Elastic Cloud allows smaller instance sizes to get temporarily boosted vCPU when under heavy load. vCPU boosting is governed by vCPU credits that instances can earn over time when vCPU usage is less than the assigned amount.
How does vCPU boosting work?edit
Based on the instance size, the vCPU resources assigned to your instance can be boosted to improve performance temporarily, by using vCPU credits. If credits are available, Elastic Cloud will automatically boost your instance when under heavy load. Boosting is available depending on the instance size:
Instance sizes up to and including 8 GB of RAM get boosted. The boosted vCPU value is
16 * vCPU ratio, the vCPU ratios are dependent on the hardware profile selected. If an instance is eligible for boosting, the Elastic Cloud console will display Up to 2.21 vCPU, depending on the hardware profile selected. The baseline, or unboosted, vCPU value is calculated as:
RAM size * vCPU ratio.
Instance sizes bigger than 8 GB of RAM do not get boosted. The vCPU value is displayed in the Elastic Cloud console and calculated as follows:
RAM size * vCPU ratio.
What are vCPU credits?edit
vCPU credits enable a smaller instance to perform as if it were assigned the vCPU resources of a larger instance, but only for a limited time. vCPU credits are available only on smaller instances up to and including 8 GB of RAM.
vCPU credits persist through cluster restarts, but they are tied to your existing instance nodes. Operations that create new instance nodes will lose existing vCPU credits. This happens when you resize your instance, or if Elastic performs system maintenance on your nodes.
How to earn vCPU credits?edit
When you initially create an instance, you receive a credit of 60 seconds worth of vCPU time. You can accumulate additional credits when your vCPU usage is less than what your instance is assigned. At most, you can accumulate one hour worth of additional vCPU time per GB of RAM for your instance.
For example: An instance with 4 GB of RAM, can at most accumulate four hours worth of additional vCPU time and can consume all of these vCPU credits within four hours when loaded heavily with requests.
If you observe declining performance on a smaller instance over time, you might have depleted your vCPU credits. In this case, increase the size of your cluster to handle the workload with consistent performance.
For more information, check Elasticsearch Service default provider instance configurations.
Intro to Kibana
ELK for Logs & Metrics