Private Clouds & Storage: Considerations

by dave on February 13, 2009



One of the current “hot” topics within the IT industry these days is the concept of cloud computing.  While there is a level of ambiguity around what cloud computing actually is, there is a definite trend towards looking at what these cloud services and storage can do for the commercial and enterprise.

Private Cloud Challenges

In a previous posting, I talked about the concept of a common cloud file system or cFS.  Some folks responded on Twitter that perhaps IBM‘s GPFS filesystem (as a part of their SOFS strategy) could indeed fulfill that role for the cloud.  As an additional twist, PaaS “connectors” such as SnapLogic, et. al. that seek to tie front-end processes or applications to underlying cloud services could simply provide a API hook to entities like Nirvanix, GoGrid, Amazon S3, etc. for handling discrete storage functions (with the obvious implications of data access/protection SLAs being separate and distinct from the middleware providers).  The provision for this changes when you start talking about privatizing the cloud.

Extrapolating this even further, Steve Todd posted the following this morning:

The Impact On Private Clouds

Customers building a private cloud will have the option to purchase storage as hardware or to purchase it as a file.

For those who purchase storage as a file, they will experience a new paradigm in how they deploy and configure block, file, and object storage.

As Steve explains, tossing another layer across the top of a cloud computing PaaS offering is the concept that now you can have a portable file environment that is completely insulated from others within a global NOC.  This concept of a “private cloud file system” can obviously be coupled to an insulated processing/grid service as well (i.e. what a basic definition of a “private cloud” could be construed to be).  Steve points out that the Centera VSA provided this level of functionality that could run within a virtualized environment. Similarly,  the EMC Celerra VSA provides similar capabilities as a “pluggable” entity that allows an enterprise to add additional capabilities to their environment.

These pluggable architectures don’t erase the need for a base file system, after all, even your virtualized appliances are run within another “sphere” of storage (i.e VMFS) but being able to seemlessly place these islands into the ether requires a level of standardization that at this point doesn’t exist.  S3’s underlying filesystem for data placement is different than that of Nirvanix, GoGrid, et al.  Couple that with object-oriented filesystems (like EMC‘s Atmos) and you’ve got a veritable plethora of options that have absolutely zero interoperability.  PaaS offerings thrive in this type of environment since their model is built upon connecting application x to filesystem y but this fundementally creates additional complexity and complicated SLA models for enterprise.

Closing Thoughts

I’ve spent a fair amount of time beating on PaaS vendors in this post and while I’ll stand by that, it’s not fair to say that they have no value or future purpose.  Quite to the contrary, in fact. I think the nature of their “connectors” is going to change from being an interpreter to being a optimizing and securing conduit for business.  By creating application-specific connectors to cFS, the I/O stream from internal systems to external cloud (albeit private) can be optimzed or maintained.  Additionally, these type of connectors could perform the same type of write-splitting that we currently see with EMC’s Recoverpoint product, for example. This would enable concurrent data replicas to allow for data redundancy and pathing across multiple private cloud providers.  Pretty exciting challenges that are already being designed and worked on by the talented individuals out there!

Thoughts and comments are always welcome!

 

Reblog this post [with Zemanta]
Share
  • http://cloudberrylab.com Andy

    here is a great example how 3rd party employ storage API provided by online storage vendors. Our CloudBerry Explorer freeware uses Amazon S3 storage API and unleashes the power of Amazon S3 to non-programmers. http://cloudberrylab.com/

  • dave

    Andy,
    thanks for your comments here. what you’re discussing makes absolutely perfect sense but the model of storage within the cloud still needs to be fundamentally changed.
    look forward to seeing more from you.

    thanks,

    Dave

  • dave

    Andy,
    thanks for your comments here. what you’re discussing makes absolutely perfect sense but the model of storage within the cloud still needs to be fundamentally changed.
    look forward to seeing more from you.

    thanks,

    Dave