, and the applications that the project’s aims, methods, metadata and you use to open them. Try to stay up to date. files. Plan where and how data are backed up, 8. Test backups regularly. Don’t assume Much of Vosshall’s early data are stored in an and develop a schedule — daily or weekly, for that your backups are working: test them. obsolete disk format, she says, meaning they’re instance — for doing so. Can you open your files? Do you have the backed up but inaccessible. “I’d have to go to an Raw data should always be saved, but necessary applications, login credentials antique store to find a reader.” Even the cloud intermediate processing files can often be and registration keys to run them? Wickes’ provides no guarantees: data-storage companies discarded.