Does the candle burn at both ends? Cisco says yes, and in what Paul Perez, Cisco UCS GM, described as the "most sign technology announcement Cisco has made this year," announced a new operational model for IoT computing at the edge and at the core. Cisco’s launch of "edge-scale" and "cloud-scale" computing platforms can be viewed on one hand as an effort to halt mid-swing the pendulum that has had compute paradigms alternate between centralized (mainframe) platforms, distributed (client server) platforms, centralized (cloud) computing and back again with to local, distributed processing. The launch also represents Cisco work to extend its developing application focus.
Introducing the new operational model in New York last week, Cisco CTO Padmasree Warrior outlined the drivers for this new take on the company’s Unified Computing System (UCS): according to Warrior, increasing reliance on IT as the source of innovation and to buttress an organization’s competitive positioning means that "every company will become a technology company." At the same time, data-intensive, IoT enabled applications are demanding "Fast IT," infrastructure characterized as "simple, secure, and programmable." Cisco has been building this type of infrastructure for five years now, she added, through UCS Director management technology that delivers a converged environment to prepare data centres for cloud computing, through FabricPath software innovation that has pushed network speed in the Nexus networking portfolio, and through ACI, a policy driven network architecture and management platform that enables an app centric view of data centre resources. Adopting an "ecosystem approach," Warrior explained, Cisco has worked with partners to deliver this integrated UCS infrastructure in consumable bites — with VCE in Vblock and with NetApp in FlexPod — for rapid IT deployment. But with the emerging sensor-riddled, IoT future have come new requirements for management of Very Large Data sets at the data centre core, as well as processing at the local level, as close to the source of the data as possible. To address this need, Warrior explained, Cisco has embarked on the next stage in its journey: "edge scale" and "cloud scale" computing.
But what exactly do these high level concepts mean in the nuts and bolts language of data centre infrastructure? Specifically, "cloud-scale computing" consists of new Cisco UCS M-Series Modular Servers that Cisco claims deliver "breakthrough levels of operational efficiency" for CSP and enterprise customers, which are increasingly reliant on application architectures that can scale. Based on Cisco System Link Technology, the UCS M-Series features a 3rd generation virtual interface card that extends UCS fabric into the server, shared local networking and storage (4 SSDs in the chassis and 40 Gb connectivity), and improved compute density with 2 16 Intel Xeon Compute nodes in each cartridge. According to Cisco, this new "disaggregated server architecture" can eliminate 95% of the system components duplicated in a traditional server design, which, when combined with UCS management, can deliver up to 36% combined TCO savings. The outcome is new levels of cloud scale.
At the New York event, Cisco also announced a series of server product and management updates, including: a new Cisco UCS C3160 Rack Server with high-capacity local disk storage designed for distributed data analytics and object stores, unstructured data repositories, and media streaming and transcoding; the UCS B200 M4 Blade Server, C220 M4 and C240 M4 Rack Servers, based on the future Intel® Xeon® processor E5-2600 v3 product family; as well as the UCS Director Express for Big Data, which automates deployment of Hadoop to enable repeatable Hadoop UCS cluster configuration and which is integrated with major open-source distributions to provide single pane of glass management across physical infrastructure and Hadoop software.
A consistent note in Warrior’s presentation was that IT infrastructure must evolve with applications, which in the world of IoT means processing of data at the edge, and the transmission of refined results rather than raw data from the device to the data centre core. As Sathinder Sethi, Cisco VP engineering for UCS and data centre solutions, noted at the New York launch, 50 percent of computing that now goes on is done outside servers in the data centre. To address this new circumstance, Cisco has expanded UCS outside the data centre with the UCS Mini, a new modular platform that combines Fabric, storage, servers and UCS management in a compact form factor — to produce a 36 percent TCO as compared to traditional rack servers.
Enumerating the potential use cases for the new Mini, Sethi explained that this all-in-one architectural solution will allow large scale IT departments to cover branch plant operations: the goal is to extend UCS management and policy control across distributed operations, introducing the operational standardization that enables application scale. At the same time, the Mini is designed to empower the small scale IT shops that may not have the in-house IT skills needed to manage increasingly complex IT needs.
Ultimately, the aims of the Mini platform and the new modular server series appear to have more to do with the current reality of data centre operators (Big Data and application scale) than with IoT, and with Cisco ability to meet head on some of the converged infrastructure solutions targeted at high end and mid-market on offer by competitors — VMware’s EVO Rails is only a recent example, which may begin with HP’s converged systems. But with "core" and "edge" concepts, Cisco is positioning for the IoT opportunity, creating a foundational platform that itself can develop and scale to meet a hyper connected future.