Skip to main content

Cloud Technologies and Services

The Cloud Technologies and Services group focuses on advanced technologies and services towards smarter and optimized cloud infrastructures. Our major focus areas are software defined networking (SDN), cloud systems analytics and optimization, and the Power platform differentiation as a base for cloud infrastructure.

The group's activities in these areas influence a wide range of IBM system products and services, including IBM Cloud, IBM SDN, IBM Power Systems, and more. We collaborate with several teams from IBM Research and contribute to productization work of IBM development peers. We are also collaborating on the EU FP7 and the Horizon 2020 programs, where the currently active projects are COSIGN and BEACON.

Software Define Network (SDN)

SDN is one of our major focus areas, where the team has a long history and experience in all aspects of network virtualization, performance, and federation.

A notable achievement in the SDN area is the Distributed Overlay Virtual nEtwork (DOVE) technology, which combines overlay networking with centralized control. The team innovated the base technology and took it from a research prototype into a product contribution (SDN-VE). DOVE presents an abstraction of 'Network-as-a-Service' where the network connectivity is determined and enforced according to different types of policy (QoS, security, etc.) that are defined at the application level (without the need to specify low level notions like ports, VLANs, etc.). Parts of the technology were contributed to open source (OpenDaylight), and as part of our contribution to the EU project FIWARE the team integrated the technology into a cloud management stack (OpenStack).

Two domains where we exploit our SDN expertise to advance the state of the art research are optical data center networking (DCN), and cloud federation. In optical DCN, we participate in the EU COSIGN project, where we research and develop SDN orchestration architecture that will optimally exploit the benefits of optical links and devices for the next generation DCN. On SDN federation we participate in the EU BEACON project, where we research and develop technology that enables efficient and seamless Interconnection of multiple clouds (on/off premises) with different intra-cloud networking approaches.

Our team also contributed to SDN data plane performance, where we innovated a technique to overcome performance degradation resulting from absence of hardware off-load due to the VXLAN encapsulation.

Our team is now participating in the IBM Cloud Innovation Lab (CIL), contributing to the effort of Next-Gen cloud data center networking, towards empowerment of IBM's hybrid and public cloud offerings (i.e., SoftLayer).

Cloud Systems Analytics and Optimization

In this research area we exploit multiple sources of cloud system data – workload performance metrics, system event messages, logs, etc., to gain insights into the system behavior and exploit it for workload and resource optimization, detection of failures, and anomalies that can indicate security issues.

In the Federated Elasticity project we provide optimized elasticity for cloud workloads in various scenarios covering local, hybrid, and federated cloud infrastructures. Based on analysis of workload metrics we generate a resource demand prediction, which is used to optimize the resource allocation in various aspects: auto-scaling parameters, resource types (e.g., virtual machine multi-dimension sizes), and location. We integrated the prediction-based auto-scaling optimization into the Cloud Services Broker – a new service being developed by IBM GTS, providing optimized selection and setup of auto-scaled resources across multiple cloud providers. This was demonstrated in the recent IBM InterConnect 2015 conference. We are working with the IBM Cloud unit to integrate a smarter, prediction-based elasticity, employing a proactive auto-scaling, into the IBM OpenStack-based cloud offerings.

Another project in this domain focuses on anomaly detection in event data, provided as multiple time series signals, primarily for the purpose of identifying potential security attacks. We developed an anomaly detection technology based on control theory, which was integrated as a pilot in an IBM service that protects bank customers, and provides security alert to the service operators.

Power Systems Differentiation

We are working to optimize the Power platform for cloud and legacy workloads by exploiting its unique features.

We are working with IBM Systems to contribute advanced policies integrated into the PowerVC product that exploit the Power specific features to provide new dimensions of pool-wise optimization (as global Capacity on Demand, Shared Processor Pools for enforcing group performance entitlement and for licensing capping optimization).

In another project that focuses on the Power/KVM platform, we optimize the performance of the Docker environment on Power for critical workloads as SPARK.