We all aware of the mind numbing growth of data being generated today. It is for that reason that according to IDC, more than 30,000 PB of new external storage system capacity was purchased in 2014 alone, an increase of 30% over the prior year. Beyond simply being aware of it however, we must learn to master the ability of maximizing the value of all of this data. In order to contend in the highly competitive global economy which we operate, IT managers must create a data structure that gives users the ability access this data, process it and bring it to value as quickly as possible. Innovation is originated from data, reinforced by data and efficiently delivered by data. In the previous ten years, enterprises adapted computer virtualization in order to achieve the states of agility and flexibility that business demands today. In the coming ten years, the challenge will be to transform the data infrastructure into highly efficient and elastic pools that innovation demands.
But it’s not just data that is driving the necessity for greater storage silos. Virtualization and automated provisioning demand storage pools as well. Server and computer endpoint deployments such as can now be reduced to minutes. In the previous ten years, enterprises adapted computer virtualization in order to achieve the states of agility and flexibility that business demands today in order to meet both predicted and unpredicted usage spikes of services and resources.
Because storage pools are the underlying infrastructure of both data and virtual endpoints, organizations require data silos that are fast, flexible, easy to manage and redundant. Obviously decentralized storage is no longer an option today, not just from a redundancy side, but because manual configurations of DAS storage is no longer viable to match the speed of business today.
Which brings us to shared storage, the pooling of multiple disk arrays into one large conglomerate that can then be divvied out into a manifold of silos. The traditional options of shared storage over the previous have been the SAN and NAS.
A SAN is a very robust device. It is designed to withstand a sizable degree of hardware failure by integrating great redundancy into the hardware design. A SAN can consist of multiple shelves of drives, each shelf composed of one or more RAIDs with designated hot spares. Shelves can then be mirrored together to form a RAID 50. To complete the redundancy, the SAN requires multiple controllers which are connected to multiple iSCSI switches or in some instances, Fibre Channel switches and converged networks. The SAN is very secure as well, presenting block level storage only to the servers themselves through a physical connection. Users must access the designated storage pools through the servers themselves, isolating and protecting the storage backend from the rest of the network. NAS on the other hand offers file level storage directly to the user. While SANs are utilized for virtual ecosystems and database backends, the NAS is mostly limited to file access.
SAN though has definite limitations however. It has a custom proprietary operating system that has to be updated and patched regularly. In addition, vendors have to spend much of their time designing and testing proprietary components and conducting interoperability testing. It is no wonder then that storage array vendors spend millions in development costs creating these monolithic robust structures, which is why SANs are so expensive. A SAN may consist of a great number of disk drives, but that doesn’t translate into scalability. There is a limit to the number of drives a controller can manage and a limit to the number of controllers that can be interconnected. SANs are designed for multi-tenant environments as there is typically a very limited number of administrators, many times only one. This serves as an obstacle for multiple workloads and self-provisioning deployments. The purchase of a SAN then locks you in to a particular vendor, increasing opportunity costs and forcing future complicated data migrations. Finally, an enterprise may have multiple storage arrays from multiple vendors, each requiring a separate unique management console that eliminates any chance of cohesion.
It is these limitations that have brought us a new paradigm in how storage is managed, provisioned and presented called Software Defined Storage. SDS is hardware agnostic, period. It is fully compatible with x86 architecture and works with nearly any storage platform, integrated or DAS. Thanks to SDS technology, servers today are being designed with small sized flash based disks that can handle petabytes of storage on their own. This means that the files for your virtual machines and applications can be stored on the disks residing on their host servers. The implications of this are enormously beneficial in that not only is the control plane separated from the hardware, but it puts the server and its corresponding storage on the same data plane. This greatly simplifies management for not just storage but for the entire ecosystem. Virtual machines can now communicate with their assigned file systems through the system bus rather than the iSCSI network, just as the virtual machines communicate with one another without accessing the network. This greatly increases speed and performance.
In the same way in which the hypervisor virtualized the bare metal server, Software Defined Storage is using a similar approach in order to create elastic and redundant storage pools. Although there will always be justifications for the traditional robust hard metal SAN, IT leaders are quickly learning the value of this storage transformation.
bradinthecloud 
Posted on January 4, 2017
0