All of they above. Archive data has elements of all the data types. However, I see companies focused on parts of the distributed application. Look at ML/AI. You'd never do it in the DC. What type of data is ML & AI in the cloud?
Getting past "checkbox" regulation compliance...with tapes after 5 years who knows if can really restore much less move to new tape media (new LTO formats!).
If I decide to deploy HANA in AWS based on elastic compute then it will be structured data. If I want to use cheap and deep storage for AI/ML then it may be unstructured. It gets to the value prop of public cloud for your hybrid cloud use case.
@dinisco we've had cloud file gateways around for a while but adoption still seems to be mixed...people get nervous about the potential lack of performance predictability for "stubbed data"
@andriven with using cloud for archive it's still "on disk" and if done right both can keep a catalog plus selectively restore (not pull back a 100 GB vmdk when just need a 100 KB file)
there's hope that using cloud for certain data types will avoid "data migration" due to changing underlying physical media types (new LTO versions, storage array upgrades)
The hard question - Do you need an hybrid cloud data strategy or better yet a global hybrid cloud file system. @dfloyer talk about this concept a lot. I tend to side with him that enterprise need to think through the distributed hybrid FS.
often cost avoidance...storage is still a big line item but gets more scrutiny every budget cycle due to its size. You have to have options if get pressed harder than last year's budget cycle.
Absolutely Archive with Backup a close second - are seeing lots of customers replace tape with cloud archives. Cloud Bursting is great in theory but still really hard in practice given the weight of data and LAN vs. WAN speeds.
Also seeing test/dev drive it...spot pricing on AWS can even help there. This works given an underlying mechanism to move data around and leverage EC2/Azure instances on the fly.
All three but bursting has broader use cases that drive variable revenue. backup and archival must be standard so it's a hard decision different both valuable use cases
I think archive and backup is a dominant use case. I'm not sure for how long. Enterprises will figure out Object storage and offer a comparable cost to cloud based storage.
But you have to think beyond the IaaS layer. SAP is a good example. Cloud hosted enterprise apps make sense as cloud providers show greater ability to support the app than the enterprise. SAP in the cloud, bolt on in the private DC.
I'm starting to see an interesting edge case where people arbitrage against discounted bandwidth from the cloud providers. For instance US-Europe connectivity between branch offices may be cheaper routed through an AWS transit VPC
@CTOAdvisor There are even some providers (Cerner for instance) that will host just not just their own app but specific other apps that integrate closely or need to be at LAN speeds to the core app.
IaaS becomes an enablement for legacy software providers to remain relevant. Elastic compute allows for a business model where they run software they write for the customer with heavy capital investment.