Research Workloads on Preemptible VMs (Cloud Next '18)

Опубликовано 25 июля 2018, 0:38

At Two Sigma, researchers run compute workloads that process enormous amounts of data in batches that are parallelized across nodes. In theory, these workloads are well suited to take advantage of Preemptible virtual machines (VM) on Google Cloud Platform (GCP). In practice though, if systems are not architected for elasticity or pre-emption, there will be breaking points on the path to migration. This talk opens with an introduction to Preemptible VMs, and will cover lessons learned from running infrastructure systems for researchers, as well as the migration path to efficiently running workloads on Google Compute Engine (GCE) Preemptible VMs.

Event schedule → g.co/next18

Watch more Infrastructure & Operations sessions here → bit.ly/2uEykpQ
Next ‘18 All Sessions playlist → bit.ly/Allsessions

Subscribe to the Google Cloud channel! → bit.ly/NextSub event: Google Cloud Next 2018; re_ty: Publish; product: Cloud - Compute - Preemptible VMs; fullname: Michael Basilyan; event: Google Cloud Next 2018;