A naukri.com initiative
Cloudblog
1M
306
Image Credit: Cloudblog
How we improved GKE volume attachments for stateful applications by up to 80%
- Google Kubernetes Engine has improved the speed at which Persistent Disks (PDs) are attached and detached by up to 80%.
- Less latency means better performance for user workloads and interactions with storage.
- The Google Compute Engine Persistent Disk (GCE PD) Container Storage Interface (CSI) Storage Plugin now allows seamless storage access for Kubernetes workloads, making it easier to manage Compute Engine PDs in a GKE cluster.
- Tens to hundreds of PDs were being attached to VMs, creating longer downtime during software upgrades.
- The solution was the introduction of operation merging in Compute Engine, which transparently parallelizes detach and attach operations and optimizes PD attachment and detachment processes.
- The best part is no customer action is needed and they automatically benefit from 80% improvement in p99 latency for attach workloads and 60% for detach workloads.
- The custom solution also introduces a tiered approach for workflow execution and decouples incoming attach and detach requests from the actual execution of volume attachments.
- Incoming attachDisk/detachDisk requests aren't queued within the same database row as the target Compute Engine instance entity to minimize database contention.
- The technology aims to allow large-scale stateful applications to run seamlessly on GKE clusters.
- Further information is available on the Google Kubernetes Engine website.
Read Full Article
18 Likes
For uninterrupted reading, download the app