Description

Cluster Dynamic Resource Scheduling (DRS) is a technology for balancing and scheduling cluster resources. Traditional DRS ensures business service continuity by detecting the resource usage of hosts and migrating VMs in time to prevent resource shortage or load imbalance. However, its effectiveness is not guaranteed, such as the inability to predict the time of threshold triggering, the lag of suggestions, only based on the computing busyness of hosts but not VMs, and only based on resources but not on the performance and reliability after scheduling.

Sangfor HCI DRS 2.0 predicts the load on cluster resources for the next two hours based on the resource history of hosts and VMs over the past five days and performs intelligent scheduling based on the resource rating results to improve both the performance score and the reliability score after the migration, thus, to optimize business services.

Precautions

1. DRS 2.0 is not supported for the VMs of the following types: non-clustered VMs, NFV VMs, rapidly-recovered VMs, VMs with a specified run location, VMs with a specified fault domain, VMs with CDP enabled, VMs with CPU exclusive mode enabled, VMs that are using a GPU, and VMs with USB mapping.
2. DRS 2.0 is not supported for the nodes whose UPS-powered VM shutdown policy is triggered.
3. It is only supported to perform disk failure detection for Seagate and Western Digital SATA disks (HDDs). The deduction deviation for the detection of unsupported disks is between 5 and 10 points (Labeled as Level 3).
4. After a cluster is deployed, disk failure detection is unavailable for the first 10 days, and memory failure detection for the first 30 days because these features require historical data for training.
5. It is only supported to perform RAID card failure detection for MegaRAID and SAS3008 RAID cards.
6. It is not supported to generate all scheduling recommendations for a cluster at a time. Up to 50 recommendations can be generated in 10 minutes.

Prerequisites

1. HCI clusters consist of multiple nodes.
2. The virtual machine is stored on external storage or virtual storage.
3. The virtual machine running location is configured to be selected automatically.
4. The CDP function is not enabled on the virtual machine, the CPU exclusive function is not configured, and the USB mapping is not configured.

Steps

Go to Reliability > DRS.

40. Select an Automation Level. If Automated is configured, VMs will be automatically migrated to another node based on the resource rating results. If Manual is configured, migration recommendations will be given but need to be manually applied. If Disabled is configured, no migration recommendations will be given.
41. Select a Scheduling Mode, based on which the optimal destination nodes and recommendations are given to optimize the cluster. The Better Performance mode balances the load on each node, and the Less Cost mode saves node resources by centralizing VMs to run on one or more nodes as much as possible.

• Better Performance: Migrates VMs on nodes with high loads to nodes with low loads to improve the performance of the nodes and VMs. The reliability level of the destination nodes cannot be lower than that of the source nodes.

• Less Cost: Migrates VMs in order of node memory load from the lowest to the highest to nodes with high loads. The VM performance score decreases after the migration cannot exceed the threshold and the reliability level of the destination nodes cannot be lower than that of the source nodes.

42. Select Conservative (less frequent), Moderate (default), or Aggressive (frequent) for Sensitivity. The node load threshold depends on Sensitivity and Scheduling Mode.
43. Select Disable, Low, Medium, High, or Highest for AI Prediction Capability based on the number of VMs. If this feature is enabled, AI algorithms will be used to predict the loads of the nodes and VMs for the next two days and schedule resources based on the load trends of the past seven days.
44. Configure the Schedule to prevent DRS from being performed during off-peak hours.
45. Configure Automation Level (Automated, Manual, Disabled) for specific VMs as needed.