Random Access
The Cloud's Role in Enterprise Data Storage Architecture
Here's some of the options IT has in tying their on-premise data storage with the cloud.
The life of a storage architect has become a heck of a lot more exciting in recent years. Unlike years before, there are now all kinds of new opportunities for storage administrators to leverage to meet the various needs of their organizations. Among the new opportunities that are out there are various cloud-based services.
Before we get into a discussion of the cloud options, let's talk about some of the needs that are associated with various kinds of tasks and how different kinds of storage options can be leveraged to meet those needs. The table below lists common workloads and how various kind of storage supports those needs.
| |
Virtual Machines |
General File Storage |
Big Data |
VDI/Desktops |
| Local Storage |
Works fine, but creates limitations; some features require shared storage |
Risks creating storage resource islands, but does not usually pose performance issues |
Depends on type of disk and capacity needs |
Not suitable due to scale issues |
| Hyperconverged |
Requires new ways of thinking; IT staff not always prepared for the adjustment |
Won't always work unless a virtual machine in the environment acts as a front end file server |
The storage architecture often lends itself well to big data needs and its proximity to the compute layer reduces latency |
One of the early "wins" for hyperconverged space, VDI is a perfect fit here thanks to low latency, high IO, high capacity and scalability |
| HDD-Only SAN |
Works fine, but will eventually hit a performance wall |
Requires a front end file server; provides ample expansion opportunities |
Will often be insufficient due to HDD limitations |
Not suitable due to massive I/O issues ineherent with VDI deployments |
| Hybrid SAN |
This is the next step beyond an HDD only SAN for those thave have hit the performance wall |
May be overkill for just file storage, but with powerful deduplication, could be more cost effective than other solutions |
Often a very god solution for big data workloads, especially when those needs are small to mid-size |
Very suitable for VDI deployments thanks to solid state based caching mechanisms |
| All Flash SAN |
Only needed for the most I/O intensive virtual machines; a waste of money in other cases |
In general, major overkill for simple file storage needs unless working with very large files or latency sensitive ones |
A perfect solution for even large bif data workloads |
Excellent choice for VDI thanks to raw I/O performance, but may be too expensive |
Table 1. Task goals for different storage options.
|
With that understanding under your belt, how would each of these workloads translate to work in a cloud-based environment?
Virtual Machines
With the rise of the hybrid cloud -- that is, a cloud service that integrates into an organization's private cloud and allows seamless migration of workloads between services -- the cloud has become a viable target for virtual machines. The common cloud players provide full compatibility for their products between on-premises and off-premises services, simplifying the process by which customers can shit workloads between environments.
General File Storage
In my opinion, general file storage is a perfect fit for the cloud and there are a number of ways that files can be stored there. First, you can use one of the companies that specialize in file storage in the cloud. Or, you can take an appliance-based approach. In this approach, you install an appliance in your data center, which acts as a file server. However, when you save files to it, they are saved to the appliance vendor's cloud service. The appliance just acts as just another file server. It caches files locally and syncs them with the cloud service s needed.
Big Data
Big data and the cloud go hand in hand. Most organizations can't afford the cost of building a complete big data analytics environment. So, many of them push the heavy lifting to cloud providers and just download the results. In this way, the organization gets the best of both worlds -- local and fast to work with data, but massive capacity to crunch the data.
VDI/Desktops
There was a day when VDI didn't belong anywhere except internal and on only the fastest possible storage. Today, though, we're seeing a rise of desktop-as-a-service vendors providing this resource-intensive service as a cloud-based product offering and it's a trend I'm happy to see. If a vendor can build a robust back end VDI environment coupled with a low-latency client and provide users with a low-latency connection, it can vastly streamline the VDI deployment procress
About the Author
Scott Lowe is a virtualization architect for ePlus Technology in Herndon, Va., and author of "Mastering VMware vSphere 4" (Sybex, 2009).