Cache policy (hot and cold cache)
Azure Data Explorer stores its ingested data in reliable storage (most commonly Azure Blob Storage), away from its actual processing (such as Azure Compute) nodes. To speed up queries on that data, Azure Data Explorer caches it, or parts of it, on its processing nodes, SSD, or even in RAM. Azure Data Explorer includes a sophisticated cache mechanism designed to intelligently decide which data objects to cache. The cache enables Azure Data Explorer to describe the data artifacts that it uses, so that more important data can take priority. For example, column indexes and column data shards,
The best query performance is achieved when all ingested data is cached. Sometimes, certain data doesn't justify the cost of keeping it "warm" in local SSD storage. For example, many teams consider that rarely accessed older log records are of lesser importance. They prefer to have reduced performance when querying this data, rather than pay to keep it warm all the time.
Azure Data Explorer cache provides a granular cache policy that customers can use to differentiate between: hot data cache and cold data cache. Azure Data Explorer cache attempts to keep all data that falls into the hot data cache category, in local SSD (or RAM), up to the defined size of the hot data cache. The remaining local SSD space will be used to hold data that isn't categorized as hot. One useful implication of this design is that queries that load lots of cold data from reliable storage won't evict data from the hot data cache. As a result, there won't be a major impact on queries involving the data in the hot data cache.
The main implications of setting the hot cache policy are:
- Cost: The cost of reliable storage can be dramatically lower than for local SSD. It's currently about 45 times cheaper in Azure.
- Performance: Data is queried faster when it's in local SSD, particularly for range queries that scan large amounts of data.
Use the cache policy command to manage the cache policy.
Azure Data Explorer is designed for ad-hoc queries with intermediate result sets fitting the cluster's total RAM. For large jobs, like map-reduce, where you want to store intermediate results in persistent storage such as an SSD, use the continuous export feature. This feature enables you to do long-running batch queries using services like HDInsight or Azure Databricks.
How cache policy is applied
When data is ingested into Azure Data Explorer, the system keeps track of the date and time of the ingestion, and of the extent that was created. The extent's ingestion date and time value (or maximum value, if an extent was built from multiple pre-existing extents), is used to evaluate the cache policy.
You can specify a value for the ingestion date and time by using the ingestion property
By default, the effective policy is
null, which means that all the data is considered hot.
null table-level policy overrides a database-level policy.
Scoping queries to hot cache
Kusto supports queries that are scoped down to hot cache data only. There are several query possibilities.
- Add a client request property called
query_datascopeto the query. Possible values:
- Use a
setstatement in the query text:
set query_datascope='...'. Possible values are the same as for the client request property.
- Add a
datascope=...text immediately after a table reference in the query body. Possible values are
default value indicates use of the cluster default settings, which determine that the query should cover all data.
If there's a discrepancy between the different methods, then
set takes precedence over the client request property. Specifying a value for a table reference takes precedence over both.
For example, in the following query all table references will use hot cache data only, except for the second reference to "T", that is scoped to all the data:
set query_datascope="hotcache"; T | union U | join (T datascope=all | where Timestamp < ago(365d) on X
Cache policy vs retention policy
Cache policy is independent of retention policy:
- Cache policy defines how to prioritize resources. Queries over important data will be faster and resistant to the impact of queries over less important data.
- Retention policy defines the extent of the queryable data in a table/database (specifically,
Configure this policy to achieve the optimal balance between cost and performance, based on the expected query pattern.
hot cache policy= 28d
In the example, the last 28 days of data will be on the cluster SSD and the additional 28 days of data will be stored in Azure blob storage. You can run queries on the full 56 days of data.