ACTiCLOUD proposes a novel cloud computing architecture for drastically improved management of cloud resources, targeting 1.5x increase in resource efficiency and more than 10x in scalability. Utilizing hardware intelligence that enables true resource disaggregation between multiple servers, ACTiCLOUD promotes holistic resource management both at the rack scale and across distributed cloud sites. Furthermore, the ACTiCLOUD novel architecture will enable the cloudification of in-memory databases, a core component for extremely demanding and highly critical classes of applications.
Cloud Service Providers, Cloud software vendors, Big data analysts, academics and researchers
Recent studies have shown that in most cloud data centers resource utilization ranges from 10% to 50%. ACTiCLOUD targets the two critical barriers that currently hinder true fluidity of cloud resources: the server barrier and the data center barrier.
Today, flexibility and elasticity in resource provisioning are true only for storage resources that can be shared between multiple servers within a single data center, and horizontal (“scale out”) scalability where additional VMs are allocated to cover increased demands. The critical resources of CPUs and main memory remain constrained to what one physical server can provide, limiting the ability of the infrastructure to host resource-demanding applications (especially large-scale in-memory processing) and increase resource utilization. ACTiCLOUD aims to extend resource disaggregation beyond storage, to processing cores and main memory, as well as provide a set of mechanisms and policies that will support application migrations between cloud sites that are geographically distributed.
ACTiCLOUD focuses on the critical family of applications that utilize resource-demanding databases of various types with a goal to provide an environment capable of serving their high resource requests. The extended scalability and fluidity of resources provided by ACTiCLOUD, will allow the support for resource-demanding, in-memory databases in ACTiCLOUD-enabled cloud systems.
By utilizing the ACTiCLOUD system software, system libraries and database systems, Cloud Service Providers will drastically increase the resource efficiency of their infrastructure and make available large pools of resources to the executing applications. Thus they will be able to support applications that today cannot be hosted on their infrastructure due to the requests being in excess of the server capacity; alternatively, they will be able to unify their infrastructures, as opposed to current practices where typical cloud workloads are serviced by typical low-end servers and resource hungry workloads by specialized machines typically outside the cloud.
Cloud software vendors will improve their resource efficiency capabilities at the server, rack-scale, local cloud site and distributed cloud levels.
ACTiCLOUD will enable a rack-scale in-memory database (MonetDB) and a rack-scale graph database (Neo4j), which will offer to Big Data analysts efficient (in terms of both performance and resource consumption) in-memory processing for an order of magnitude higher data volumes than provided today in typical cloud offerings.
Finally, ACTiCLOUD will develop a complete set of algorithms and methodologies for optimized resource allocation at the rack-scale and distributed clouds that can be utilized by researchers to further advance findings on the field of resource-efficient computing.