Composable/Disaggregated Infrastructure Systems: Catalysts For Digital .
IDC TECHNOLOGY SPOTLIGHTSponsored by: FungibleModern workloads demand modern infrastructure. Composable/disaggregatedinfrastructure systems provide the agility and elasticity necessary for deployingcurrent- and next-generation applications at cloud scale.Composable/Disaggregated InfrastructureSystems: Catalysts for Digital TransformationMarch 2021Written by: Ashish Nadkarni, Group Vice President, Infrastructure Systems, Platforms, and TechnologiesIntroductionDigital transformation (DX) is upon us. Companies of all sizes, across allindustries, and on all continents are in the process of transformation withthe objective of maintaining their differentiation in the future. IDC predictsthat by 2022, approximately 65% of GDP will be digitalized. Competitivedifferentiation in the future means newer approaches to customerengagement, newer ways of business process optimization, and the abilityto develop and deliver new products and services at a much faster pace.DX initiatives compel organizations to reinvent themselves and create newrevenue sources while maintaining their current sources of revenue.DX initiatives lead to a slew of next-generation workloads in thedatacenter, each demanding a level of agility and scale that cannot be metby current infrastructure architecture. IT organizations must invest innewer approaches to infrastructure architecture. At the same time, theymust maintain the infrastructure supporting current-generationworkloads, as they in turn support revenue-generating outcomes.Maintaining separate infrastructure for current-generation andnext-generation infrastructure can be an expensive proposition that, if notmanaged properly, can lead to runaway costs, poor service quality, andloss of business.AT A GLANCEKEY STATSDigital transformation will drive demand forcomposable/disaggregated infrastructureworldwide. According to IDC:» By 2022, approximately 65% of GDP willbe digitalized.» The market for composable/disaggregatedinfrastructure solutions will grow at asix-year CAGR of 58.2% and reach 4.7 billion in 2023.KEY TAKEAWAYSNext-generation datacenter infrastructureneeds modern architecture that is flexibleand can scale linearly. Investing in abest-in-class composable/disaggregatedinfrastructure enables enterprises toembrace a common datacenterarchitecture.The emerging composable/disaggregated infrastructure solutions provide the necessary scale, agility, and flexibilityneeded for cohosting current-generation and next-generation workloads on a common datacenter infrastructurearchitecture. According to IDC estimates, the market for such solutions will grow at a six-year CAGR of 58.2% and reach 4.7 billion in 2023. The solutions disaggregate compute and data persistence (storage) elements into pools of resourcesthat can be composed in a software-defined manner and consumed as code. Such approaches are secure and can bedeployed in a connected, hybrid cloud manner at a competitive total cost of ownership relative to all-in public cloudinfrastructure-as-a-service investments.
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationScaling Limitations with Current-Generation InfrastructureDatacenter infrastructure is in a state of transition thanks to DX initiatives at most modern enterprises. Firms aredemonstrating a growing appetite for next-generation workloads as a vehicle for aggressively transforming themselves.Meanwhile, their business operates on current-generation workloads, which require high-quality service.Next-generation workloads are often custom developed and deployed using newer methodologies such ascontinuous integration/continuous delivery (CI/CD) and DevOps. They require that IT operations and developers workas a team to continuously release updates to the application to accomplish a quicker release cadence. Such workloadsare designed to be horizontally scalable, with state maintained in resilient data layer processes provided by theinfrastructure itself. They are designed to run on industry-standard computing platforms and inside lightweightapplication containers and carry no bias toward specific silicon architecture. Examples of applications include mobile,online gaming, artificial intelligence/machine learning (AI/ML), Internet of Things (IoT), and analytics and big data.In contrast, most current-generation workloads assume infrastructure resiliency and require computing technologiessuch as virtualization and clustering that provide portability and transparently preserve application state. They alsodepend on external shared storage arrays that provide multiple layers of resiliency and redundancy. Such workloads havesluggish upgrade cycles that are driven by vendor patches, requiring more planning ahead of time. Examples includeenterprise resource planning (ERP), customer relationship management (CRM), and