Service providers who offer bare metal as a service to their enterprise customers should consider incorporating new open standard datacenter technologies such as Intel® Rack Scale Architecture to gain better efficiencies – from resource discovery, pooling and disassembling of the cloud resources. This will help them take another step closer to realizing the benefits of software-defined infrastructure.
slow migration to public cloud
Public cloud services are gaining steam but most enterprises are still slow in migrating their workloads to the public cloud. There are many reasons. One concern is unpredictable and unreliable cloud performance when many different types of workloads are placed in a multi-tenant environment. Performance degradation, noisy neighbors, and virtual resource constraints are all culprits. Also, it is hit or miss for enterprise customers to figure out which cloud services and virtual resources they need to run their workloads in the cloud. It takes time and resources to get within a certain range.
An alternative solution for enterprises is bare-metal servers, where enterprise workloads perform much better, since there is no hypervisor overhead and the server does not have I/O limitations. It is possible to run bigger workloads that do not have to compete for virtual resources. Some service providers recognize this need and created bare-metal cloud or bare metal as a service. Customers like this solution because it reduces their operating expenses (opex) and capital expenditures (capex) and lets them realize the flexibility and scalability of cloud computing.
Challenges of bare-metal cloud
But, for service providers, there are challenges with this bare-metal cloud:
- How can we best provision servers and the service in a timely way?
- What happens to hardware and cost structure when customers complete their projects and terminate their workloads?
Of course, the hardware can be repurposed for other uses, but this takes time and resources if resources are still managed manually―especially when compute, storage (solid-state drives and memory), and networking are all still sitting on a physical server. Just the thought of conducting an inventory of the physical hardware is a headache.
Implementing SDI at the rack level
A simple way to manage this challenge is through automation or by implementing software-defined infrastructure at the rack level, and by using open APIs such as Redfish along with protocols like Intelligent Platform Management Interface (IPMI) to manage the discovery, composition, and decomposition of resources in the datacenter for bare-metal service instance provisioning.
Image courtesy of Intel
Intel developed rack-scale architecture hardware and software so service providers don’t have to scale cloud services capacity one or a few servers at a time. Intel® Rack Scale Architecture offers service providers the ability to grow their datacenter capacity incrementally at the rack level. They can inventory and pool cloud resource to provide the optimal conditions for each bare-metal instance to run performance-hungry workloads. With rack-scale architecture, bare-metal cloud services can be created and terminated with ease.