But some of the most compelling attributes of the public cloud are best suited to ephemeral or (relatively!) short-term use cases. You can spin up a cloud server in minutes. You can scale a cloud-based application to cope with the peaks and troughs of demand. You can control all of this through a web console, with no more than a credit card and a laptop. Silicon Valley, SoMa, Silicon Alley, Silicon Roundabout, Silicon Allee, Silicon Wadi, Silicon Forest, Silicon Welly, and the Silicon Bog (only one of those was made up, I think) are full to bursting with bright young things building exciting new products (and silly photo sharing sites) powered only by the cloud and expensive coffee.
And then you have government, private, and commercial Archives, with an over-riding imperative to keep stuff for a very, very long time. These Archives clearly can (and do) use cloud computing in the same ways as everyone else. They use clouds to cost-effectively transform data from one format to another, they use clouds to stream large and popular media files to the public, and they use clouds in all sorts of other ways to make innumerable workflows and processes easier, cheaper, or more robust. For those use cases, even the biggest, grandest, and most important of archives is actually pretty much like any other user. Cloud’s as useful to them as it is to the rest of us, and that’s great.
Does it make sense, though, for Archives to entrust any of their long-term preservation role to the cloud? I’m not sure (yet), but The National Archives (TNA) here in the UK wants to find out. They’ve commissioned a study from a small consultancy, Charles Beagrie, and I’m subcontracted to provide a bit of cloud knowledge to the team.
Out of the box, you’d have to question the sense of an archive entrusting anything to the public cloud for purposes of long-term preservation. That’s not really what Amazon’s Simple Storage Service or Rackspace’s Cloud Files or any of the other cloud-based filestores are for. Their Service Level Agreements and their technical underpinnings are all about cost-effectively storing lots of stuff and losing as little as possible. If a file is lost or damaged, the service provider might pay out a few service credits, and/or the customer might restore from a backup, and everyone continues on their way. Archivists, we were reminded at one of the project’s focus groups, have this peculiar expectation that the systems they use to preserve their primary materials won’t lose anything at all. A couple of service credits don’t really help when you just lost, truncated, or changed a few words in the digital equivalent of the Magna Carta or the Domesday Book or the Book of Kells or the Declaration of Arbroath. And, just to be totally clear, losing a digital copy of the Declaration of Arbroath would be ok. The National Archives of Scotland still has the vellum (I presume their copy was written on vellum?) in a climate-controlled vault. They probably also have a CD or two of backups for the digital images. Things become a bit more serious when the content is ‘born digital,’ and the file you’re preserving is the thing itself and not just an image of some physical artefact.
is designed to provide average annual durability of 99.999999999% for an archive. The service redundantly stores data in multiple facilities and on multiple devices within each facility. To increase durability, Amazon Glacier synchronously stores your data across multiple facilities before returning SUCCESS on uploading archives. Unlike traditional systems that can require laborious data verification and manual repair, Glacier performs regular, systematic data integrity checks and is built to be automatically self-healing,
the big public cloud providers aren’t really in the business of supporting the extreme needs of an Archive. Archives demand a whole extra level of error checking, resilience, redundancy and integrity, and it would be cost-prohibitive for AWS and their competitors to do all that across their sprawling data centres when most customers are actually perfectly happy with “redundantly stores data in multiple facilities” and “automatically self-healing.”
Interestingly, Seagate sees value in offering a Glacier competitor capable of storing data “intact for decades” and offering access instantly rather than in a matter of hours as Glacier does. As it’s based in Utah I doubt that European government archives would touch it, but it will be interesting to see whether their North American cousins show any interest…
One thing, of course, that most public cloud providers are good at is offering a platform upon which others can build. Archivists, like others, have begun to layer rules, policies, procedures and processes on top of the bare-bones cloud infrastructure offerings, to build something a little more robust and dependable. Services like DuraCloud take AWS and Rackspace (currently only in their US data centres, but that could change), and add things like proactive error checking and even more backups to deliver something that an archivist might be prepared to trust.
There’s a use case here, and there are plenty of (mostly university) archives in the States putting DuraCloud and similar cloud-powered tools to work as part of their preservation strategy.
But I can’t help wondering if some great big enterprise data management solution, with multiply redundant disks, multiply redundant backups and a whole heap of watertight, ironclad, fault tolerant, and ridiculously over-specified policies might be a better (albeit eye-wateringly expensive) way to preserve the truly irreplaceable? Either that, or archives and archivists need to explicitly embrace a more pragmatic approach to what they’re attempting with these systems.
‘Design for failure’ is a core tenet of cloud-powered systems. What’s the archival equivalent? ‘Lose nothing, ever’ just won’t cut it.
Disclaimer: Charles Beagrie is a client. TNA is a client of theirs. This post is not part of the project. Any opinions expressed here are my own, a work in progress… and subject to change!