Azure Data Explorer support of Lsv3-series SKUs was announced in June 2022. The Lsv3-series comes in 2 flavors: Lsv3 (Intel) and Lasv3 (AMD). Since the announcement 100s of Kusto clusters and 1000s of cores have migrated to the new SKUs.
The Lsv3/Lasv3 SKU families are now the Storage Optimized preferred SKUs of Kusto. Specifically, when comparing to “older” SKU families such as Dsv2, Dv2, Ls and Lsv2, the benefits of migrating to the Lsv3/Lasv3 families are remarkable:
Cost reduction
Depending on your region and current SKU, the cost savings range of migrating from Dsv2 to Lv3 SKUs is ~30-40%. This is due to the following factors:
Following the migration, the recommendation is to turn on the Auto-scale feature. In many cases following the SKU migration, the cluster size will scale-in , reducing the nodes count of the cluster. Smaller clusters have less Compute resources, further reducing the cluster costs.
Better Performance
The new SKUs are built on modern CPUs. Following migrations clusters experience improved performance: average CPU usage reduction and improved query latencies. In addition, the new SKUs have local SSD disks. The old SKUs had remote attached Premium Storage disks. Kusto clusters perform better with local disks.
New Seamless SKU Migration process
We have recently released a new seamless SKU migration process. Previously, the SKU change was a sequential process. During the creation of the new “target” VMSS and its cache warming process, the cluster service was not available. In the new and improved process, we prepare the new VMs in parallel- while the “old” source cluster continues to provide service. The “warming” of the cache of the target SKUs is performed in the background by reading the data from “cold” storage.
The switchover to the new VMSS occurs only after the new VMSS nodes cache warming process is complete.
The Lsv3 and Lasv3 are available in many Azure regions. When available, the Lsv3/Lasv3 SKUs are listed as the Recommended SKUs for the Cluster Create and SKU Migrate/Scaleup operations in the Azure Portal.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.