Abstract: Methods and systems may provide for reducing workloads of neighboring virtual machine tenants in a cloud environment with shared storage using pro-active log offloading. Additionally, logging activity may be redirected to reduce input/output resource consumption. Trends in future input/output activities may be determined and preemptive action may be implemented to reduce performance impact of the neighboring tenants.

Abstract: A system and method for iteratively deploying a workload pattern are provided. The system and method determines a current set of requirements for at least one piece of the workload pattern that is initiated in a designer and generates a stability metric for at least one of the current set of requirements. The system and method further compares the stability metric to an acceptance threshold and calculates an estimated time to deploy the at least one piece of the workload pattern based on the comparing of the stability metric to the acceptance threshold.

Abstract: A system and method for iteratively deploying a workload pattern are provided. The system and method determines a current set of requirements for at least one piece of the workload pattern that is initiated in a designer and generates a stability metric for at least one of the current set of requirements. The system and method further compares the stability metric to an acceptance threshold and calculates an estimated time to deploy the at least one piece of the workload pattern based on the comparing of the stability metric to the acceptance threshold.

Abstract: Methods and systems may provide for reducing workloads of neighboring virtual machine tenants in a cloud environment with shared storage using pro-active log offloading. Additionally, logging activity may be redirected to reduce input/output resource consumption. Trends in future input/output activities may be determined and preemptive action may be implemented to reduce performance impact of the neighboring tenants.

Abstract: A computer implemented method and system for managing resources available to an application in a shared pool of configurable network computing resources is provided. The method includes receiving, by a metric collector, one or more service level metrics correlated with one or more time intervals. The service level metrics are converted to hypervisor weights and a hypervisor is configured to allocate computing resources to a virtual machine according to the hypervisor weights. In response to translating and implementing the hypervisor weights, a dynamic cost model is created. The cost model is correlated with chosen service level metrics as well as with the duration of time a chosen service level metric is implemented.

Abstract: A technique for caching evidence for answering questions in a cache memory of a data processing system (that is configured to answer questions) includes receiving a first question. The first question is analyzed to identify a first set of characteristics of the first question. A first set of evidence for answering the first question is loaded into the cache memory. A second question is received. The second question is analyzed to identify a second set of characteristics of the second question. A portion of the first set of evidence, whose expected usage in answering the second question is below a determined threshold, is unloaded from the cache memory.

Abstract: A method, system and computer program product for selecting virtual machines to be migrated. The page consumption is monitored and recorded for each virtual machine running on a physical machine in a cloud computing node. Those virtual machines with a positive gradient of page consumption that is less than a threshold are placed in a list of virtual machines to be ranked. The virtual machines in the list are ranked based on an increasing order of the relative page consumption value at each sample interval within an observation window of time. A final rank for each virtual machine in the list is computed based on averaging its ranking across each sample interval within the observation window of time. One or more virtual machines are then selected to be migrated to another physical machine that have the lowest final ranking (i.e., with the lowest relative page consumption values).

Abstract: A method, system and computer program product for selecting virtual machines to be migrated to a public cloud. The current resource usage for virtual machine instances running in the private cloud is determined. Furthermore, any scaling policies attached to the virtual machine instances running in the private cloud are obtained. Additional resource usages for any of the virtual machine instances with a scaling policy are computed for when these virtual machine instances are scaled out. A cost of running a virtual machine instance in the public cloud is then determined using its current resource usage as well as any additional resource usage if a scaling policy is attached to the virtual machine instance based on the cost for running virtual machine instances in a public cloud. If the cost is less than a threshold cost, then the virtual machine instance is selected to be migrated to the public cloud.

Abstract: A method, system and computer program product for selecting virtual machines to be migrated to a public cloud. The current resource usage for virtual machine instances running in the private cloud is determined. Furthermore, any scaling policies attached to the virtual machine instances running in the private cloud are obtained. Additional resource usages for any of the virtual machine instances with a scaling policy are computed for when these virtual machine instances are scaled out. A cost of running a virtual machine instance in the public cloud is then determined using its current resource usage as well as any additional resource usage if a scaling policy is attached to the virtual machine instance based on the cost for running virtual machine instances in a public cloud. If the cost is less than a threshold cost, then the virtual machine instance is selected to be migrated to the public cloud.

Abstract: Mechanisms are provided for scheduling execution of pre-execution operations of an annotator of a question and answer (QA) system pipeline. A model is used to represent a system of annotators of the QA system pipeline, where the model represents each annotator as a node having one or more performance parameters indicating a performance of an execution of an annotator corresponding to the node. For each annotator in a set of annotators of the system of annotators, an effective response time for the annotator is calculated based on the performance parameters. A pre-execution start interval for a first annotator based on an effective response time of a second annotator is calculated where execution of the first annotator is sequentially after execution of the second annotator. Execution of pre-execution operations associated with the first annotator is scheduled based on the calculated pre-execution start interval for the first annotator.

Abstract: A method, system and computer program product for selecting hypervisor platforms that are best suited to process application workloads. Attribute requirements for an application workload, such as high CPU capacity, high power and low cost, are received. A ranking algorithm is then applied to a list of pools of compute nodes to identify an ordered list of pools of compute nodes that are best suited for satisfying the attribute requirements of the application workload by comparing hypervisor characteristics of the pools of compute nodes with the attribute requirements of the application workload. Each pool of compute nodes runs on a particular hypervisor platform which has a unique combination of characteristics that correspond to a combination of a set of attribute requirements (e.g., medium CPU/memory/disk capacity; high CPU and memory performance). In this manner, the hypervisor platforms that are best suited for satisfying the application workload requirements are identified.

