Opportunities for Flash in the Data Center
These conferences described a data center infrastructure that is undergoing substantial technology refresh and substantial innovation, in support of pressing demands faced by IT organizations to support Big Data Analytics, Cloud Computing, Mobility and Social Media.
I’m interested in looking at the role of flash storage and memory in all of these changes. Many of the technical sessions at both conferences provided insights on areas in which flash will play important roles in this transformation.
From my work here at SanDisk®, I know that some of the key workloads driving this IT growth for flash deployments include: indexing, front-end web workloads (especially for web-scale computing); transactional processing, database workloads and Virtual Desktop Infrastructure (VDI). I’ve written a recent blog post that goes into more detail about the role of applications in shaping infrastructure deployment. See it here.
Supporting a Shift in Data Center Infrastructure
Data center infrastructure is changing from a siloed world into a more unified “fabric” of resources that can be tapped to support specific workloads. As the fabric gets built out, it will become increasingly possible to move the work, or the data, to the resources – and to manage the workflow via policy-based automation and management software.
Flash will increasingly provide storage capacity across the data center that keeps the most active data nearby to the available computing resources – and moves that data, if needed, to other resources.
The Breadth and Depth of Flash-Enabled Systems: Use-Case Deployments in the Data Center
Inspired by the discussions about data center transformation at the Gartner Infrastructure and Operations Management conferences , I began to think of how well flash technologies support these new types of deployments
Here, I’m listing a number of IT scenarios that involve flash-enabled deployments as part of the solution. It is by no means encyclopedic, but rather presents a view of the breadth and depth of flash-enabled systems in enterprise and cloud data centers.
Supporting the Virtualized Data Center
Today, many of the servers are virtualized in many data centers—and in some sites, more than that. Data centers that aren’t yet at that level can realize near-term benefits by increasing their virtualization level, so that they can pool more server and storage resources.
Data centers with high virtualization rates are ready to connect the IT system “silos” by moving workloads across the data center, shifting them to available resources, and managing them by policy and software automation.
At the Gartner data center conferences , Gartner vice president Philip Dawson, an analyst studying software-defined infrastructure, distilled his insights on the highly virtualized data center. He outlined how new, data-intensive workloads are being added into the portfolio of applications and databases in the data center. He specifically outlined the “hierarchy” of software-defined anything (SDx) being adopted in the data center, as follows
Hierarchy of Software-Defined Anything
Provisioning and management agility while optimizing process and service delivery costs — (reducing opex)
- Movement of functionality from hardware to software layer, fostering innovations as well as potentially extending investment in legacy hardware
- Reduces overprovisioning of hardware
- Fosters greater use of industry-standard server hardware (reduces capex)
- Faster provisioning and change management
- Be cautious of single points of failure — SDx drives high availability
Supporting a Shift in Data Center Infrastructure
Gartner’s view of storage trends was another spur to write this blog about flash in the data center, and its adoption. In their Five-Year Storage Scenario presentation at the Data Center Conference, Alan Dayley and Arun Chandrasekharan outlined the changes they expect to see in data-center storage over the next 2-3 years:
Technology disruption ahead:
- Software-defined storage
- Integrated systems
- Pervasive flash
- Cloud integrated storage
- Data management evolution:
- Blending of backup and DR
- Archiving, analytics and governance gains importance
My POV: What this Means for the Data Center
Clearly, the enterprise data center cannot stay as it is. Rather, IT managers need to think long and hard about what can be updated, replaced, migrated – or installed as “net-new”, having been designed from the ground up to support cloud computing, Big Data Analytics, Mobility and Social Media.
These four trends interact with each other: That is, many workloads will combine with cloud-enablement: Big Data and Cloud; Analytics and Cloud; or Mobility and Cloud. We’ve also seen dramatic growth in Web Analytics, providing near-real-time feedback on web workloads, which allows businesses to offer more data services to consumers – and to do so while the end-user is still online with their first request. Clearly, speed and time-to-results are needed to achieve this result.
Following are five deployment scenarios that I see as flowing from this kind of change in the data center infrastructure:
- Faster Server Platforms.
New servers leverage flash to increase their data-transfer rates, which helps to accelerate workloads, including enterprise applications and large databases. Flash-enabled servers are shipping from all major server vendors, and flash is part of the combined infrastructure shipped inside of integrated systems and hyper-converged systems. Adoption of flash is growing, given IT demands for faster I/O performance and higher data capacity on server platforms.
- All-Flash Arrays.
Adoption of all-flash arrays is increasing, including products from emerging vendors, in addition to well-established system OEMs. Although initially a small percentage of all storage arrays shipped, the proportion of all-flash arrays is increasing quickly over time, given their high data density, high capacity, and total aggregate IOPS performance. All-flash arrays will be helpful in addressing the pressing demands of Big Data Analytics, which require rapid processing in order to ensure faster time-to-results for the business organization.
- Addressing Storage Bottlenecks.
As the fabric within data centers is built out, with faster network switches, higher bandwidth, and faster servers, storage performance must improve over time. The need to reduce storage bottlenecks in enterprise data centers is one reason why flash is being increasingly adopted. Leveraging flash reduces the number of servers required to process workloads (applications and databases), reducing power/cooling requirements, while improving workload performance.
- Accelerating Database Workloads and Analyzing Big Data
Databases are getting larger in many enterprises, reaching the multi-TB levels – and beyond for the largest corporations with multi-PB data centers. Not only are they getting larger, but the size of the data-sets being analyzed is increasing. In short, there is a business need for time-to-results that is driving demand for accelerated database performance, which is a perfect use case for flash. Flash will play a strong role for in-memory computing (IMC), as that market segment grows.
- Supporting Cloud Infrastructure.
Cloud service providers (CSPs) often leverage large amounts of flash for fast data-services delivery, and are increasingly looking to flash for write-once, read-many workloads to store photos and other files related to social media. This cloud usage model is a growing opportunity for flash storage. But there is another cloud-based opportunity: Flash will help enable on-prem/off-prem workloads spanning data centers. It does this by facilitating large data transfers between these two types of data centers.
Today, flash storage is already widely deployed in enterprise and cloud data centers, and is expected to continue its strong adoption in years ahead. I’ve been involved with enterprise IT for more than 20 years as a journalist, industry analyst and, now, as enterprise solutions manager for SanDisk. These types of technology disruptions only come along once every decade or so. I am excited to be part of this transformation and will continue to blog about it here.
 Footnote 1: Software-Defined Anything (SDx) Scenario … Beyond Virtualization, Philip Dawson
 The Five-Year Storage Scenario, Alan Dayley and Arun Chandrasekharan