In Alibaba Cloud Elasticsearch Agent Engine Edition, Branch Slice shortens the data preparation time for knowledge bases, code repositories, and application experiments. Within the same Collection, you can derive a new Slice from an existing Slice and then read and independently modify the data without waiting for a full copy to complete. A branch reuses the existing immutable data, so creating one does not copy documents one by one.
For example, an application can use the Slice tenant-a to create tenant-a-preview and then immediately modify the preview data, compare query parameters, or run evaluations. Document modifications made during an experiment do not affect the source Slice, so you can independently validate different approaches on the same initial data.
When to use Branch Slice
| Requirement | Option |
| You need a data copy that you can read from and write to independently as soon as possible. | Branch Slice (Beta): read and write a single Slice immediately after creation. Use it to quickly prepare a copy of a knowledge base, test data, or application experiment data. Functional limits apply to the branch before it converts into a regular Slice. For more information, see . |
| You need full capabilities immediately, or you want the target to use a different number of primary shards or routing partitions. | Copy Slice: copies documents one by one and rebuilds the index based on the shard configuration of the target backing index. After the copy completes and the data becomes searchable, the result is a fully functional regular Slice without the limits that apply to a Branch Slice before conversion. However, you incur the time and read/write overhead of a full copy. |
| You need to validate different field types or vector index configurations. | Create a new Collection. Slices within the same Collection share mapping and index settings. |
A branch can also be used long term. Your application can keep accessing data through the same Slice name. You do not need to make another copy just for long-term retention. When you choose between Branch Slice and Copy Slice, consider the creation speed, the copy overhead, and the query capabilities that your business requires.
Branch Slice limits before conversion
Branch Slice is currently in Beta. A newly created branch supports reading from and writing to a single Slice. It does not yet support querying the branch together with other Slices or running cache warming. These limits are lifted after the branch automatically converts into a regular Slice in the background.
When you create a branch, you choose a wait method. The available wait methods differ in whether the data starts from the same point after fault recovery. For the behavioral differences and the recovery guarantees during creation, read/write, and the Active period, see Branch Slice (Beta) reference.