Creating load-based scaling

Applies to: Kyvos Enterprise Kyvos Cloud (SaaS on AWS) Kyvos AWS Marketplace

Kyvos Azure Marketplace Kyvos GCP Marketplace Kyvos Single Node Installation (Kyvos SNI)

When the cluster usage duration is not fixed, you must use load-based scaling. This type of scaling allows you to set rules to automatically scale up or scale down the query engines based on usage patterns. By managing cluster scaling, you can enhance the utilization of your cloud cluster and save on compute costs. Additionally, you can set rules to automatically shut down or start the BI Servers and Query Engines to optimize resource utilization further.

To implement a scale-down rule, you can configure a setting to automatically reduce the capacity of your Query Engines when the cluster is under usage. Additionally, you can set a rule to automatically shut down the Query Engines and BI Servers if the cluster is not in use for a specified period. Similarly, for a scale-up rule, you can set a rule to automatically start the BI Servers and Query Engines with the required capacity when the cluster utilization exceeds over a defined period.

The Kyvos 2024.2 release comes with enhanced Load-based scaling.
Scaling rules are implemented based on the CPU load and memory usage of the Query Engine instances. System resources will be monitored for all BI Servers and Query Engines every 30 seconds, and Query Engines’ scaling will be performed based on this data.

To set a load-based scheduling, from the Toolbox, click Setup > Cluster Scaling. The Cluster Scaling page is displayed. The Load option is selected by default.

Kyvos will provide the following scaling modes for Load-based scaling:

Managed: In Managed scaling, Kyvos intelligently manages the cluster capacity to scale up and scale down.
Custom: In Custom scaling, you can set up rules based on your environment's usage patterns. Currently, the custom scaling mode supports CPU and Memory utilization-based scaling. When the CPU or memory load condition meets the configured parameters, the Query Engine cluster can be scaled up or down.

On this page, you can specify rules to scale down or scale up the query engines, as well as set rules to shut down or start the BI Servers and query engines.

Note

The capacity of the BI Servers cannot be changed.
All BI Servers can be shut down except the Coordination Master. It is required to provide a response to a query. If there is only one BI Server, this BI Server is treated as the Coordination Master.
The Settings option and Add Schedule option are disabled on the Load screen. You can set the default settings for the query engine cluster or create a new schedule using the Schedule option displayed on the Cluster Scheduling page. See the Creating schedule-based scaling section for more details.
You can view on-screen notifications that provide you with timely information about the state of the cluster.
When you scale down the Query Engines, you reduce the capacity of the node, including the number of cores and memory. Conversely, when you scale up the Query Engines, you increase the capacity of the node by adding more cores and memory.

Points to know

From Kyvos 2023.2 onwards, you can specify the time (in minutes) to hold the queries when query engines are started by using the QE_STARTUP_QUERY_HOLD_TIME property. Note that if the query engines become active before the configured time, the query will be served; otherwise, it will fail.
Upon completion of the Change Cluster State schedule, the load-based rules will be implemented automatically if they have been saved. If no rules have been saved on the Load screen, the cluster will run using the Change Cluster State capacity.
In the future, it will be possible to scale the cluster even further. For instance, if the cluster was previously scaled down from maximum to moderate capacity, it can be scaled down again to the next lower capacity. Similarly, it can also be scaled up to increase its capacity.
If the query engine scales down to moderate capacity, you cannot further scale down the Query Engine to low capacity.
The Kyvos cluster and the Query Engines automatically scale up at full capacity for any MOLAP query.
The Query Engines do not start for any ROLAP queries.
The scale-down duration cannot be less than 30 minutes.
The duration to shut down the query engine must be at least 30 minutes greater than the scale-down duration.
If the cluster is down and a query is executed, the first query triggers the cluster startup process, and all queries fail until the cluster is up and running. In this case, the following messages are displayed:
- Message 1: "Could not serve the query as Query Engine Cluster is not available. Query Engine is launched. Please try after some time."
- Message 2: "Could not serve the query as Query Engine Cluster is starting. Please try after some time."
During the transitioning period of Query Engines; such as scale-up, scale-down, or shut down; you can design and refine the semantic model because the Coordination Master is always up and running .
The following tables specify the approximate time required to complete the process during the transitioning period.

To set the scaling modes for Load-Based scaling, perform the following steps.

On the page, select the Load option. The Load page is displayed.
To set the scaling mode, select one of the following:
1. Managed: Select the capacity option as Low, Moderate, High or Full from the list to start Query Engine and BI Server when any query is fired.
2. Custom: To set scale up and scale down the query engines and BI Server.
  - To scale up,
    1. To start the BI Servers and query engines, select the capacity option as Low, Moderate, High or Full from the list when any query is fired.
    2. Enter a percentage to scale up the cluster if CPU and memory utilization threshold goes above the specified percentage. Also, specify the number of data points and the total number of data points to set.
  - To scale down,
    1. To shut down the BI Server and/or Query Engine, when no queries is fired for specified period of time.
    2. Enter a percentage to scale down the cluster if CPU and memory utilization threshold remains below the specified percentage. Also, specify the number of data points and the total number of data points to set.
Click Save. The load-based scaling mode is set.

NOTE

A data point is information on resource utilization captured every 30 seconds.

Changing cluster state

You can change the cluster capacity as required. This change comes into effect immediately and overrides any existing schedules or rules.

Note

When you change the cluster state, the system enters a transition phase from one schedule to another. During this phase, you cannot modify the Query Engine capacity.
Once the transition phase is complete, you can change the cluster capacity as needed.

To change the cluster state, perform the following steps.

Click the Change Cluster State link available in the Cluster Current State section.
The Change Cluster State dialog is displayed.
Every cluster capacity option is associated with a title. The associated title is displayed in the Title box when you select a required capacity option.
For example, if you choose the capacity option as Full, the title associated with the Full cluster capacity option is displayed in the Title box. You can use the default title or change the title for that schedule as needed.
Select the date and time when you want to change the cluster state.

Note

The difference between the end time and the current time cannot be less than an hour.

Select whether the service should remain online or offline.
Click to select a query engine capacity.
Click Change Now.
The Schedule entry is displayed in the schedule.

Subscribe to notifications

Click the Actions menu (...) located in the upper right corner of a cluster scaling page, and then click Subscribe Cluster Events to subscribe to receive notifications for cluster events. These events include the cluster shutting down as per schedule, the cluster state scaling down due to Cluster scheduling, all query engines being up and in a running state, the cluster state scaling down, entity modifications, or a license expiring soon.

You can set up notifications and subscriptions as needed.

Tip

If you subscribe to the Notify if the cluster shuts down as per schedule, you will receive an email with details on how to restart the cluster if you need to manually.

View cluster audit logs

Click the Actions menu (...) located in the upper right corner of a cluster scaling page, and then click View Audit Logs to view audit logs for cluster scaling events. This opens the Monitor page and displays all the events that have occurred on your cluster along with the details. You can view details about default and user-created schedules and download logs of any cluster scaling event.