In this blog you will learn about Platform9 Private Cloud Director’s Dynamic Resource Rebalancing (DRR), an automated feature optimizing VM workload distribution. Discover how DRR ensures efficient resource utilization via live migration, preventing performance bottlenecks. Compare DRR to VMware DRS and understand its proactive, intelligent cluster optimization.
Introduction
In any dynamic virtualized environment, workloads fluctuate. Some virtual machines (VMs) might become resource-intensive, while others remain idle. This can lead to imbalances across the physical hypervisor hosts within a cluster – some hosts become overloaded, causing performance degradation for their VMs, while other hosts sit underutilized. Platform9 Private Cloud Director addresses this challenge with its integrated Dynamic Resource Rebalancing (DRR) feature.
What is Platform9 DRR?
Dynamic Resource Rebalancing (DRR) is Platform9’s intelligent, automated mechanism designed to optimize the distribution of VM workloads across the hypervisor hosts within a Private Cloud Director cluster. Think of it as the direct equivalent to VMware’s Distributed Resource Scheduler (DRS). DRR works continuously to ensure that resources like CPU and memory are utilized efficiently and that no single host becomes a performance bottleneck.
How DRR works: continuous optimization
DRR functions as an ongoing optimization engine for your cluster:
- Continuous Monitoring: DRR monitors key resource utilization metrics, such as CPU and memory usage, across all active hosts within the cluster. It leverages real-time and historical data to understand workload patterns.
- Imbalance Detection: DRR looks for overloaded hosts in the cluster. When DRR finds a host with CPU or memory utilization greater than 80%, it determines that host is overloaded. It then identifies underutilized hosts with sufficient spare capacity, sorting all compatible hosts by utilization and selecting the host with the lowest utilization as the migration target. DRR also takes into account per-host overcommitment or over-allocation ratios when finding a suitable candidate.
- Automated Live Migration: Once an overloaded host and a suitable target are identified, DRR selects VMs for migration based on their Migration Priority (see Configuration and Control below). VMs with High priority are selected first, followed by Normal, then Low. VMs with an Excluded priority are never migrated by DRR. DRR initiates live migrations one VM at a time, ensuring no downtime for the VM or its applications. After each migration, DRR re-evaluates whether the source host is still overloaded before continuing.
- Intelligent Placement: Beyond just rebalancing running VMs, DRR’s logic also contributes to making smarter initial placement decisions when new VMs are powered on, assigning them to the most suitable host based on current cluster load and VM requirements.
Configuration and control
DRR gives administrators several controls for tuning its behavior:
- Cluster-Level Setting: DRR is configured at the cluster level within Platform9 Private Cloud Director. When you create a new virtualized cluster, you have the option to enable DRR. You can also enable or disable DRR on an existing cluster at any time.
- Configurable Frequency: When you enable DRR, you choose how often it evaluates the cluster for rebalancing. The default is 20 minutes. You can also set it to 10 minutes or 30 minutes.
- Migration Priority (per-VM): Administrators can assign a migration priority to individual VMs, controlling whether and in what order DRR selects them for migration. Note that migration priority only applies to DRR. It does not affect VM HA evacuations or manual migrations initiated by an administrator. The supported values are:
- Unset – The default state. VMs without an explicit priority and without soft affinity rules are treated as Normal.
- High – Selected first for migration. Good candidates are dev/test VMs that are smaller or less sensitive to migration overhead.
- Normal – Selected after High-priority VMs, before Low-priority VMs.
- Low – Selected last. Assign this to VMs that incur a heavier cost from migration and should only be moved when other options are exhausted. VMs with soft affinity rules that don’t have an explicit priority are also treated as Low.
- Excluded – Never migrated by DRR. Use this for VMs you don’t want DRR to move under any circumstances.
Benefits of using Platform9 DRR
Enabling DRR in your Platform9 Private Cloud Director environment delivers tangible benefits:
- Consistent VM Performance: Helps prevent performance issues caused by resource contention on overloaded hosts.
- Optimized Resource Utilization: Ensures that your hardware investment is used more efficiently across the cluster.
- Proactive Problem Avoidance: Identifies and resolves potential resource bottlenecks before they negatively impact applications.
- Reduced Operational Overhead: Automates the complex task of monitoring and balancing VM workloads, freeing up administrator time.