Abstract: A method, system and computer program product for selecting hypervisor platforms that are best suited to process application workloads. Attribute requirements for an application workload, such as high CPU capacity, high power and low cost, are received. A ranking algorithm is then applied to a list of pools of compute nodes to identify an ordered list of pools of compute nodes that are best suited for satisfying the attribute requirements of the application workload by comparing hypervisor characteristics of the pools of compute nodes with the attribute requirements of the application workload. Each pool of compute nodes runs on a particular hypervisor platform which has a unique combination of characteristics that correspond to a combination of a set of attribute requirements (e.g., medium CPU/memory/disk capacity; high CPU and memory performance). In this manner, the hypervisor platforms that are best suited for satisfying the application workload requirements are identified.

Abstract: A method, system and computer program product for optimizing runtime performance of an application workload. Network input/output (I/O) operations between virtual machines of a pattern of virtual machines servicing the application workload in a private cloud are measured over a period of time and depicted in a histogram. A score is generated for each virtual machine or group of virtual machines in the pattern of virtual machines based on which range in the ranges of I/O operations per seconds (IOPS) depicted in the histogram has the largest sample size and the number of virtual machines in the same pattern that are allowed to be in the public cloud. In this manner, the runtime performance of the application workload is improved by minimizing the network input/output communications between the two cloud environments by migrating those virtual machine(s) or group(s) of virtual machines with a score that exceeds a threshold value.

Abstract: A method, system and computer program product for optimizing runtime performance of an application workload. Network input/output (I/O) operations between virtual machines of a pattern of virtual machines servicing the application workload in a private cloud are measured over a period of time and depicted in a histogram. A score is generated for each virtual machine or group of virtual machines in the pattern of virtual machines based on which range in the ranges of I/O operations per seconds (IOPS) depicted in the histogram has the largest sample size and the number of virtual machines in the same pattern that are allowed to be in the public cloud. In this manner, the runtime performance of the application workload is improved by minimizing the network input/output communications between the two cloud environments by migrating those virtual machine(s) or group(s) of virtual machines with a score that exceeds a threshold value.

Abstract: A technique for cloud infrastructure backup in a virtualized environment utilizing shared storage includes obtaining a workload input/output (I/O) profile to the shared storage over a time period. An attempt to locate one or more time windows in the workload I/O profile for which a cloud infrastructure backup can be staged is initiated. In response to determining the cloud infrastructure backup can be staged during at least one of the time windows, staging of the cloud infrastructure backup is scheduled during a selected one of the time windows. In response to determining the cloud infrastructure backup cannot be staged during at least one of the time windows, an interference tolerance approach is employed for accessing the shared storage for active workloads and the cloud infrastructure backup during the staging of the cloud infrastructure backup.

Abstract: A technique for cloud infrastructure backup in a virtualized environment utilizing shared storage includes obtaining a workload input/output (I/O) profile to the shared storage over a time period. An attempt to locate one or more time windows in the workload I/O profile for which a cloud infrastructure backup can be staged is initiated. In response to determining the cloud infrastructure backup can be staged during at least one of the time windows, staging of the cloud infrastructure backup is scheduled during a selected one of the time windows. In response to determining the cloud infrastructure backup cannot be staged during at least one of the time windows, an interference tolerance approach is employed for accessing the shared storage for active workloads and the cloud infrastructure backup during the staging of the cloud infrastructure backup.

Abstract: A technique of backing up a workload in a virtual environment includes identifying one or more files that are associated with the workload. One or more source volumes that include the one or more files are identified. A respective target volume is provisioned for each of the one or more source volumes identified. Only dirty blocks are copied (in a snapshot mode that prevents an update to the one or more source volumes during the snapshot mode) from each of the one or more source volumes to its respective target volume. The one or more dirty blocks are then copied from each target volume to a backup medium.

Abstract: A technique of backing up a workload in a virtual environment includes identifying one or more files that are associated with the workload. One or more source volumes that include the one or more files are identified. A respective target volume is provisioned for each of the one or more source volumes identified. Only dirty blocks are copied (in a snapshot mode that prevents an update to the one or more source volumes during the snapshot mode) from each of the one or more source volumes to its respective target volume. The one or more dirty blocks are then copied from each target volume to a backup medium.

Abstract: Mechanisms are provided for scheduling execution of pre-execution operations of an annotator of a question and answer (QA) system pipeline. A model is used to represent a system of annotators of the QA system pipeline, where the model represents each annotator as a node having one or more performance parameters indicating a performance of an execution of an annotator corresponding to the node. For each annotator in a set of annotators of the system of annotators, an effective response time for the annotator is calculated based on the performance parameters. A pre-execution start interval for a first annotator based on an effective response time of a second annotator is calculated where execution of the first annotator is sequentially after execution of the second annotator. Execution of pre-execution operations associated with the first annotator is scheduled based on the calculated pre-execution start interval for the first annotator.

Abstract: An approach is provided for an information handling system to select a cloud environment for a workload. In the approach a number of provisioning delays are identified with each of the provisioning delays corresponding to the workload being provisioned at a different cloud-based environment. One of the cloud-based environments is selected, with the selected cloud-based environment being based on the provisioning delay that was identified to provision the workload at the selected cloud-based environment. The workload is then deployed at the selected cloud-based environment.