My take on unified storage
I’m sure I’ll get some comments here from zealots who disagree with parts, but here goes. Unified storage refers to a single logical system that is capable of presenting both dedicated LUNs (SAN) as well as folders within a file system (NAS). Where do the zealots play into this? Basically, in how this is performed.
There are unified storage systems where the SAN blocks are actually sitting on top of a file system and the translation happens within the storage controller. There are other unified systems where there are dedicated controllers for SAN functions and separate controllers for NAS functions, but there is a management wrapper that creates a single logical system to manage. Who am I to say which is right, or for that matter, why a NAS gateway connected to a SAN array couldn’t be a form of unified storage?
Since we’ve already covered SAN and NAS to some degree of depth (and I’ve presented that unified is largely a congruence of these two under a common logical system umbrella), I’m going to brief here. However, I will suggest that customers need to very carefully examine the individual application performance attributes of what they are storing and how they are storing it vs. automatically jumping to the conclusion that “unified” is somehow the best of both worlds.
Depending on how that unified storage is deployed, you might end up with sub-par application performance in the pursuit of perceived management simplicity. The truth is that SAN management has come a long way since unified storage first came on the scene and now many dedicated disk arrays are designed to be managed by IT generalists using server-oriented management tools. Simplicity is not really the best decision criteria to use when looking at dedicated NAS and SAN versus unified approaches. As Forrest Gump might say, “that’s all I’ve got to say about that.”
Now this is where things get fun: object storage
With the move towards cloud computing and service delivery over the public internet, traditional storage access paradigms like NAS and SAN are starting to see new options move into the neighborhood. While probably more closely related to NAS vs. SAN, object-based storage does not provide a block-oriented interface or access via a file system and file folders, but rather it organizes data into variable sized containers, or objects. Each object has the data itself (a stream of bytes) as well as metadata (data about the data). Objects are then organized in a flat hierarchical structure of buckets and accounts.
Unlike low-level SCSI-based commands to manipulate the blocks, access to the object is achieved through higher-level commands that manipulate the object in its entirety…create, delete, get, put, etc. Information about where the object physically resides and security mechanisms associated with the object are stored as metadata.
Just as NAS and other file-based systems are accessed via protocols like CIFS and NFS, object-based systems are similarly accessed. In this case, object storage is accessed by web service APIs such as REST, SOAP, or Amazon S3 via protocols such as HTTP, SMTP, or XML. The main thing to note is that the commands to locate and act on the object are embedded—for example, in the URL string within the HTTP request.
While not the same thing as cloud storage, these two are closely related. Most cloud storage offerings are utilizing object storage devices and mechanisms as the method to deliver those services. The stateless nature of object storage means it is not generally an appropriate play for high-performance applications that need low latency and guaranteed performance levels.
But in a world where people are trying to squeeze every ounce of an IT budget, many are turning to the cloud for storage where it is appropriate. When compared to NAS and especially SAN, Object storage provides a latency tolerant access pattern to a data set. For some applications, this is tolerable and provides customers the ability to geographically separate the physical storage from the end user. Think about archive and backup as key use models where object storage is growing rapidly and people are more willing to move data off premise.
This area is changing rapidly as companies struggle to find a better way to deal with explosive growth in unstructured content. Recent directions include the emergence of cloud gateway and on-ramp solutions to bridge on premise storage to cloud storage and enable non-object based applications to access data stored in the object pool using traditional iSCSI or NAS out the front.
A final word. . .
It’s more than a cliché. Information really is the lifeblood of business and almost everything we Facebook, Tweet, or Snap, so it’s important to understand the how, where, and why of data storage. I spent almost 10 years as an IT manager at a small hosting provider before coming to work at HP, and in that role was often overwhelmed with the variance of technology options. Storage is probably one of the worst offenders when it comes to jargon, nuance, and cryptic knowledge. Hopefully by sharing my thoughts here, I have helped you to decode some of the marketing-speak that folks like myself are often responsible for disseminating in our day jobs, and perhaps, in the spirit of my recent trip through Asia, I’ve even helped to generate some positive storage karma.
I’d love to keep this conversation going. Please let me know your views of the changing storage landscape. I also welcome your suggestions on other topics you’d like to discuss. Let me know and I’ll dive in. You can also follow me on Twitter: @HPBradParks
Learn more about storage innovation from HP.
If you missed it, here are the first two blogs in this series: