|
|
United States-English | |
|
|
 |
|
HP Capacity Advisor Version 4.1 User's Guide > Chapter 4 Planning with Capacity Advisor
Using the Smart Solver |
|
|
The Smart Solver automates the manual process of repeatedly moving workloads onto different systems to determine an optimal solution to a problem. Below, one example use of the Smart Solver is shown; then, the various types of automated scenario changes that the Smart Solver can provide are explained. For detailed steps while using Capacity Advisor, see the contextual online help, accessed by pressing the Help or ? within the Capacity Advisor screens. Suppose you have an existing data center with two new homogenous HP Proliant servers that act as VM hosts with the following VM guests:
Existing in the data center is also a heterogeneous legacy environment consisting of older standalone servers. These servers have various workloads and uses; an abbreviated table of these is shown below: Table 4-2 Legacy Standalone Servers
Suppose you want to eliminate the old servers, consolidating them to be VM guests on your existing VM hosts. Furthermore, you want to do the following:
You could manually attempt the various permutations of consolidating the legacy systems to the VM hosts while maintaining the existing rules for utilization limits using the utilization data you have gathered from the various workloads, keeping in mind that you want to create new VM hosts only when absolutely needed and the virtualization overhead, and then load balance the workloads. Or you could run the Smart Solver to do this for you. Here's how: Assumptions The process below assumes you already have the following:
The Process Below is the general process for this situation to utilize the Smart Solver. (Detailed steps and screen descriptions for the required screens below are described in the online help.)
Viewing the Results Once the Smart Solver has completed, you will be shown its results. In this case, it shows the best configuration for converting and placing the legacy systems onto VM hosts based on the above parameters. You can expand the various sections for detailed information. Saving and Reviewing the Results Additionally, you can save the results into a new scenario by selecting the Save As button. Once the scenario is saved using the Save As button, you should be taken back to the list of scenarios. From there, you can create a Scenario Comparison report, a comparative report of the original scenario with its legacy standalone servers and the new scenario with these servers now converted and running as VM guests. Depending upon your report options and the data you've collected for the systems, you can see everything from how the consolidation uses previously wasted (unused) CPU cycles to changes in power costs. The following sections describe the various Smart Solver solutions that you can choose to run on an existing scenario. The solutions show the systems converted to VM guests on VM hosts. These VM hosts are either existing VM hosts, what-if generated template VM hosts, or a combination, depending on what destinations you selected. When a combination of VM hosts are chosen, the placement of VM guests goes first to existing VM hosts, and then to the template VM hosts. Further, as part of the input parameters, you can select to load balance the resulting VM hosts. This load balancing occurs after and only amongst the target VM hosts that are required for the consolidation solution. In other words, if the consolidation solution results in any VM hosts being unused (and therefore, not required in the solution), the unused systems will not be involved in the load balancing. Only the required systems in the solution are involved in the load balancing. The resulting solution is the configuration requiring the fewest number of systems with the minimal requirement for headroom, while taking into consideration resource utilization and utilization limits. Resources The placement of VM guests takes into consideration CPU, Memory, Network I/O, and Disk I/O capacity, and utilization limits. If load balancing (balancing resource utilization across the resulting systems) was selected, the VM guests were load balanced across the systems that had 1 or more VM guests. Resource Capacity. Workloads that specify utilization limits for a metric (for example, memory or disk I/O) can only be placed on resources that define a capacity for the corresponding metric. In other words, if you specified that a workload never exceed 100% memory utilization, that workload can only be placed on a system for which total memory capacity is known. Utilization Limits. Every workload selected must have at least one utilization limit applied before using the Smart Solver. This can be any type of utilization limit, including the default global utilization limit. Headroom Ranking The headroom ranking shows the amount of available resource above the existing resource utilization that will exist for the resulting solution in the simulation. Among the solutions that require the same target systems, the solution with the tightest fit is shown. Fewer Systems Shown. The Smart Solver solution can contain fewer VM host targets than were originally selected. This occurs when the workloads fit on fewer systems than originally selected. For example, if systems A, B, and C are selected as target VM hosts, but all the workloads can fit into VM hosts A and B, then only VM hosts A and B are shown in the solution. Systems Involved in Load Balancing. When load balancing is performed, the loads are balanced only across the resulting systems in the solution. For example, if only VM hosts A and B are used (and VM host C is not), then load balancing is performed only across VM hosts A and B. VM host C is not included for the load balancing calculation. Headroom Ranking Shows Zero (0) Stars. The headroom ranking shows zero stars even though it appears that there is sufficient room on the VM host for the workloads. This happens whenever one or more of the original servers violates a utilization limit