SELF-ORGANISING, SELF-MANAGING HETEROGENEOUS

CLOUD

A Brief OverviewProf J. P. Morrison

Overview Project Funding and Consortium

Specific Challenge

Typical IaaS Cloud Usage

Project Goals and Ambitions

Our Approach

The CloudLightning Architecture

Beneficiaries

Challenges

The principal goals are to ensure Europe produces world-class science, removes barriers to innovation and enabling the public and private sectors to work together in delivering innovation. 

The emphasis is on: excellent science, industrial leadership and tackling societal challenges.

The CloudLightning project was funded under Call H2020-ICT-2014-1   Advanced Cloud Infrastructures and Services

High performance heterogeneous cloud infrastructures and runs from Feb 2014 - January 2017

Horizon 2020

Project Consortium

Cloud computing is being transformed by new requirements such as • heterogeneity of resources and devices• software-defined data centres • cloud networking, security, and • the rising demands for better quality of user experience. 

Cloud computing research will be oriented towards • new computational and data management models (at both infrastructure and

services levels) that respond to the advent of faster and more efficient machines, • rising heterogeneity of access modes and devices, • demand for low energy solutions, • widespread use of big data, • federated clouds and • secure multi-actor environments including public administrations.

The aim is to develop infrastructures, methods and tools for high performance, adaptive cloud applications and services that go beyond the current capabilities

Specific Challengehttps://ec.europa.eu/research/participants/portal/desktop/en/opportunities/h2020/topics/290-ict-07-2014.html

Cloud computing is being transformed by new requirements such as • heterogeneity of resources and devices• software-defined data centres • cloud networking, security, and • the rising demands for better quality of user experience. 

Cloud computing research will be oriented towards • new computational and data management models (at both infrastructure and

services levels) that respond to the advent of faster and more efficient machines, • rising heterogeneity of access modes and devices, • demand for low energy solutions, • widespread use of big data, • federated clouds and • secure multi-actor environments including public administrations.

The aim is to develop infrastructures, methods and tools for high performance, adaptive cloud applications and services that go beyond the current capabilities

Specific Challengehttps://ec.europa.eu/research/participants/portal/desktop/en/opportunities/h2020/topics/290-ict-07-2014.html

Customer must do the hard work

Research various offerings and build/compile solutions accordingly.

Target the lowest common denominator to facilitate portability

Solution often end up either being completely generic opportunity cost

Or, they are focused on using some special features (inevitably tying them to particular providers)

portability lost

Providers support this usage pattern with over-provisioning

Typical IaaS Cloud Usage

Make Cloud Computing more accessible to the average customer.

Allow the provider to make their offering more efficient• The current model is not sustainable. The cloud is now approaching 10% of the world’s

electricity consumption.

Exploit heterogeneous hardware type

Demonstrate our approach in a very challenging application domain – HPC

Project Goals and Ambitions

We see the adoption of a Service Interface as key.

• Provides a “clean” interface between the customer and provider

• This interface should not require the customer to specify resource requirements. Rather, function requirements, workflows and SLAs

However, this implies moving the management complexity from the customer to the provider, which in turn, gives rise to a large complex system.

Project Goals and Ambitions

The BP Creator forms the work-flow and stores the Blueprint in the Blueprint Catalogue;

The Operator selects a Blueprint from the Blueprint Catalogue and optionally constrains and parameterizes it.

The Operator launches the Blueprint by:

(1) requesting an appropriate solution from the CL and (2) deploying the Blueprint on the resources returned as part of that solution.

The End User then interacts with the deployed Blueprint.

Our Approach

Managing complexity of this scale can be done using self-organisation.

• Synergetic activities of elements when no single element acts as a coordinator and the global patterns of behaviour are distributed

• Prevalent in Nature

• Already being used to develop many control systems, sensor networks, economic systems, ...

“Global order can arise from local interactions”. Alan Turing.

Conceptual Architecture

Architecture Components

Basic tenets:• component autonomy• awareness of the environment• goal-driven behaviour of individual components• self-configuration

Goals include:• minimize energy consumption• Improve service delivery

Goals are achieved by collaboration.

Self-configuration allows the system to create coalitions of resources, working in concert to respond to the needs of a specific service request, rather than offering a

menu of a limited number of resource packages.

Self-Organisation

The CL system uses a single abstract concept of resource, known as a CL-Resource.

In response to a service request, the CL system identifies a specific CL-Resourcethat will be used for the delivery of that service. The physical realization of a CL-Resource depends on what aspect of the underlying physical hardware is being exposed to the CL system.

CL-Resources can be • bare metal, • virtual machines, • containers, • networked commodity hardware (either offered as a bare metal cluster or as a cluster pre-

configured to host distributed applications), • servers with attached accelerators such as GPUs, MICs and FPGAs.

CloudLightning Resources

CL-Resources aggregated together and given a specific identity, known as a Coalition.

Coalitions formed by a vRack Manager in response to specific service requirements.

Coalitions may be persisted for improved service delivery

The constituent CL-Resources of a Coalition may span multiple servers but are restricted to a single vRack.

Resource Coalitions

vRack Manager Types and Groups

Plug and Play

Leveraging Existing OpenStack Components

BeneficiariesThe primary beneficiary is the Infrastructure-as-a-Service provider. They benefit from activating the HPC in the cloud market and a reduction in cost related to better performance per cost and performance per watt.

This increased energy efficiency can result in lower costs throughout the cloud ecosystem and can increase the accessibility and performance in a wide range of use cases including Oil and Gas discovery, Genomics and Ray Tracing (e.g. 3D Image Rendering)

• Oil and Gas

Improved physics simulations and higher resolution RTM imaging.

Energy and cost efficient scalable solution for RTM and OPM/DUNE simulations.

Reduced risk and costs of dry exploratory wells.

Genomics

Improved performance/cost and performance/Watt

Faster speed of genome sequence computation.

Reduced development times.

Increased volume and quality of related research.

Ray Tracing (3D Image Rendering)

Reduced CAPEX and IT associated costs.

Extra capacity for overflow (“surge”) workloads.

Faster workload processing to meet project timelines.

In Conclusion

The Challenges Ahead

Separate the concerns of the IaaS consumer and the CSP

Create a Service Oriented Architecture for the emerging heterogeneous cloud

Reduce energy consumption by improved IaaS management

Improve service delivery

Leverage heterogeneity to bring HPC to the cloud

Resource management in hyper-scale cloud deployments

THANK YOUJohn Morrison j.morrison@cs.ucc.ie