DRR interoperation with other services
Host Aggregates
DRR respects host aggregate assignments. It will only find candidate target hosts that satisfy a VM’s host aggregate requirement. If no suitable host exists within the aggregate, DRR will not migrate the VM.
VM HA
DRR and VM HA are designed to work well together. If a host failure occurs while DRR is actively rebalancing VMs, VM HA will detect the failure and initiate evacuations. The evacuations may create a cluster imbalance, and DRR will detect and address that imbalance during the current or next run.
Affinity and Anti-Affinity Rules
- Hard affinity: DRR will skip VMs with a hard affinity rule. This may change in the future once DRR supports bulk VM migrations.
- Hard anti-affinity: DRR will find a candidate host that satisfies the anti-affinity rule. If no suitable host is available, DRR will not migrate the VM.
- Soft affinity: DRR will try to find a target host that satisfies the affinity policy. If it can’t, it will select another target host. VMs with soft affinity rules that don’t have an explicit migration priority are treated as Low priority.
- Soft anti-affinity: DRR will try to find a target host that satisfies the anti-affinity policy. If it can’t, it will select another target host. VMs with soft anti-affinity rules that don’t have an explicit migration priority are treated as Low priority.
VM and Host States
DRR only live migrates VMs in an Active (powered on) state. VMs in all other states are skipped. DRR only operates on hosts in an online state with the hypervisor role assigned. Hosts that are offline, in error state, or without the hypervisor role are ignored.
VMs with Special Properties
- DRR will live migrate VMs that have hot-added CPU or memory resources.
- DRR does not currently support live migration of vTPM-enabled VMs. This is planned for a future release.
VMware DRS vs. Platform9 Private Cloud Director DRR comparison
This table compares VMware Distributed Resource Scheduler (DRS) with Platform9 Private Cloud Director’s Dynamic Resource Rebalancing (DRR).
| Feature | Platform9 Private Cloud Director DRR | VMware DRS |
| Primary Goal | Optimize VM distribution & workload balance (CPU/Memory) across hosts in a cluster. | Balance VM workloads (Primarily CPU/Memory) across hosts in a cluster to ensure performance and prevent overload. |
| Mechanism | Continuous monitoring of CPU/Memory utilization with configurable evaluation frequency (10, 20, or 30 minutes). | Periodic calculation of cluster balance & recommendations/actions for VM migration. |
| Trigger for Action | Host CPU or memory utilization exceeds 80%. DRR selects the least-utilized compatible host as the migration target. | Cluster imbalance calculation exceeding a defined threshold. |
| Automation Control | Fully automated. Administrators control behavior through configurable frequency and per-VM migration priority. | Configurable Automation Levels (Manual, Partially Automated, Fully Automated). |
| Migration Control | Per-VM migration priority (High, Normal, Low, Excluded) determines selection order. VMs migrated one at a time. DRR re-evaluates after each migration. | Configurable Migration Threshold (Conservative to Aggressive) controls sensitivity to imbalance. |
| Configuration Scope | Cluster-level setting with per-VM migration priority overrides. | Cluster-level setting, with options for per-VM overrides. |
| Migration Technology | Uses Platform9 Live Migration. | Uses VMware vMotion. |
| Placement Logic | Influences initial VM placement using enhanced logic/metadata filters. Respects host aggregates and affinity/anti-affinity rules. | Performs initial VM placement based on cluster load (in automated modes). |
| Advanced Features | Per-VM migration priority, configurable frequency, host aggregate awareness, affinity/anti-affinity rule awareness. | Offers Predictive DRS (using vROps data), VM Distribution rules (spread VMs evenly). |
| Default State | Enabled by default when a cluster is created. Can be disabled per cluster. | Requires explicit configuration and enabling. |
Conclusion
Platform9’s Dynamic Resource Rebalancing (DRR) is a vital feature for maintaining a healthy, efficient, and high-performing private cloud. By automatically monitoring resource usage and intelligently migrating VMs using live migration, DRR ensures that workloads are optimally distributed across your cluster hosts. With configurable evaluation frequency, per-VM migration priority, and awareness of host aggregates and affinity rules, DRR provides both automated intelligence and operator control for handling dynamic workload demands.
Continue learning
Explore our eight learning modules and become a Private Cloud Director expert.
