Hybrid Cloud: Why Converged Modular Servers are just right for IoT Infrastructures
As mentioned within my last blog post, nearly every important market researcher and analyst agrees to the following: hybrid technology will be one of the main drivers of the digital transformation. The reasons for this are obvious: Big data, the Internet of Things (IoT), and mobility will most likely lead an explosion in the use of hybrid clouds. Because as the number of sensors, applications, mobile devices and the use of social media increases, there will also be a tremendous growth of mobile, structured, and unstructured data. The hybrid cloud will provide an ideal platform for these major trends.
Currently there is no question, that hybrid cloud environments offer a better and safer solution for end-to end enablement of IoT deployments than the public cloud. But in the course of this, companies now have to face the question which architecture will bring better scalability, improved rackspace density, reduced power consumption/cooling requirements and flexibility. A new architecture which is driving the scalable IoT hybrid cloud is the converged modular server. This new breed of commercial-off-the-shelf (COTS) modular server hardware incorporates the best characteristics of rackmount and blade servers, integrating high-density compute, storage, and network switching into one platform.
Target IoT Application Needs
Compared to a traditional server using a scaled-up processor such as the Intel Xeon processor E5 with up to 18 cores, converged modular servers are reaching into the many core realm to deliver denser, more flexible compute resources. For example, the Kontron SYMKLOUD Series currently packs up to 36 Intel Xeon processor E3 cores or up to 72 Intel Core i7 processor cores in a 2U enclosure – leading to perhaps 756 or 1512 cores in a 42U rack. Near term, these new processors will push core counts over 100 per 2U. And of course, this framework introduces many new possibilities for partitioning a server for an optimized IoT infrastructure use. Here are some of the most relevant advantages:
- By using lower-power multicore processors, modular servers can reduce overall power consumption and space while enabling scale-out performance increases within and between platforms. So rather than splitting a large workload across several cores, the converged server approach is best for many threads of execution.
- Converged modular server architecture connects multiple processors via a fast, low latency, configurable fabric within a single enclosure. Integrated fabric enables low-latency load balancing between processor cores.
- An Ethernet-based platform can drop into an existing enterprise network, distributing and shaping traffic.
- Redundant, integrated IP switching provides more control over latency and quality of service, and guarantees each node its allocated bandwidth. Traffic shaping also segregates management traffic from data, providing continuous network integrity under all loading conditions.
- Instead of a unified design, many converged modular servers are implemented in a compact rackmount chassis, with self-contained modular server subsystems each having their own fabric connection, cooling, and monitoring. Adding or replacing tool-less subsystem modules with the platform powered on makes for easier maintenance and upgrades.
- Some modular servers manage power on a per-server basis. Careful task partitioning means processors can sleep, or even completely power down, until needed.
And a hybrid cloud approach to IoT infrastructure featuring modular servers also helps free application developers from many other boundaries:
- On a private modular server, virtualization environments are fully controlled – everything running on the platform is known. This opens the potential for network functions virtualization (NFV), reducing both capital expenditures and lifecycle costs as appliances are consolidated.
- Developers can select and deploy the right software approach. Operating systems can mix and match between real-time, Linux, or Windows. Installing and configuring middleware such as OSGi, DDS, MQTT, or other services is unconstrained on a private platform in the hybrid cloud.
- Many-core processing and fabric interconnect mean the ability to ingest and process more IoT data and provide faster and more detailed analytics. This can lead to improvements or cost reduction in existing services, or incremental revenue from new service offerings.
- The highest potential of the IoT lies in monetizing analytics and data, and creating powerful new business models. With a hybrid cloud in place, data assets can be developed and shared or fully protected. Ability to grow and adapt the infrastructure quickly means improved competitiveness.
If you want to learn more about converged modular servers, you can also read the Kontron Whitepaper “Powering a scalable, hybrid cloud for IoT infrastructure”.
Are you already modernizing your in-house applications and making them “cloud ready”? And if so, are there a lot of technical and compliance-related hurdles to overcome? Tell us about your experiences with hybrid cloud environments (maybe even converged modular servers) by leaving a comment!