During backup, each asset is divided into smaller slices based on threshold values. Each slice is then serviced by an individual stream.
- Number of Slices = (Assets Size) / (Slice Size)
The threshold slice size is 200 GB and/or file count of 1 million, with a tolerance of 30%. - When determining the number of Protection Engines, using a factor of 1.2-1.5x size of the preceding slice count is recommended.
- Each Protection Engine supports up to 24 concurrent streams.
To achieve optimized throughput, follow these guidelines:
- Number of Protection Engines = (Number of Slices) / 24 (24 is the total count of streams per Protection Engine, where eight streams a piece are served by a different container).
- With the current PowerProtect Data Manager v19.9 release, the recommendation is to scale up to 11 Protection Engines for larger shares (for example 50 TB or larger).
- To achieve optimum performance, it is recommended to use a dedicated 10 GbE network per Protection Engine. The Protection Engine throughput is bounded by underlying network stack on ESXi host. Hence, a dedicated 10 GbE network per Protection Engine would achieve better performance.
- Multiple NAS Protection Engines with a dedicated 10 GbE network can achieve better net aggregated throughput. This includes reading the data from NAS array and writing it to protection storage.
- If the whole environment is 10 GbE network (NAS array, PowerProtect DD series and multiple Protection Engines), the overall throughput is bound by 10 GbE network speed.
- If the read throughput from the NAS array and write throughput to PowerProtect DD series causes a bottleneck with multiple Protection Engines, it is recommended to have more network ports on NAS array and on PowerProtect DD series.
- The asset parallelism per asset helps to load balance the number of streams across multiple shares. Asset Level Parallelism enables all asset backups to run in parallel, and each of these assets has many concurrent streams (as per user Input of asset parallelism). Also, if there are enough containers available, all these assets will run in parallel. The Asset Level Parallelism parameter maximum-supported count with PowerProtect Data Manager v19.9 release is 256 concurrent streams per asset.
- Also, we can use the sizing tool created by Dell Technologies to determine the number of protection engines which must be deployed for a certain protection load. Contact Dell Technologies support to download and use a recent version of the tool. The yellow highlighted sections in the tool can be edited to give us the approx. number of proxy engines which need to be deployed without any manual calculation. Some of the inputs required from the customer are:
- Array type (PowerScale, Dell Unity, PowerStore, or generic).
- Total amount of NAS data to be protected in TB.
- Total number of files in millions.
- Expected backup duration for full (Gen 0) backup.
- Network parameters such as the number of array nodes/ports, PowerProtect DD series ports, and proxy ESXi ports involved in the backup. The same data will be used to account for any bottlenecks in the proxy engine calculation.
- Expected backup duration for synthetic full/incremental (Gen 1) and the approximate change rate expected between backups.