Software-Led Infrastructure: The Evolution Of Software-Centric IT Operations
A new era of datacenter infrastructure is emerging that will result in operational and application changes on a scale that we have not experienced in more than 15 years. At the core of this change is a shift to increasingly software-centric IT operations with the virtualization of the majority of hardware devices across the IT value chain. Pervasive virtualization across compute, storage, and networking technologies means that the point of control for the datacenter will exist in software, not hardware components. This fundamental shift in datacenter architecture will have the following primary implications:
- The datacenter will be increasingly dynamic (rather than static), with resources deployed, managed, and better aligned with rapidly changing business requirements.
- Management of devices will be moved to a single point of control to synchronize tasks that are executed across multiple IT environments to fully exploit the benefits of dynamic resource allocation.
- IT operations will be highly automated, greatly improving efficiencies and the ability to dial resources up and down as required while providing the ability to quickly make changes to SLA’s at an application level.
- Process change becomes just as important as new IT adoption. IT organizations will have to work more closely with lines of business to encapsulate SLA requirements at the application level. The result is a set of autonomic systems that can intelligently balance application performance and availability against defined costs and business priorities.
- Applications will be presented as “services” within a catalog of IT resources directly available to the IT budget holder.
The objectives of IT organizations in this new era will not only be to deliver greater levels of efficiency and agility, but also to integrate with the objectives of business lines more deeply. Moreover, IT organizations must endeavor not simply to “compete” with external service providers (e.g. IaaS, PaaS and SaaS offerings), but rather to deliver sustainable competitive advantages that complement external services and deliver clear business value.
What’s Really Different About A Software-Led Datacenter?
This notion of a services-driven datacenter is not new. There have been many iterations of an agile datacenter that can shift resources as business demands change. Today many cloud computing solutions are sold based on this premise. So what’s really different that sets the next five years apart from the last 50? Four major elements differentiate a software-led data center summarized in Figure 1 and detailed below
- Pervasive Virtualization and Encapsulation: All key resources in the software-led data center are virtualized, including processors, storage, and networks. This pervasive virtualization allows dynamic distribution of workloads across a wide set of resources to meet availability and performance SLAs. Applications, management services and hardware services are codified and orchestrated to provide for IT resources (only) as needed.
- Software Defined Networking: This will include user, storage, and processor-to-processor networks. Networking has been a bottleneck for IT transformation; resolving this is critical to enabling next generation datacenters. By virtualizing the network and moving network control into software, traffic can be prioritized and dynamically managed to guarantee predictable end-to-end application performance and latency, even as system resources, demands on the system, and SLAs change. Networking hardware must be a reliable platform on which software services can be deployed. Management of the network will shift from today’s network-specific silo approach to a datacenter approach.
- Unified Metadata about System, Application, & Data: The future datacenter will work only if all the individual components can communicate and interact in a synchronized manner. Unified metadata for datacenter devices, services, and applications will be a vital element to manage the infinite number of connections and exchanges between systems. For example, as a new virtual machine (VM) is deployed, storage devices will need to understand proper data placement, while network devices will need to set priorities based on IO requirements and application value. Software-led components will be constantly generating, ingesting, and analyzing data about themselves and other components. This management can only be achieved with a metadata structure that is deployed along with common APIs, protocols, and standards.
- Distributed High Performance Persistent Storage: Persisent storage — storage that survives a power outage — was until recently held on low-performance magnetic disk drives and/or tapes. With the introduction of flash, high performance persistent storage at equivalent or lower cost will be distributed throughout and between data centers. Data-in-flash will be directly addressed by the processor, be near the processor, be shared between processors and be distributed geographically.
- Increasingly Intelligent Data Management Capabilities: Sophisticated software-led management for active and passive data and metadata will ensure integrity, availability, protection from disaster, device optimization, and location optimization. Storage and data management services will become more device independent. Efficient storage management will be a continuing theme, with an emphasis on end-to-end capabilities across the infrastructure hierarchy. Storage space and additional unnecessary copies of data will be reduced by end-to-end storage reduction technologies, such as compression and de-duplication. Magnetic media will be used as bulk low-cost store for inactive data.
Defining Software-Led Infrastructure
Software functionality will increasingly drive business value for infrastructure investments delivered as services layered on top of standardized hardware. While this has been a trend for the last five years, unified metadata management structures will emerge and evolve to deliver new levels of business capability. In particular, software-led infrastructure (SLI) refers to datacenter technologies that form the underpinning for application portability & functionality, data placement, and automation within and between datacenters. SLI technologies have a design point with the aim to synchronize with other datacenter technologies and optimize for:
- Speed, Performance or Time to Market,
- Availability and/or Business Continuity,
- Efficiency and Cost Reduction,
- Business and IT Productivity.
Software-led technologies will serve to minimize hardware differentiation in favor of software-based service delivery. Software-led Infrastructure differs from other infrastructure solutions in that:
- The majority of technology value is in software integration that manages for seamless operation of different datacenter technologies. Multi-component awareness across the entire datacenter is a core differentiator from traditional standalone components and a fundamental design point of SLI.
- SLI leverages standard APIs, protocols, and definitions that are primarily driven by open source communities or broadly sanctioned industry consortia.
- SLI maximizes the use of commodity high-volume components where possible and minimizes the requirements for specialized or standalone hardware. In doing so, hardware and software leverages the volume economics of consumer technologies, the “consumerization” of IT.