other enterprise applications thatutilize relational databases.Diverging requirements presented by these two generations of workloads present IT organizations with two options:» To fork their infrastructure and maintain two separate stacks — one for each generation» To maintain a common infrastructure stack for both generationsThe second option is more practical in theory and presents an opportunity for organizations to make the most oftheir investments. There is only one problem. Many current-generation infrastructure solutions use architecturalapproaches that are a few decades old. In the era of DX, these infrastructure solutions present severe scaling challenges,especially as IT organizations choose to run next-generation workloads adjacent to current-generation workloads.These challenges include:» Inflexibility and slow to respond to change. Infrastructure is immutable after deployment. In addition, thedeployment processes are slow and arduous.» Poor utilization. Hyperconverged/direct-attached-type architecture leads to expensive resources being maroonedinto silos, resulting in an increasingly large number of SKUs in order to address the variety of workloads.» Performance degradation with workloads. The addition of workloads to the system results in performance falloffor inconsistency. This is due to the overhead of a software-based hardware abstraction layer such as a hypervisor.» Simplicity not possible for some workloads. For newer workloads such as big data, AI/ML, and cloud-nativeapplications, organizations may want virtualization-like simplicity but cannot afford to trade performance for simplicity.» Security threats that cannot be intrinsically tackled. Workload isolation is needed to protect against escalating threats.#US47518821Page 2
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationFigures 1 and 2 illustrate key IT objectives when modernizing infrastructure. Saving on IT infrastructure costs andincreasing employee productivity are the top 2 objectives for organizations when deploying modern IT technologies fortheir business-critical workloads. IT modernization and IT automation are the top 2 initiatives for investments in 2021.FIGURE 1: IT Objectives for Business-Critical WorkloadsQ What value does your organization see in using IT modernization technologies for itsbusiness-critical workloads?Saving on IT infrastructure costs40.2Increasing employee productivity35.6Enhancing customer experience29.2Free IT staff time to focus on new business innovation/growth28.9Responding more quickly to new business needs28.7Driving more business value28.2Becoming more competitive in our market27.9Enabling new business opportunities23.7Delivering new products to market19.4Don't see much/any value1.2Don't know0.70510152025(% of respondents)30354045n 636Source: IDC's IT Infrastructure Plans for 2021 Survey, December 2020FIGURE 2: Key Areas of IT InvestmentsQ What initiatives will drive investments into IT infrastructure in 2021 in your organization?IT modernizationIT automationData protection (backup, recovery, etc.)AI (machine learning or deep learning)Digital transformation (DX)General IT refreshBusiness continuity/disaster recoveryNet-new or additional investment in public cloud infrastructure and/or servicesImplementation of new workloadsHigh-performance computing (HPC)Deployment of private cloud environmentEdge environments051015202530(% of respondents)35404550n 636Source: IDC's IT Infrastructure Plans for 2021 Survey, December 2020#US47518821Page 3
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationIT organizations need another option that builds on a consolidated infrastructure approach but uses a different set ofbuilding blocks. Composable/disaggregated infrastructure addresses key objectives such as modernization andautomation, which in turn drive increased productivity while reducing costs.Composable/Disaggregated Infrastructure Is a New Infrastructure ParadigmThe emerging category of composable/disaggregated infrastructure refers to systems that make use of high-bandwidth,low-latency connectivity (such as high-speed Ethernet) and special silicon engines called function-offload accelerators (FAs)(also called data processing units or DPUs and SmartNICs) to disaggregate compute, storage, and networking resources intoshared resource pools, which in turn can be allocated to workloads on demand in a software-defined manner.These systems employ a scaling architecture to break down the physical boundaries of the server systems, enablingresources inside each node to be available for on-demand allocation. Individual IT assets are aggregated into fluidlymanaged resource pools that can be dynamically provisioned and deprovisioned via a common application programminginterface (API). True composable/disaggregated infrastructure requires "composable" and "disaggregated" platforms towork in