Early Holiday Present: Composable Infrastructure with OpenStack
By Figen Ulgen, General Manager, Intel® Rack Scale Design
Here in the Rose City—Portland, Oregon— the Thanksgiving Day turkeys have been consumed, the holiday decorations are up, and we’re flying toward the end of 2017! This was a whirlwind year in many ways, especially for Intel® Rack Scale Design (Intel® RSD). Our ecosystem partners continue to announce new solutions based on the Intel® RSD architecture – they’re using it as a blueprint on their path toward a fully composable and disaggregated infrastructure – and we’re seeing a lot of interest in Intel® RSD from the open source cloud community as well. That’s no surprise since Intel RSD is a natural complement to virtualized cloud environments, which are all about flexibility, speed, rapid scaling, efficient use of resources and advanced management automation.
Red Hat Enables Redfish APIs in OpenStack, Paving the Way to Intel RSD
A case in point is the Red Hat* announcement that OpenStack Platform 12* (OSP 12) now includes support for the Distributed Management Task Force’s* (DMTF) Redfish* open APIs for composable infrastructure. This allows OSP 12 to interoperate with vendor solutions that implement the Intel RSD architecture, which is based on Redfish.
Intel® RSD Composition and Bare Metal Provisioning with OpenStack
Intel Enables Composition and Bare Metal Provisioning in OpenStack for Intel® RSD Compatible Products
In collaboration with the OpenStack* community, Intel has enabled hardware resource composition, bare metal provisioning and workload distribution for Intel RSD compatible products, using OpenStack Nova and Ironic services. To implement this functionality, Intel created a Python* library called “rsd-lib,” and an OpenStack Client (OSC) plugin called “python-rsdclient” and upstreamed them to both projects https://github.com/openstack/rsd-lib and https://github.com/openstack/python-rsdclient.
These modules extend the standard OpenStack services, enabling composition calls to the Pod Manager (PODM) software component of Intel® RSD, and enable node composition in an OpenStack environment, which can then be handed off to Ironic. PODM carries out node composition requests using the Intel RSD RESTful APIs. PODM in turn communicates with hardware resources in a rack. The diagram above shows the simple OpenStack bare metal provisioning process:
- Request a new composed node.
- Enroll the new node into the Ironic database.
- Commission a workload on the new node using Nova.
Intel has published a paper describing how the integration works, how to install it, and examples of composition, enrollment and provisioning commands. It is available at https://github.com/openstack/python-rsdclient/blob/master/doc/source/user/User_Guide.pdf
Why are OpenStack users excited about the Intel Rack Scale Design architecture? Because it allows compute, storage and accelerator sharing via resource pools, which can be “composed” into any desired configuration on the fly, adding a new level of agility to cloud-oriented data centers. For example, a “logical server” (i.e., a composed node) can be scaled out in minutes to meet a new workload requirement. You could just as easily compose and provision 10 racks at a time. If you need more performance you could include fast non-volatile storage like Intel® Optane™, and (in future releases of Intel RSD) accelerators like Intel® Xeon® Phi or Intel® FPGAs. Intel RSD enables hardware and system suppliers to deliver composability in their products through RESTful APIs based on the DMTF Redfish framework. This addresses the challenge of interoperability, which is a critical requirement for enterprises and providers who need to efficiently manage their heterogeneous data centers.
Intel® RSD—Industry-Aligned Architecture for Composable Disaggregated Infrastructure
Intel and its ecosystem partners are moving quickly to deliver the future of the data center—composable, disaggregated infrastructure based on open standards. Intel RSD compatible products are available now from major vendors including Dell EMC*, Ericsson*, HPE*, Huawei*, Inspur*, QCT*, Radisys*, Supermicro*, Wiwynn* and others. Ask your equipment and software providers how to start coming up to speed on this new data center architecture, and visit https://www.intel.com/intelrsd.
*Other names and brands may be claimed as the property of others.
Figen Ülgen, Ph.D.
General Manager, Rack Scale Design Group at Intel Corporation
Figen Ülgen is General Manager of the Rack Scale Design Group within the Datacenter Group at Intel Corporation. Ülgen oversees solutions roadmap, solutions enablement as well as operations, support and ecosystem and partnership strategy for Rack Scale Design. She has held positions at Microsoft, McKinsey & Co., Motorola and at Justsystem Corporation in Japan. Ülgen earned a Bachelor’s Degree in Computer Engineering, a Master’s Degree in Expert Systems under a Fulbright Scholarship and a Ph.D. in Machine Learning under the Japanese Government Ministry of Education Scholarship. She holds two U.S. patents.