Random Access

Options Abound in the Modern Storage Market

What's old is new again in the physical storage choices enterprises have.

There was a time not all that long ago when buying storage meant going out and buying a SAN or a NAS, cabling it up and connecting it to servers.  Today, the enterprise storage market is a far more interesting place with far more options provided to potential customers.  Further, as often seems to be the case in enterprise IT, what is old seems to be new again as server-based storage is once again en vogue and is more than a viable option for supporting even the most critical workloads.  With that said, let's take a look at some of the options that are available in today's storage market.

Direct-Attached Storage
Although buying tons of storage for local servers kind of went by the wayside with the rise of virtualization and its insistence on external shared storage, when it comes to running physical services, many organizations still opt for this old standby option.  From a simplicity standpoint, server-based storage is often easy to manage and configure, even when run behind a hardware RAID adapter.

However, scalability can be somewhat limited in traditional direct-attached storage systems, particularly when this is limited to in-server storage only.  In addition, in a traditional model, the end result is often the creation of islands of storage, each which is managed separately.  This can result in increased administrative overhead and less efficient capacity usage.

Server-Side Storage/Scalable
Today, we're seeing resurgence in the use of server-side storage, but with a really big twist.  Now, rather than simply deploying a bunch of storage islands, there are solutions on the market that aggregate and harness this storage power as a distributed, scalable pool of capacity.  In some cases, this aggregation requires the use of a mix of solid state and rotational storage, which does one (or both) of the following:

  • Create a storage tier.
  • Create a mega cache to accelerate everything.

These kinds of systems are gaining steam with the rise of hyperconverged infrastructure, which brings the storage layer back into the compute layer and then uses software mechanisms to create scalability and which unifies the storage into a single resource pool.  In some cases, scalability is achieved through the use of a controller virtual machine that resides on each host while in others, dedicated hardware handles much of the scalability work.

When done right, such systems are also providing enterprise storage features, including deduplication and compression. Many of them have eschewed RAID as a data protection mechanism in favor of data replication, which carries with it less complexity and less processing overhead than is needed to perform RAID parity calculations.

Server-Side Caching/Tier
In addition to seeing all of the storage moved server side, we're also seeing just a fast caching/storage tier moved to the server in the form of PCI-e based storage.  These kinds of systems are often used to improve the performance of traditional SAN devices.  They work by creating a huge cache that sits between the host and the SAN and can accelerate both reads and writes.  In short, these kinds of systems reorder write operations in a way that is favorable to rotational storage.  Legacy storage doesn't really like random write operations since it makes the hardware have to jump all over the place to work.  So, a caching layer will reorder random I/O into a sequential pattern that can be written with much more efficiently.

Traditional SAN/NAS
This class of storage is what has become the most common deployment scenario in data centers.  Perhaps the biggest change today is the inclusion of solid-state storage in these arrays.  There are now arrays that include 100% flash storage that provide massive performance benefits, which comes at the cost of capacity.

However, hybrid storage arrays have also hit the market.  These arrays leverage a combination of both solid state and rotational storage in ways that provide a whole lot of performance while also providing a whole lot of capacity.  In fact, it's possible to get an array in this class that provides 22 TB of raw storage and 30,000 IOPS in a single unit for way under $100,000.  Not bad!

 

About the Author

Scott D. Lowe is the founder and managing consultant of The 1610 Group, a strategic and tactical IT consulting firm based in the Midwest. Scott has been in the IT field for close to 20 years and spent 10 of those years in filling the CIO role for various organizations. He's also either authored or co-authored four books and is the creator of 10 video training courses for TrainSignal.


Featured

comments powered by Disqus

Subscribe on YouTube