unison with each other.Disaggregation occurs in hardware and includes the ability to decouple "system-level" component resources such asCPUs, memory, storage, and peripherals into subsystem "pools" that operate autonomously. Such resource pools canthen be virtually sliced from commodity off-the-shelf servers and spliced on demand to create complete hardwareplatforms via APIs — such aggregation being controlled at an embedded firmware level and supported by FAs andhigh-bandwidth, low-latency fabrics.Composability is software defined. The ability to create a set of virtually consumable infrastructure entities from fully orpartially disparate/disaggregated/autonomous resources enables workloads to be deployed in a bare metal, virtualized,or containerized manner. Control software accesses hardware resources via a unified set of low-level APIs.As Figure 3 illustrates, taxonomically, IDC considers composable/disaggregated infrastructure to be the next phase in theevolution of converged systems, an overlay infrastructure. The first generation of converged systems comprisesintegrated infrastructure; the second generation consists of hyperconverged infrastructure (as illustrated on the left-handside of the figure); and the third generation comprises composable/disaggregated infrastructure, which is borne out ofthe limitations posed by first-generation and second-generation converged systems.#US47518821Page 4
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationFIGURE 3: Composable/Disaggregated Infrastructure Follows Hyperconverged InfrastructureSource: IDC, 2021Modern composable/disaggregated infrastructure solutions are powered by platform-independent FAs that can beinstalled in commodity servers as PCIe cards or integrated into the motherboard. The solutions are designed to functionindependent of the operating system instance and the CPU. A key differentiator of platform-independent FAs is that theyare not dependent on the CPU architecture itself (i.e., the same FA can be installed in servers with x86 or ARM-based CPUs).FAs offer scaling approaches that overcome current limitations of computing platforms. FAs are designed to operateindependently of the CPU and outside the control of main processing system from a code execution perspective. In otherwords, the CPU is aware of their presence but cannot control them. FAs enable:» Decentralized or distributed computing platform architectures for embedded and management function offloads.This includes data storage and networking services, security services, and even some compute offload services(such as clustering and failover).» An additional abstraction layer in the boot-up and operational state of the platform to control access to physicalresources such as data persistence and network interfaces through which sensitive data can be accessed.» The ability to offload or take over direct execution of crucial portions of the embedded and management payloadfunctions such as security, hypervisor root partition, and virtual networking.Payloads that FAs can handle include:» Embedded payloads that are part of the core operating system and often run as privileged operations in kernelspace (Compute, storage, and networking functions lie here.)» Management payloads that straddle embedded and user space payloads and that are gaining importance incloud-native and scale-out deployments» User space payloads (i.e., programmatically assigned portions of workloads that make up an application and itsaccompanying data set); not the primary function but can be tooled to do so#US47518821Page 5
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationFor function-accelerated composable/disaggregated infrastructure approaches to be successful, the software stack mustbe as robust as the hardware stack and be tightly coupled. The software stack must support:» Delivery of full offload of embedded and management functions in a distributed manner» Disaggregation of the virtualization layer, wherein privileged hypervisor functions are offloaded» Multiplatform hardware and heterogeneous computing, allowing buyers to mix and match CPU and acceleratortypes with a common accelerator fabric controlling the cluster state» Compute, storage, and networking composability relying on hardware with varying levels of disaggregation foreach resource pool (The stack must tightly couple these resources for scaling purposes.)» Bare metal and virtualized workloads (Bare metal can then support direct or containerized workloads.)Modernizing Infrastructure with FungibleFrom an IDC taxonomy perspective, Fungible Datacenter is a composable/disaggregated infrastructure system. It is acomprehensive, multitenant, secure, and turnkey solution delivering bare metal performance at the cost efficiencies andsimplicity of hyperscale datacenters.Powered by Fungible Data Processing UnitFungible Datacenters are powered