This article can also be found in the Premium Editorial Download "Virtual Data Center: What is cloud computing, and what can it really do?."
Download it now to read this article plus other related content.
Capacity planning tools carry out a number of functions that help data centers operate at peak efficiency. These tools allow data center managers to monitor their current physical servers and virtual machines (VMs) as well as the virtual machine hosts on which VMs run.
With the proper capacity planning tool, data center managers can size their virtual environments to allow for better server utilization and plan for growth. Picking the best capacity planning tool, however, means you need to do your homework. You must define your requirements carefully and pick the tool that best matches your needs.
Capacity planning tools: Getting started
Why is capacity planning so important? For example, a successful migration requires capacity planning. Enterprises considering virtualization must determine server utilization levels before migrating VMs.
Before a physical-to-virtual (P2V) migration, server-level utilization must be determined so as not to underestimate the physical targets. It is also essential to determine servers' memory needs before migrating VMs to help prevent costly memory add-ons to improve performance levels after the fact.
The ideal capacity planning tool should include built-in modeling to enable model creations and allow for memory, CPU, disk, and network change values. This will also allow visibility for the number of VMs on the virtual platform.
A capacity planning tool will monitor physical servers and identify potential candidates to be consolidated into a virtual environment. For the best results, the tool should run for at least 30 days to capture accurate information about peaks and spikes in the CPU memory network and disk utilization.
A spike can be related to something happening during a certain time, such as when a backup job is running. A peak can be caused by applications or when a server runs out of resources. A peak typically has a longer duration than a spike.
During a 30-day period, servers should be monitored regularly for highly utilized values showing up in the chosen capacity planning tool. If the server's CPU utilization run rate is in the neighborhood of 80% to 90%, it may be due to older servers running Pentium 3 600 Mhz. This is not an issue if the virtualization target platforms that have been migrated are powerful servers.
For example, moving a 500 Mhz physical machine onto a Hewlett-Packard DL585 that is running four-way quad-core processors, 3 Ghz, will not be an issue because the newer hardware platform delivers more processing power. Also, once the physical machine has been migrated over, the CPU utilization drops because it's running on the more powerful platform.
The other values to be monitored are memory paging. Many physical servers running at 256 MB of memory are paging to disk. Once these servers have been migrated over, the memory can be increased to reduce the paging to disk by each VM. The value of the physical network adaptor in the server should not be a concern because most of the time the network adaptor can be an older 100 Mbps or a 1 Gbps, but the bandwidth may not be used to its full potential. The highly utilized values are always memory and CPU. This exercise easily identifies servers best suited for migration.
<<b>Using capacity planning to monitor VMs
Organizations with existing virtualized servers can benefit from capacity planning to help manage virtual machines, prevent bottlenecks and maximize utilization on virtual machine hosts. Many capacity planning tools allow the import of existing virtual machine hosts. This will help determine the efficiency of the current servers, be it VMware Infrastructure 3 (VI3), Citrix Xen Server or Microsoft Hyper-V.
Running the capacity planning tool will help determine the number of VMs that can run on the virtual hosts. Depending on the existing utilization levels of current virtual machine hosts in the data center, VMs can be added to increase utilization on those hosts. Some who don't run a capacity planning tool will find out later, if they do run one, that they can add many more VMs and still have room for growth.
The standard percentage of desired virtualization across virtual machine hosts can be:
- 80% memory utilization
- 70% CPU utilization
- 60% disk utilization.
These values can be adjusted according to each organization's needs. Once these values are reached after running the capacity planning tool, another virtual machine host can be added to the environment. This allows for adding new virtual machine hosts as needed rather than adding new servers without backing up data. It is also possible to run a capacity planning tool after migrating physical servers to VMs to verify that options used during the migrations were properly set.
Capacity planning to forecast growth needs
Many capacity planning tools offer an option to monitor virtual machine hosts and forecast growth needs.
Akorri Balance Point is a performance-monitoring tool that also allows for capacity planning.
With the correct capacity planning tool, data center managers can determine the level of storage area network (SAN) disk performance when adding more VMs onto virtual machine hosts. It can also help troubleshoot VMs running on virtual machine hosts.
For instance, there may be too many applications running on one virtual machine host, which can be very memory-intensive. The capacity planning tool can help spread these applications across other virtual machine hosts, along with other applications that are CPU-, disk- and network-intensive, to obtain a good balance of performance across virtual machine hosts.
Many enterprises run multiple SANs. With the appropriate capacity planning tool, each SAN can be examined to determine if increased performance is possible by moving VMs onto another SAN platform. This is a nice option because users may have thought about adding more disks to their SAN to increase performance -- with capacity planning, they have other options besides adding more disks, which increases costs. If users find they can move onto another SAN and get better performance without adding more costs, they will go this down route without any issues.
Best practices for selecting capacity planning tools
If your organization is considering capacity planning tools, you need to determine the following:
- If you can set up and use the tool in-house or if vendor services are required to install and run it. VMware's Capacity Planner tool allows consultants to run the tool on-site and create a report for review. This option also provides the ability to run a consolidation estimate assessment or consolidation assessment. A consolidation estimate assessment provides basic information about servers and a simple report. A consolidation assessment provides a more detailed examination of servers and more granular reporting.
- VMware's virtual platform migration tools allow the use of vCenter, which offers a Guided Consolidation feature (recommended for smaller environments). VCenter allows physical server monitoring, which, in turn, allows the migration onto virtual machine hosts. This by no means provides advanced features compared with running a full capacity planning tool.
- For long-term capacity planning, Platespin Recon and Akorri Balance Point are the recommended tools. There are many features that organizations must take into account, including price. Costs involved with a capacity tool include how the product is licensed. Some are licensed by number of servers and the duration of monitoring. Other tools are licensed by how much disk space will be monitored. Even if running an Intel x86-based virtualization platform, many capacity planning tools will allow the monitoring of other types of servers. This can include Unix-based servers, which can be virtualized at a later date.
Some capacity planning tools allow integration with current monitoring tools. Platespin Recon, for instance, will allow the import of data from Microsoft Operations Manager 2005, Microsoft Windows Performance Monitor and HP Reporter.
Once the tool has been set up to import the data, it can be used for the existing data. It can also be used to determine the best performance options for setting up virtual and physical machines.
Heads up on security issues
Security issues can crop up when running a capacity planning tool. Many capacity planning tools require opening up ports on the firewall to allow communications between the capacity planning tool and the servers it is monitoring.
Monitoring Windows-based servers will require access to Windows Management Instrumentation, Remote Registry, and File and Print Sharing Services. You will also need to have a user account with proper access to each server you want to monitor. Any administrator account that has local administrator rights will work. Most users will make use of domain admin accounts, which simplify the selection of which account to use for the tool.
Monitoring Unix-based servers requires an Secure Shell, or SSH, to be installed on each Unix host. You also need to have user accounts and passwords to access these servers. These user accounts can be the root account.
Depending on the tool, data center managers may have to install some software on the server being monitored. Platespin Recon requires GSX COM API installed on the GSX server if you plan to monitor GSX servers.
ABOUT THE AUTHOR: Raymond McDonald is principal systems engineer at Expert Server Group, a virtualization
consulting group in Bedford, N.H., which specializes in analyzing complex functions, procedures and
problems to find creative, logical and effective solutions.
This was first published in February 2010