- Unified metadata management is a core requirement. SLI must have the ability to provide, maintain and utilize metadata so that data and devices are available for use by any process anywhere.
- SLI technologies are architected with a scale-out model, rather than scale-up, as the default.
- A key requirement for software-led infrastructure is “playing nice”, maximizing product choice to infrastructure and application architects, and practitioners.
Key Disruptive Technologies
The key disruptive technologies that will drive and be assimilated into SLI solutions include:
- Flash-based persistent storage will continue to greatly reduce latency and costs for data access and update, leading to new architectures for file systems, databases and applications (see Designing Systems and Infrastructure in the Big Data IO Centric Era. Together with persistent metadata, Big data transactional and analytic applications will provide significant improvement in business productivity and the potential to provide new business opportunities.
- Software Defined Networking (SDN) will include user, storage, and processor-to-processor networks. By virtualizing the networks and moving network control into software, traffic can be prioritized and dynamically managed to guarantee predictable end-to-end application performance and latency (see Networking Revolution: Software Defined Networking and Network Virtualization).
- Software-led Storage (SLS) will allow the migration from fixed storage components (storage arrays and direct access storage (DAS) in servers) where each component provides storage services to a topology where the the storage services are provided at the system level and can be called and integrated into system level storage management and system level SLI management.
- Management and Automation Software will be controlled by services interfacing with APIs from hardware and software components, rather than the current norm where each infrastructure component has its own set of management software that is siloed and independent from other resources.
- End-to-end Security relies on protocols and mechanisms implemented at the endpoints of a connection. This allows security to be defined within an application or system service. The endpoints will continue to need traditional security mechanisms, again built on industry standards.
- Low-cost Mobile Devices are replacing PCs as the predominant source of data access and data creation, creating significant change and challenge for data center infrastructure and data management. The appeal of mobile devices is the interface. However the real power comes from the backend networks and data stores and the ability to rapidly access data services.
- New generation Converged Infrastructure, pre-tested building blocks based on open and/or consortia standards that can “Lego-fit” with other building blocks (see Wikibon’s Converged Infrastructure Market Forecast and Solution Primer).
- Modular Datacenters that focus on dramatically improving PUE.
- Low-Power servers (at the moment from ARM-based or Intel Atom chipsets), which enable greater computing density without the need for high-performance and high-cost cooling infrastructure (see an example of this in Wikibon’s coverage of HP Project Moonshot).
The Business Case For Software-Led Infrastructure
The key elements of a business case for the SLI journey are:
- Decrease the entire operational people component of overall system management cost by data center automation. IT labor costs account for more than 60% of spending in many IT organizations – this expense must be attacked to deliver better agility.
- Reallocate the spend on systems management software from many specific vendor management solutions tied to hardware (e.g., replication in storage arrays) to fewer, hardware-agnostic, datacenter-wide, systems-management solutions built to industry standard APIs. This trend will probably decrease software unit costs, but more importantly will allow much faster time-to-change and time-to-implement application software.
- Enable tighter SLA management and decrease overall costs through the ability to measure end-to-end performance and availability of all system components (especially networking and storage components).
- Create business value through a schema-less storage architecture and network that significantly enhances the value of new and updated applications through flexible real-time access to all Big Data, enabled by (and only possible with) unified metadata and high-speed storage and networks. What data will be needed, and how it will be used, will not be known when the original application is designed.
Practitioners should keep in mind the following factors when making datacenter infrastructure decisions:
- SLI will impact every hardware component and infrastructure decision of the datacenter. Systems management software will be abstracted to a higher level, either replacing or overriding hardware-based functionality. Look for vendors that open up all their hardware-based data to API’s to ensure that the technology you purchase in the near future can be reused as part of an SLI architecture.
- SLI is particularly impactful to existing Cloud-based strategies and solutions, as it provides much of the functionality that sophisticated cloud offerings provide today. SLI is an opportunity to rethink both private- and public-cloud investments, both for company-owned and provider-enabled datacenters. Cost comparisons against existing service providers will be a critical element to cloud considerations, and the choices of service providers for CIOs and CTOs must include the ability to share meta-data and own the IP of meta-data.
- SLI should be thought of as key enabler for existing Big Data projects today, not a replacement. However, the evolution of both markets is very nascent and as such, careful consideration should be made periodically to make sure both strategies are aligned and to watch for potential areas of overlap or inconsistencies in architecture. At the end of the day, both architectures fundamentally depend upon access to metadata to be successful, so cross-leveraging these projects is important but not critical just yet.
- Work with ISVs that have adopted SLI architectures as part of their future roadmap. Working with software providers that cannot leverage SLI will severely limit your ability to deliver competitive IT services. Accessing data and providing real-time capabilities to your software will be the underpinning for many new business development activities over the next five years.
Action Item: Software-led infrastructure is a game changer for businesses and organizations, on the same scale as Internet was in 1995. CIOs and CTOs must move quickly to perform an assessment of their existing IT strategy that focuses on the business impact of an SLI architecture, particularly as it relates to new projects and applications where access to metadata can significantly enhance application functionality and analytic insights. Cost savings from commodity hardware and a reduction in manual tasks by administrators will be an important element of the business case, however it is the ability to build new applications and access data in real time that will lead to significant business value delivery.
Footnotes: This definition of Software-led Infrastructure is the first in a series of Wikibon research pieces designed to assist our members in understanding the implications and opportunities of this dramatic evolution. Please contribute with comments and/or hit edit and improve!