To learn where the major inefficiencies are in your storage system, or to address them, it helps to know how they got there.
There are a variety of explanations for why so many firms have rolled
The blame can be shared by both vendors and consumers. Vendors often point an accusing finger at consumers, noting that expensive efforts undertaken in the past to implement sophisticated, standards-based management capabilities on their wares have gone largely unrewarded in the marketplace. This assertion is borne out in consumer surveys where respondents admit that "management" rarely finds its way onto their list of array selection criteria. Instead of seeking a comprehensive storage resource management (SRM) strategy, most user concerns are typically assuaged by the element manager software provided on each box, usually accessed by typing in a URL corresponding to the array's management webpage. Element management works until several storage arrays are deployed and it becomes necessary to surf each array's webpage to collect status information or troubleshoot a problem.
On the other hand, vendors have demonstrated little interest in enabling their wares to be managed in common with their competitors' products. They also often find the benefits of cross-platform management at odds with their pursuit of product-discriminating and revenue-enhancing value-added software features. For example, a feature such as thin provisioning -- which consumers are willing to pay a premium for on their arrays -- may actually obfuscate the ability of management software to collect the actual available capacity of an array. In another case, support added by the vendor to a NAS appliance for fast file-system access may impair the ability of files to be migrated from that array to lower-cost file archive storage, preventing efficient file tiering or archiving.
Whatever the explanation, unified resource management in the storage infrastructure is lacking, and inefficiency is the inevitable result. It follows then that improving our capability to monitor and manage the storage infrastructure hardware layer should have a positive impact on efficiency.
Theoretically, a homogenous infrastructure (infrastructure comprising gear purchased from the same vendor) should be an easier fit for resource management than a heterogeneous infrastructure (gear purchased from different vendors). Indeed, you will hear some industry analysts claim that fielding a homogeneous infrastructure is the only practical way to drive down cost and improve storage efficiency. The problem is that very few vendors have a sufficiently deep bench of storage gear to meet all the needs of an enterprise consumer. Those that have various platforms to meet different user needs have rarely developed the gear internally. Instead, they have purchased smaller firms to obtain their products or have simply entered into OEM agreements where they represent the arrays developed elsewhere -- and that have no architectural affinity to the gear built in the original company -- as part of their product family. Hence, going the homogenous route to obtain efficiency is not a valid solution.
There has also been some activity recently, within IBM and elsewhere, to field manageable infrastructure, including storage, by linking it to a hardware/software stack or platform intended for a specific purpose, such as cloud computing. IBM recently announced a comprehensive management solution for both physical storage resources and storage services (the value-added software usually bundled on array controllers) that works with the OpenStack cloud computing platform. Unfortunately, the integrated management tools recently introduced by IBM do not work with hardware that is not part of an OpenStack configuration.
Choosing the right SRM software for your infrastructure
Few companies have rolled out infrastructure with management in mind, so the question is how best to retrofit existing infrastructure for better monitoring and management. The current answer is to implement SRM software from a third-party vendor, such as EMC, Hewlett-Packard, IBM Tivoli, SolarWinds, Symantec and a number of other players. But installing them after the infrastructure is deployed can be challenging and expensive.
To choose the right product, you need to assess your current infrastructure. If the brand names on your storage arrays are those of industry hardware leaders, the chances are good that the products are supported for management by leading third-party SRM tools. If there are some off-brand products in your infrastructure mix, you may need to look at what onboard management hooks the arrays provide.
SRM is usually accomplished via a combination of vendor proprietary application programming interfaces (APIs) that provide doorways into array element management software, plus any or all of the following "standards-based" management hooks, including Simple Network Management Protocol (SNMP), Management Information Bases (MIBs), Storage Management Initiative Specification (SMI-S) providers, and/or Representational State Transfer (REST) Web services. Vendors like SolarWinds freely admit that, in the absence of a dominant management methodology, an all-of-the-above approach is typically required.
The smart move is to download a trial version of the software, try it out, and note the capabilities and limitations of the SRM dashboard and report data. Be aware that hardware vendor APIs may provide the richest and most current status data, but these access paths are the most vulnerable to interruption. When a hardware vendor makes a change to an API, it often causes the monitored resource to disappear from the SRM dashboard radar until the SRM software vendor amends its own code. SNMP is faulted for its lack of real-time data presentation, while SMI-S is implemented differently by vendors and may reveal various kinds of status information inconsistently between vendor products. RESTful management is the best approach for managing infrastructure resources and services, but the vendor community has been slow to adopt REST, arguably because of its commoditizing effect.
After trying various SRM tools, select the one that best fits your needs. Then let your vendors know that storage hardware purchases going forward will be contingent in part on the support of the storage product for the SRM tool being used for resource management. Arrays and other storage-related gear that cannot be monitored using your preferred SRM tool will not be purchased. This step alone can help bring your infrastructure under better management so that steps can be taken to improve the efficiency of resource allocation and break/fix processes.
This was first published in June 2014