by Fungible's own data processing unit — a specialized programmable processor(which IDC refers to as an FA) that is designed to address performance, utilization, and scaling challenges with today'sinfrastructure such as:» Inefficient execution of data-centric computations within a server (Data-centric computations refer to storage,networking, security services.)» Inefficient movement of data within a pool of compute and storage servers» Trade-off between programmability (or flexibility) and performanceThe Fungible DPU is capable of:» Executing data-centric computations an order of magnitude more efficiently than general-purpose CPUs incompute and storage servers, enabling each server to deliver its resources at a near optimal point in the trade-offbetween performance, cost, and power.» Implementing the end point of a high-performance network fabric (called TrueFabric) that provides deterministiclow latency, full cross-section bandwidth, congestion and error control, and high security across a wide scale ofdeployments. This capability enables datacenter-scale composition of the disaggregated compute and storageservers at datacenter scale.» Offering true data-path programmability through an architecture that allows fine-grained feature and performanceoptimization for different data-centric computations. Programmability is offered at the infrastructure softwarelevel using high-level programming languages, and no changes in application code are necessary.#US47518821Page 6
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationThe Fungible DPU is versatile and can be deployed in a multitude of use cases, including storage (initiator as well astarget), AI/analytics servers, Layer 4 to Layer 7 systems, virtualization, security appliances, and NFV applications.Components of a Fungible DatacenterA full rack Fungible Datacenter system comprises key hardware and software platforms that work in unison with each other:» Standard servers with Fungible Data Services PCIe Function Accelerator Cards. These cards support 2x100, 2x50 or2x25 Gbps (PCIe Gen3 or Gen4). They offer bare metal workload support, with an offloaded hypervisor data pathand a complete offloaded networking, storage client, and security stack. This frees up nearly 50% of CPU cycles.Workload accelerator support is available via a virtual PCIe switch that enables access to a pool of GPU resources.» An NVMe-based all-flash storage platform provides disaggregated multitenant storage with advanced datareduction and protection. The platform can offer up to 15 million read IOPS and 6 million write IOPS (4KB chunks).Performance scaling is linear, while all data is compressed, encrypted, and erasure coded across multiple platforms.» Industry-standard "top of rack" switches for data, control (BMC), and management traffic. The BMC network isused for gathering server telemetry, providing a bare metal console to the server, and performing firmwareupdates. The boot is from locally presented NVMe drives.Fungible Datacenter is powered by Data Center Composer (DCC) Software — a datacenter-scale multitenantbare metal virtualization solution that provides composition, provisioning, management, and orchestration of hardwareresources. The solution provides:» Bare metal composition. IT administrators can compose a bare metal server instance with the desired amount ofdisk (flash), accelerator, and network resources. Furthermore, a real-time load balancing feature enables the mostoptimal resource utilization.» One-click deployment of apps. IT administrators can manage and deploy applications in minutes using prebuilttemplates from the application marketplace. The DCC interface plugs seamlessly into existing DevOps stacks.» Role-based access control. Two levels of IT administrator roles (one for the overall datacenter and one for eachcustomer) support independent authentication and management planes, enabling seamless customer onboarding.The DCC plugs into existing identity management systems.» Multitenancy and customer isolation. IT administrators can partition servers, storage, GPUs, and networkingresources to customers. Servers and GPUs are hard partitioned, while storage and networking resources areallocated from pools. Resources can be rebalanced to partitions in real time and true self-sufficiency via a tenant'sindependent management portal and APIs.» Frictionless network integration. IT administrators can leverage existing infrastructure to extend investment.Virtual networking functions and hardware-accelerated network devices such as load balancers and firewalls areprovided via built-in network controllers. These built-in network controllers can be bypassed with a customerprovided network controller without loss of other composer functionality.» Management of all datacenter assets. IT administrators can discover, register, describe, and decommissionhardware assets through single-pane-of-glass management software. The software can monitor telemetry data,manage failures, manage firmware updates, and maintain awareness via notifications.#US47518821Page 7
