Server virtualization can bring your business significant benefits, especially in the initial stages of deployment. Companies we speak with in the early stages of adoption often cite more flexible and automated management of both infrastructure and apps, along with CAPEX and OPEX savings resulting from workload consolidation. However, as an increasing number of apps are virtualized, many of these organizations encounter significant storage performance challenges. As more virtualized workloads are consolidated on a given host, aggregate IO demands put tremendous pressure on shared storage, server and networking resources, with the strain further exacerbated by the IO blender effect, in which IO streams processed by the hypervisor become random and unpredictable. Together, these conditions reduce host productivity—e.g. by lowering data and transactional throughput and increasing application response time—and may prevent you from meeting performance requirements for your business-critical applications.
How can you best address these storage performance challenges in your virtual infrastructure? Adding solid-state or flash storage will provide a significant performance boost, but where should it be deployed to give your critical applications the biggest improvement per dollar spent? How can you ensure that the additional storage fits effortlessly into your existing environment, without requiring disruptive and costly changes to your infrastructure, applications, or management capabilities?
We believe that server-side acceleration provides the best answer to all of these questions. In particular, we like server solutions that combine intelligent caching with high-performance PCIe memory, which are tightly integrated with the virtualization platform, and enable sharing of storage across multiple hosts or an entire cluster. The Flash Virtualization System from SanDisk is an outstanding example of such a solution. As we’ll see, Flash Virtualization enables a shared cache resource across a cluster of hosts in a VMware environment, improving application performance and response time without disrupting primary storage or host servers. This solution will allow you to satisfy SLAs and keep your users happy, without breaking the bank.
All Flash Arrays (AFAs) have had an impressive run of growth. From less than 5% of total array revenue in 2011, they’re expected to approach 50% of total revenue by the end of 2016, roughly a 60% CAGR. This isn’t surprising, really. Even though they’ve historically cost more on a $$/GB level (the gap is rapidly narrowing), they offer large advantages over hybrid and HDD-based arrays in every other area.
The most obvious advantage that SSDs have over HDDs is in performance. With no moving parts to slow them down, they can be over a thousand times faster than HDDs by some measures. Using them to eliminate storage bottlenecks, CIOs can squeeze more utility out of their servers. The high performance of SSD’s has allowed storage vendors to implement storage capacity optimization techniques such as thin deduplication within AFAs. Breathtaking performance combined with affordable capacity optimization has been the major driving force behind AFA market gains to date.
While people are generally aware that SSDs outperform HDDs by a large margin, they usually have less visibility into the other advantages that they bring to the table. SSDs are also superior to HDDs in the areas of reliability (and thus warranty), power consumption, cooling requirements and physical footprint. As we’ll see, these TCO advantages allow users to run at significantly lower OPEX levels when switching to AFAs from traditional, HDD-based arrays.
When looking at the total cost envelope, factoring in their superior performance, AFAs are already the intelligent purchase decision, particularly for Tier 1 mission critical workloads. Now, a new generation of high capacity SSDs is coming and it’s poised to accelerate the AFA takeover. We believe the Flash revolution in storage that started in 2011will outpace even the most optimistic forecast in 2016 easily eclipsing the 50% of total revenue predicted for external arrays. Let’s take a look at how and why.
Flash technology has burst on the IT scene within the past few years with a vengeance. Initially seen simply as a replacement for HDDs, flash now is triggering IT and business to rethink a lot of practices that have been well established for decades. One of those is data protection. Do you protect data the same way when it is sitting on flash as you did when HDDs ruled the day? How do you take into account that at raw cost/capacity levels, flash is still more expensive than HDDs? Do data deduplication and compression technologies change how you work with flash? Does the fact that flash technology is injected most often to alleviate severe application performance issues require you to rethink how you should protect, manage, and move this data?
These questions apply across the board when flash is injected into storage arrays but even more so when you consider all-flash arrays (AFAs), which are often associated with the most mission-critical applications an enterprise possesses. The expectations for application service levels and data protection recovery time objectives (RTOs) and recovery point objectives (RPOs) are vastly different in these environments. Given this, are existing data protection tools adequate? Or is there a better way to utilize these expensive assets and yet achieve far superior results? The short answer is yes to both.
In this Opinion piece we will focus on answering these questions broadly through the data protection lens. We will then look at a specific case of how data protection can be designed with flash in mind by considering the combination of flash-optimized HPE 3PAR StoreServ Storage, HPE StoreOnce System backup appliances, and HPE Recovery Management Central (RMC) software. These elements combine to produce an exceptional solution that meets the stringent application service requirements and data protection RTOs and RPOs that one finds in flash storage environments while keeping costs in check.
Virtual Instruments, the company created by the combination of the original Virtual Instruments and Load DynamiX, recently made available a free cloud-based service and community called WorkloadCentral. The service is designed to help storage professionals understand workload behavior and improve their knowledge of storage performance. Most will find valuable insights into storage performance with the simple use of this free service. For those who want to get a deeper understanding of workload behavior over time, or evaluate different storage products to determine which one is right for their specific application environment, or optimize their storage configurations for maximum efficiency, they can buy additional Load DynamiX Enterprise products available from the company.
The intent with WorkloadCentral is to create a web-based community that can share information about a variety of application workloads, perform workload analysis and create workload simulations. In an industry where workload sharing has been almost absent, this service will be well received by storage developers and IT users alike.
Read on to understand where WorkloadCentral fits into the overall application and storage performance spectrum...