Page 488 - Cloud computing: From paradigm to operation
P. 488

2                                            Cloud Computing management


            6.4     Resource management for a single cloud service provider

            The overall resource management should be viewed from the point of view of the lifecycle management of
            a cloud application. The application, as it passes through its lifecycle, must be acted upon by traditional
            business  processes  associated  with  management  system  functions  such  as  administration,  provisioning,
            configuration, service assurance and charging.
            As shown in Figure 1, in the simpler case of an application that resides on a single cloud computing system,
            it becomes dependent on two distinct categories of virtualized resources. The dotted arrows depict the active
            coordinated relationship that must be maintained between resources at each level.



                                                     Application



                                                       Cloud
                                                     virtualized
                                                      resources
                                                      Network
                                                     virtualized
                                                      resources
                                                                   Y.3520(13)_F01


                              Figure 1 – Applications residing on a single cloud computing system
            NOTE  – Although Figure 1 divides virtualized resources into "cloud" and "network", cloud computing considers all
            resources at the same level [ITU-T Y.3502].
            A resource management issue requiring further work is how to use existing cloud management systems to
            maintain awareness of which logical and physical resources are actually relevant to a specific instance of a
            specific application at any given point in time.
            Due  to  the  rapid  elasticity  and  scalability  characteristic  of  cloud  computing  [ITU-T  Y.3500],  the  cloud
            computing system can configure additional resources to handle changing application demands; there are
            additional  requirements,  needing  further  analysis,  for  dynamically  reconfiguring  the  underlying  network
            configurations in response to the changing resources at various components of the cloud computing system.
            This issue arises both within the internal network fabric of large cloud computing data centres, between the
            interconnecting networks in hybrid scenarios, and across transport networks and content delivery networks.

            Another issue that arises is the division of responsibility between an internal cloud computing virtualization
            management  system  and an  external management system.  Although  the  cloud  computing virtualization
            function can typically manage its own physical and logical resource allocations for supported applications, an
            external management system may be desired to dynamically reallocate resources in a coordinated fashion
            across the three levels shown in Figure 1 or to track and have knowledge of those changing relationships.
            As shown in Figure 2, the capability of a management system to both manage resource allocations and track
            their instantaneous state could enable that management system to provide the information necessary to
            display the status of a given service and all of the underlying relevant resources, at any given point in time.

            From the point of view of the quality of service of resource management, the issue is how to ensure that the
            service assurance systems are receiving relevant telemetry from the cloud computing or network resources
            actually  involved  in  delivering  a  particular  instance  of  a  service.  The  issue  is  less  concerned  with  what
            telemetry  data  needs  to  be  managed,  as  each  dataset  is  often  unique  to  a  given  management  system
            implementation, but is more concerned with how to use the cloud computing system to do so effectively.










            480
   483   484   485   486   487   488   489   490   491   492   493