Historically, dedicated silicon, such as an ASIC, or proprietary hardware and “firmware” designs have been the preferred delivery mechanism for purpose-built technology devices such as storage controllers, network switches, layer 4-7 devices, and other datacenter hardware. Control, performance and reliability can be very high, but this often comes at the expense of flexibility, with infrequent updates and rigid interfaces.
A desire to more rapidly deploy (or re-deploy) resources has been one of the major driving factors for the “software defined” movement, with the emphasis in this case placed on orchestration and coordination of resources to aid accelerated (re)deployment and configuration. However the proliferation of cheap, high performance CPU cycles is driving a parallel “sister” software defined movement, which is the migration of basic “hardware” capabilities previously delivered via ASCIs/Hardware/Firmware as “software” running atop a Linux kernel on a standard server.
As technology previously firmly separated in dedicated physical hardware components instead becomes merely modules of code running on the same server, a final new trend is emerging, which is the convergence of these (now) software components into a monolithic stack. The convergence of system software promises both lowered labor costs, increased feature/function and higher performance, but risks vendor lock in.
In this blog, I am going to overview logical components of a Converged Software Defined Environments (SDE) stack that blurs the traditional data center silos or pillars (storage, network and compute) and creates an integrated Data Center OS. With converged SDE, we can create a single, integrated System Software stack to provide:
A Storage Dataplane: This software provides storage functions such as RAID or erasure coding, tiering, read and write caching, snapshots, wear leveling, block/file/object exporters, etc. Key features and functions include data protection, transparent tiering, replication, and globally integrated interfaces. IBM’s Elastic Storage, one of the Software Defined Storage capabilities provides simplified data management and integrated information lifecycle tools capable of managing petabytes of data and billions of files across many types of media, in order to arrest the growing cost of managing ever growing amounts of data.
A Compute Dataplane: Deeply integrated with the storage dataplane, upcoming generations of Kernel-based Virtual Machine (KVM) represents the next generation in Opensource hypervisors and promise unprecedented performance. KVM is fully integrated into the Linux operating system both as a host and a guest and span multiple platforms including x86, IBM Power Systems (PowerKVM) and IBM System z. Unlike other hypervisors, KVM makes no distinction between running in either host or hypervisor mode. This duality in design has helped KVM to rapidly mature into a stable, high performing hypervisor, positioned to outperform other hypervisors available on the market today. IBM's contributions to the KVM hypervisor are consistent with its decade-plus Linux mission and commitment to open technology.
A Network Dataplane: IBM Software Defined Network for Virtual Environments (SDN VE) creates a more responsive network infrastructure by creating virtual networks for dynamic workloads. This virtual network is isolated from the physical network much like a virtual machine is isolated from its host server hardware. IBM SDN for Virtual Environments includes a rich set of APIs for easy integration with orchestration managers.
- A Single Control Plane: Spans across all three cited above, based on IBM SDE solutions in Networking, Platform Computing and IBM Cloud Manager with OpenStack to provision, orchestrate, control, and monitor those three dataplane paths.
Migration of function outside of the hardware will lead to drastic simplification of the hardware and tighter integration with systems capabilities that migrate into the software defined layer. Abstraction of the hardware layer may shift traditional spending towards software differentiation, but still leaves room for unique integration between innovative hardware and system software designed to leverage it, such as the IBM Power 8 processor and the new CAPI (Coherent Accelerator Processor Interface) capabilities to accelerate all of the dataplane components. SDE has become a powerful way of orchestrating all of the core components running within a data center, and as these themselves migrate to become software, convergence will play a role in the final integration and performance of the aggregate “Datacenter OS” stack. IBM is uniquely positioned with best of breed components across almost all categories, and remains committed and consistent in introducing open, interoperable solutions developed with community support towards an integrated vision.
Organizations managing IT infrastructure must choose a path as system software vertically integrates to realize the necessary cost reductions, efficiencies and flexibility required to rapidly deploy networking, servers and storage “Infrastructure” at unprecedented speeds, regardless if the infrastructure is public, private or a hybrid. What are your thoughts? I look forward to hearing from you. Write to me at email@example.com or connect with us on Twitter @IBMSDE.