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationFungible Datacenter is a comprehensive, turnkey infrastructure solution that can be used as a common foundation forthe entire datacenter. It provides:» Simplicity. IT can deploy and manage turnkey multitenanted datacenters with single-pane-of-glass management,enabling deployment and management of complex scale-out workloads without any application changes.» Agility. To handle workload hotspots, IT can redeploy compute, storage, and network resources across workloadsin minutes.» Performance. Bare metal performance for data-centric applications enables full offloading of data-centric I/Oprocessing to the Fungible DPU, which in turn frees up cores for applications on the host side.» Scalability. IT can scale to hundreds of racks to handle the most complex datacenter deployments.» Economics. Reducing server SKUs to a minimal set enables IT to gain economies of scale and managementsimplicity. Disaggregation and pooling of server, storage, network, and GPU resources enable higher utilizationunder changing workload demands. Just-in-time composition of independent compute, storage, network, and GPUresources optimizes consumption to exactly meet workload requirements with no wastage.» Security. Independent hardware-accelerated security domains provide fine-grained segmentation and deliverrobust quality of service (QoS) with line rate encryption.ChallengesFrom an architecture perspective, any composable/disaggregated infrastructure is made up of two parts: The first part isthe ability to (dis)aggregate IT resources into compute, storage, and fabric pools, and the second part is the ability tocompose consumable resources from such disaggregated pools via a unified API. Therefore, intelligent software isneeded to manage all the distinct assets and to compose the optimal configuration for a specific application. All theelements inside the installed infrastructure are pooled and require management, including discovery and location ofresources. Monitoring and life-cycle management software is necessary to provide full awareness of the hardware assetsand application workloads, including the following:» Self-discovery, provisioning of workloads, orchestration, and healing. It will be essential to have visibility into theutilization of the discrete resources and understand the load on the resource elements.» Orchestration layer. It will be critical to enable a catalog of compute, storage, networking, and memory in anorchestration layer and define resource requirements for specific applications and compose them in a set that isoptimized for the workload.Composable/disaggregated infrastructure necessitates infrastructure analytics (i.e., data metrics are needed tounderstand how to optimally configure hardware for the applications). It is essential to have visibility into the utilization ofthe discrete resources and understand the load on the resource elements in order to answer questions such as:» How much is the application consuming?» Which resources (such as processor and storage) are reaching thresholds or are underutilized?#US47518821Page 8
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationThese data metrics also provide valuable feedback to enable orchestration and template provisioning.The dynamic nature of application capacity means manual processes, which consume time and are a source of potentialerrors, must be reduced as much as possible.IDC sees these as necessary factors for composable/disaggregated infrastructure systems to be successful in theenterprise — in accelerating provisioning times, improving IT utilization, and simplifying overall IT operations.For system vendors such as Fungible to emerge as leaders in this space, they must work with businesses to alleviate theirconcerns with composable/disaggregated infrastructure. Concerns about technical maturity and reliability, the ability ofstaff to manage a new platform, doubts about promised ROI, and difficulty integrating with existing tools are top of mindfor many IT decision makers.ConclusionConcepts of disaggregating IT systems down to discrete resources have beendiscussed in the industry for some time, even though the enabling technologies haveyet to mature. DX and demand for next-generation workloads driven by shifts inbusiness strategy will fuel the demand for composable/disaggregated infrastructure.This emerging category of systems seeks to create greater business agility, loweroperational costs, and increase application performance. Businesses should stayopen to embracing these systems provided concerns such as ease of managementand promised ROI are addressed.Digital transformationand demand for nextgeneration workloadsdriven by shifts inbusiness strategy willfuel the demand forcomposable/disaggregatedinfrastructure.About the AnalystAshish Nadkarni, Group Vice President, Infrastructure Systems, Platforms, andTechnologiesAshish Nadkarni is Group Vice President within IDC's worldwide infrastructure practice. He leads a team ofanalysts who engage in delivering qualitative and quantitative research on computing, storage, and datamanagement infrastructure platforms and technologies via syndicated research programs (subscriptionservices), data products (IDC Trackers), and custom engagements.#US47518821Page 9
