Non-replicated disk placement groups
When creating non-replicated disks, resources are automatically distributed across the Yandex Cloud physical infrastructure. Your disks might get hosted on the same physical media. To make multiple disk failures less likely, you can combine your disks into placement groups.
Grouping non-replicated disks reduces the risk of their simultaneous failure. However, the placement requirements are more demanding, so a physical resource shortage is more likely for non-replicated disks in a placement group rather than for the same number of ungrouped disks. You might fail to create a non-replicated disk in a placement group, but succeed to create one outside the group without any issue.
A non-replicated disk can only be placed in a single group and must reside in the same availability zone with this group.
For more information about the organizational and technical limits of placement groups, see Quotas and limits in Compute Cloud.
In a placement group, non-replicated disks are located in the server racks of the data center of the selected availability zone according to one of the placement strategies.
Spread placement
Spread placement is a non-replicated disk placement strategy which ensures that each disk resides on a separate server rack in the data center of the selected availability zone. If one of the server racks hosting the group's disks fails, only one disk will be affected.
Partition placement
Partition placement is a non-replicated disk placement strategy which ensures that partitions with disks reside on different server racks in the data center of the selected availability zone. If one of the server racks hosting the group partitions fails, only one partition will be affected. A single partition may host multiple disks. A single placement group may contain up to 5 partitions.