There are many features that help with load balancing in virtual environments, but they differ from platform to...
platform and it can be challenging to learn how to use them to meet your needs. When these features aren't used correctly, it can result in severe performance problems stemming from any number of potential causes. In vSphere, for example, changing the configuration of one feature can have a negative impact on another, as well as the VMs it enables. By understanding how each feature works both individually and together, you can better manage your environment. Read these five quick tips to learn more about VMware load balancing.
VMware load balancing features
VMware load balancing is mainly done through Distributed Resource Scheduler (DRS), but there are other features that come into play. The first step to mastering load balancing is to know which feature does what. DRS uses vMotion to automatically balance workloads across multiple hosts based on available resources in a virtual environment. When resource contention occurs, DRS migrates VMs from one host to another in the cluster. High Availability (HA) is a feature that pools VMs and hosts into the cluster and moves VMs in the event of a host failure. VMware load balancing through DRS and automatic failover through HA work together to ultimately achieve faster workload rebalancing. In some cases, however, it makes more sense to use Fault Tolerance (FT) instead of HA. Because FT maintains a copy of the VM ready to swap out when a failure occurs, it allows for consistent uptime.
DRS and HA configurations
If you don't configure DRS and HA properly, VMware load balancing won't be as effective as it could be. The rules for these features overlap, so, without proper planning, settings for one could negatively affect performance or uptime. The first step to remedying this is to figure out which VMs should share a physical server. VMs that need to communicate with one another should, ideally, reside on the same server, but also consider what will happen if HA is triggered. If startup order is important, place VMs in a VMware vApp so you have control and can add a delay, if needed. Consider delaying startup of test and development VMs, in favor of reserving more resources for production VMs. When separating VMs, plan for different types of issues -- host failures, rack outages and so on -- by setting preferred hardware in addition to DRS and HA rules. Finally, document everything.
Troubleshooting DRS issues
In addition to moving workloads around, VMware load balancing with DRS can also power off lightly used servers to conserve energy and power them back on once demand increases. A common issue administrators encounter are servers that won't power down; DRS rules can be a contributing factor to this problem. Check your documentation to confirm DRS rules don't conflict and to make sure the feature isn't configured with a high-migration threshold setting. If VMs can't be migrated off the server, it might be because the destination server doesn't have the available resources to take on the VMs. Hypervisor incompatibility could be another contributing factor If you still haven't solved the problem, check other software configurations, such as vMotion and Distributed Power Management, and hardware configurations, such as Wake-on-LAN and Intelligent Platform Management Interface.
Introducing Storage DRS
Starting with vSphere 5.0, the vSphere Enterprise Plus license includes a new VMware load balancing feature: Storage DRS. Storage DRS, which is based on the same principles as DRS, balances VM storage consumption with data store clusters. Storage DRS checks the data stores' capacity and constraints every eight hours and moves VMs off any data stores that are at an 80% capacity or higher. As with DRS, there are different rules you can enable to keep VM storage on specific clusters or to balance storage and I/O across data stores. Be sure to fine-tune your setup once you've installed Storage DRS to ensure maximum performance, but remember that it's not a good idea to mix data store types. Before changing advanced settings, thoroughly evaluate what the potential impact will be. In larger environments with high storage requirements and variable workloads, Storage DRS can be a valuable tool for VMware load balancing, but only if it's configured properly.
DRS and HA improvements
Upgrade to vSphere 6.5 and you'll also see improvements to both DRS and HA to help with VMware load balancing. DRS will now take network use into account when using vMotion to move VMs from one host to another. In addition, there are new options for rebalancing clusters in DRS, including VM Distribution and CPU over-commitment. Proactive HA -- which is actually a DRS feature as opposed to an HA feature -- puts degraded hosts into Quarantine Mode, causing DRS to start moving VMs off those hosts and to keep other VMs from being moved to them. HA has also been enhanced and renamed vSphere Availability in version 6.5. This feature works by setting failover capacity with host failures to tolerate in Admission Control, which now defaults to percentage rather than slots.
Understand vSphere licensing changes