prior to the Smart Solver being run. Before running the Smart Solver, ensure that your source servers are not already violating a utilization limit (Choose What-if Actions Edit System...) The solutions show workloads that are balanced across the selected existing servers or VM hosts. Load balancing is based upon CPU, memory, network I/O, and disk I/O capacity, utilization limits, and headroom, where the goal is to distribute workloads so that each system has comparable headroom and therefore, similar headroom ranking. Resource Capacity. Workloads that specify utilization limits for a metric (for example, memory or disk I/O) can only be placed on resources that define a capacity for the corresponding metric. In other words, if you specified that a workload never exceed 100% memory utilization, that workload can only be placed on a system for which total memory capacity is known. Utilization Limits. Every workload selected must have at least one utilization limit applied before using the Smart Solver. This can be any type of utilization limit, including the default global utilization limit. Headroom Ranking The headroom ranking shows the amount of available resource above the existing resource utilization that will exist for the resulting solution in the simulation. Among the solutions that require the same target systems, the solution with the tightest fit is shown. The Load Balanced Results Appear Unbalanced. The solution may not look balanced because smaller systems generally are assigned a smaller percentage of usage than larger systems, and very small systems may end up with no workloads at all. For example, a large 16 GB system at 87% memory usage has 2 GB of headroom, and a smaller 4 GB system at 87% has only 500 MB of headroom. Aiming for 87% usage on both systems would not yield a balanced solution. Instead, a balanced solution is to fill the larger system to 87% and fill the smaller system to only 50%. With this placement, workloads placed on either system will have the same amount of headroom to grow (2 GB). No Apparent Change from Original Configuration. The solution may be the same as the original scenario, and it looks as though no computation was performed. Actually, with the current attributes and constraints, the Smart Solver could not find a better solution than the current configuration of systems. This means that the current configuration is the current best solution. A message displayed in BLUE text indicates that the results are not an error (errors are displayed in RED text). No Apparent Change on 1 or More Systems. The solution may show no change on 1 or more destination systems. Thus, it may appear that the Smart Solver did not include the server in its computations. In actuality, the Smart Solver determined that as part of the best solution, it was best to leave these target systems with their original configurations. Fewer Systems Shown. The Smart Solver solution can contain fewer VM host targets than were originally selected. This occurs when the workloads fit on fewer systems than originally selected. For example, if systems A, B, and C are selected as target VM hosts, but all the workloads can fit into VM hosts A and B, then only VM hosts A and B are shown in the solution. Smaller Systems Appear Unused. The solution may not display smaller systems, making it appear as if the smaller systems were not included in the Smart Solver computations. In actuality, when the Smart Solver attempts to place the workloads on target systems, it accounts for the robustness of those systems. If the workloads fit on larger, more robust systems and the smaller systems go unused, the solution will display only the larger, used systems. For example, if there are two large systems and two small systems, the solution may show only the two large systems, and load balancing will occur only on those two systems. The solutions show the target systems with the selected workloads placed to require as few of the target systems as possible. The target systems are either existing VM hosts, what-if generated template VM hosts, or a combination, depending on what targets you selected. When a combination of VM hosts are chosen, the placement of VM guests goes first to existing VM hosts. and then to the template-generated VM hosts. Resource Capacity. Workloads that specify utilization limits for a metric (for example, memory or disk I/O) can only be placed on resources that define a capacity for the corresponding metric. In other words, if you specified that a workload never exceed 100% memory utilization, that workload can only be placed on a system for which total memory capacity is known. Utilization Limits. Every workload selected must have at least one utilization limit applied before using the Smart Solver. This can be any type of utilization limit, including the default global utilization limit. Headroom Ranking The headroom ranking shows the amount of available resource above the existing resource utilization that will exist for the resulting solution in the simulation. Among the solutions that require the same target systems, the solution with the tightest fit is shown. Fewer Systems Shown. The Smart Solver solution can contain fewer VM host targets than were originally selected. This occurs when the workloads fit on fewer systems than originally selected. For example, if systems A, B, and C are selected as target VM hosts, but all the workloads can fit into VM hosts A and B, then only VM hosts A and B are shown in the solution. Systems Involved in Load Balancing. When load balancing is performed, the loads are balanced only across the resulting systems in the solution. For example, if only VM hosts A and B are used (and VM host C is not), then load balancing is performed only across VM hosts A and B. VM host C is not included for the load balancing calculation. In addition to the solutions shown, you can do the following with the solution:
Related Topics
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 Printable version |
|
|
|
|
|
|||||||||||||||
|
|||||||||||||||
