Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
If your clients upload data by using small block sizes, you can improve performance by choosing an efficient naming scheme. A small block size is less than 256 KiB. Larger blocks aren't affected by partition naming.
Partition keys and efficient naming schemes
The partition key for a blob is account name + container name + blob name. The partition key is used to partition data into ranges and these ranges are load-balanced across the system.
To help the system partition data more efficiently, avoid sequential naming schemes such as log20160101, log20160102, log20160103. These schemes concentrate traffic on one server, which can exceed scalability targets and cause latency issues.
Instead, add a hash character sequence (such as three digits) as early as possible in the partition key of a blob. If you plan to use timestamps in names, consider adding a seconds value to the beginning of that timestamp (for example: ssyyyymmdd).
If you use timestamps or numerical identifiers, avoid append-only or prepend-only patterns. These patterns route all traffic to a single partition which prevents load balancing. However, if you plan to use these patterns, consider splitting data into multiple blobs. Apply a hash prefix to each blob that represents a time interval such as seconds (ss) or minutes (mm). That way traffic isn't repeatedly directed to a single blob on a single partition server which could exceed scalability limits.