IDC TECHNOLOGY SPOTLIGHTComposable/Disaggregated Infrastructure Systems: Catalysts for Digital TransformationThe content in this paper was adapted from existing IDC research published on www.idc.com.IDC Research, Inc.140 Kendrick StreetBuilding BNeedham, MA 02494, USAThis publication was produced by IDC Custom Solutions. The opinion, analysis, and research results presented herein are drawn frommore detailed research and analysis independently conducted and published by IDC, unless specific vendor sponsorship is noted. IDCCustom Solutions makes IDC content available in a wide range of formats for distribution by various companies. A license to distributeIDC content does not imply endorsement of or opinion about the licensee.F 508.935.4015External Publication of IDC Information and Data — Any IDC information that is to be used in advertising, press releases, or promotionalmaterials requires prior written approval from the appropriate IDC Vice President or Country Manager. A draft of the proposeddocument should accompany any such request. IDC reserves the right to deny approval of external usage for any reason.Twitter @IDCCopyright 2021 IDC. Reproduction without written permission is completely forbidden.T US47518821Page 10
Composable/disaggregated infrastructure addresses key objectives such as modernization and automation, which in turn drive increased productivity while reducing costs. Composable/Disaggregated Infrastructure Is a New Infrastructure Paradigm The emerging category of composable/disaggregated infrastructure refers to systems that make use of high .
Why Composable Infrastructure? Composable Infrastructure helps IT to respond to the challenges laid down by new organizations. It can be considered as the next phase of hyper-converged infrastructure or the next version of on-premise infrastructure. Composable Infrastructure consists of multiple resources including, but not limited to, compute,
In 2015, Cisco introduced its first set of Cisco UCS composable infrastructure solutions. Cisco's approach to composable infrastructure builds on their converged infrastructure portfolio. It does so by taking advantage of the combined compute, network, and storage resources, by decomposing the infrastructure resources, and by leveraging the
HPE Composable Infrastructure in action HPE Composable Infrastructure lets IT administrators and developers use infrastructure as code to control their internal environments. The unified API integrates dozens of popular management tools. Here's an example of how it works using Chef , a powerful automation tool that enables the
Introducing the first Composable Infrastructure HPE Synergy, the first platform built from the ground up for Composable Infrastructure, offers an experience that empowers IT to create and deliver new value instantly and continuously. It is a single infrastructure that reduces operational complexity for traditional workloads
the PGM-free and PGM catalysts accounts partly for the substantially lower power density delivered by PGM-free catalysts in practical H 2-air PEMFCs ( 0.57 W·cm2) (4) than that of PGM catalysts ( 1 W·cm2). The most active PGM-free ORR catalysts are pyrolyzed transition metal-nitrogen-carbon (M-N-C, M Fe or Co) catalysts (4-10). This group .
Composable Infrastructure with ACI and Terraform. About Me Sess2 5 years at Cisco All things OSS and ACI Big focus on Cloud Native Automation junkie Love coding Double VMware VCDX CCIE Groove Metal addict. . Cisco Public Composable Infrastructure
Build a single, consistent hybrid IT experience based on Composable Infrastructure Private cloud Public cloud Single platform, Software Defined, Multiple Clouds Consistent and simple to operate Compatible across on and off premises Ready for VMs and containers Fluid Resource Pools Software-Defined Intelligence Composable Infrastructure Unified API
The manufactured configurations were tested according to the ASTM D 3039 and ASTM D 4255 for the in-plane mechanical properties and according to ASTM G 99 for the friction coefficient. Also, specific wear rate and through-thickness compression test were performed according to [29]. The effects of the MWCNTs on the composite were determined from the tests results. Experimental description .
