Exadata Cloud Development team is pleased to announce "Multiple VMs per Exadata system (Multi-VM)" for Gen 2 Exadata Cloud@Customer
The Multi-VM capability allows customers to create Multiple VMs in a compute node of an Exadata Cloud@Customer to enable better isolation and consolidation.
All the Gen 2 ExaC@C systems deployed up until now have been deployed with a single VM cluster per ExaC@C. This single VM cluster consumed all of the memory, local disk space and Exadata storage available in the system. Customers decided the number of OCPUs they wanted to allocate to the single VM cluster. For example, for a quarter rack system, the VM cluster would be a 2-node RAC cluster with one guest VM in each node that occupies all of the local disk space as well as all of the memory of that node. The total OCPUs enabled during the VM cluster creation process were equally divided between the two nodes and this VM cluster consumed all of the available Exadata storage distributed among DATA, RECO and SPARSE (If selected) disk groups.
Below are the details on how existing and new customers can use this new feature:
Scaling down Single VM Cluster (Pre-requisite for Existing Customers):
Since the first VM cluster consumed all of the local disk, memory and Exadata storage, the first task all customers deployed with a single VM cluster will need to do, in order to create multiple VM clusters, is to shrink their first and the only VM cluster before resources can be freed up to create additional VM clusters. The new UI interface “Scale VM Cluster” introduced with Multi-VM enables customers to scale down the OCPUs, memory, local disk and exadata storage for the VM cluster allowing them to shrink their existing VM cluster as shown in the screenshot below.
Once resources are freed up after the first VM cluster is shrunk, customers will be able to create additional VM clusters as well as shrink or grow existing VM clusters.
Notice that the interface above allows customer to provide a new value for all of the four resources – OCPUs, Memory, local file system size and Exadata storage - at once. Clicking on the “save changes” button will create a master work request with four child work requests, one for each resource, that can be monitored from the work request page. The OCPUs and Exadata storage changes are online while the memory and local file system changes are done in a rolling manner, one node at a time.
Creating New VM Cluster (Existing and New Customers):
New customers or existing customers who have scaled down their single VM cluster and whose systems have received the Multi-VM update for their region (see rollout plan section at the bottom), can now create multiple VM clusters as shown in the screen below. It is imperative that these customers do a careful planning involving their consolidation and Exadata deployment strategy to decide the size and number of VM clusters they want to deploy per ExaC@C rack. Even though Multi-VM gives them the flexibility to adjust the VM cluster size after deployment or even to add or remove clusters, changing memory and local disk space does require rolling changes and could affect their workloads.
Multi-VM introduces the following service minimums and maximums.
One additional change that is accompanying the Multi-VM rollout is the default Release Update (RU) used for creating a new VM cluster or database. The default RU, which is currently April 2019, is changing to Jan 2020. This means when the new customers whose systems were deployed after the Multi-VM feature was rolled out to their region creates a new VM cluster or a database, it will be created with Jan 2020 RU. Existing customers will need to use "dbaascli cswlib download" command to download and "dbaascli dbimage activateBP" command to activate any image other than the default image.
Multi-VM delivers a highly sought-after functionality to Gen 2 ExaC@C customers. This capability now allows enterprise customers to enable database consolidation along with workload isolation, ensuring that critical application workloads are not impacted by unexpected demand surges for other applications deployed in the same system.
* Limited by local disk space
Are you looking for more Gen 2 ExaC@C Blog Posts?