Current Filter: Storage>>>>>Opinion>

   

For IT professionals tasked with updating and refreshing data centre storage infrastructures, there are many first-rate all-flash arrays (AFAs) on the market that offer attractive density, performance, and value. This article offers tips for integrating AFAs successfully while minimising or removing the challenges of creating another storage silo or manually managing a separate suite of data services.

Unless a business has the budget and capability to shift all workloads onto a single AFA or set of identical AFAs most data centres will likely have a mix of AFAs, hybrid arrays, and HDD storage resources that each require management and support. This creates a complicated mix of tools, operational practices, and (usually) proprietary capabilities that actually increase risk and cost. Despite this fact, the benefits of all-flash arrays are compelling enough for this problem to be addressed and resolved.

One of the most interesting and successful resolutions to this problem comes from putting in place a heterogeneous, software-defined storage (SDS) platform that enables a consistent, unified storage infrastructure across storage platforms, regardless of hardware, manufacturer or protocols. The end result produces a common feature set that reduces complexity, eases the administrative burden, and simplifies integration. But not all SDS platforms deliver on this promise.

Clearly, the most compelling reason to add AFAs is for higher performance, but it will likely be one of several storage platforms, as not all workloads are optimised to take advantage of flash. While most SDS solutions boast about performance capabilities, its unavoidable that any centralised storage management adds latency. SDS platforms that are tuned or developed with flash media in mind minimise this latency pitfall through clever engineering, while also optimising legacy, non-flash arrays. SDS solutions that offer high-end data efficiency and optimisation features, like deduplication and IO path optimisation across storage systems, can maximise utilisation while ensuring the best possible performance.

The first challenge encountered when adding an all-flash array is data migration. Moving or copying data using traditional migration tools will almost always encounter performance and complexity issues surrounding AFA native thin-provisioning, deduplication and trimming functions. SDS migration features, if array-aware, will override or bypass the arrays built-in thin provisioning and optimise data transfer (by not copying zero-blocks), far surpassing the efficiency of traditional block or replication technologies. Additionally, tuned SDS migration platforms will use private storage network transfer (or isolated IP), so as to not interfere with online, production applications. Following the initial migration, a good SDS platform will handle moving data among the various underlying storage systems non-disruptively, maximising the price/performance value of the AFA by removing stale data, and maintaining the performance for applications that need it.

The second primary challenge faced when adopting AFA into a data centre lies in maintaining key disaster recovery and business continuity SLAs. AFAs generally include (or offer as options) some basic data protection features, but those add to the proliferation of siloed tools and capabilities. Further, critical enterprise-class requirements like synchronous and asynchronous replication or always-on clustering are frequently missing. Among the key benefits of SDS are data protection and continuity that safeguard data on AFAs without the need to use additional point solutions. At a minimum, an enterprise-class SDS platform should provide data protection, replication, and recovery, as well as be able to monitor storage and network connections and volumes to identify any failures, shift operations elsewhere, and keep applications and data available.

Organisations looking to expand with all-flash arrays benefit greatly from SDS solutions because they not only gain those important management advantages, they can disregard most of the arrays own software features and focus instead on getting the best, most cost-effective hardware. When coupled with a solid, enterprise-class SDS platform, all-flash arrays can be regarded as extremely high-performance commodity hardware that can be brought on board with little to no disruption at a lower cost.
More info: www.falconstor.com