To improve the performance (upload speed and server load), File Copy can upload data to the specified destination in parallel chunks and or parallel threads.
You can set the number of 1-MB chunks that will be simultaneously sent to the destination. By increasing the number of parallel chunks, you will decrease the time to complete the job, but will also have an adverse affect on server performance. Configure this value as necessary to obtain optimal performance.
For example, if you are performing File Copy for a 10-MB file and set the number of 1-MB chunks to 2, then File Copy will write 10 chunks, two at a time. If you see that this is taking too long to complete the job, change this value to 4. The time to complete the job will then decrease because File Copy will now be writing 10 chunks, four at a time, but the load on your server will increase.
File Copy lets you copy more than one file at a time. By default, File Copy transfers 8 files in parallel when the destination is configured to File Systems and transfers 32 files in parallel when the destination is configured to Cloud. If you see that File Copy is taking too long to transfer the data, increase the number of threads up to 32, to optimize performance. However, if you experience a problem on a machine with less memory, decrease the number of threads.
Chunk Value and Threads for Archive Value can be used together to control the speed of the File Copy. If you increase Chunk Value and Threads for Archive Value, you see File Copy is performed faster.
For example, if you are transferring 8 files with 10 MB each and set the number of 1-MB chunks to 2, then File Copy will write 16 at a time (8 files X 2-MB chunks), but the load on your server will increase. When you see the load on the server has increased to a point where it becomes a problem, decrease the number of threads. If the destination is a Cloud location, it is recommended that you configure these settings in such a way that produces at least 20 writes, to optimize performance.
Restore from a File Copy lets you download more than one file at a time. By default, restores from file copies downloads 8 files when the File Copy location is configured to File Systems and downloads 32 files in parallel when the file copy location is configured to Cloud. If you see that the restore from a File Copy is taking too long to transfer the data, increase the number of threads up to 32.
Note: Chunk Value does not apply to restore jobs.
Catalog Synchronization jobs lets you use multiple threads to optimize performance.
If you see that the catalog synchronization job is taking too long to transfer the data, increase the number of threads up to 10. You will see that the job is performed faster and the load on the server increases. When you see the load on the server has increased to a point where it becomes a problem, decrease the number of threads.
To configure the file copy settings to optimize performance, set the corresponding DWORD values as follows:
"HKEY_LOCAL_MACHINE\SOFTWARE\Arcserve\Unified Data Protection\Engine\AfArchiveDll"
Note: The same registry key is used when your File Copy destination is either File Systems or a Cloud location.
The available range for the number of chunks is 1 - 4 chunks.
Default: 4 chunks
Maximum: 4 chunks
The available range for the number of files is 1 - 32 files.
Default: 8 files when the destination is configured to File Systems and 32 files when the destination is configured to a Cloud location.
Maximum: 32
The available range for the number of files is 1 - 32 files.
Default: 8 files when the copy destination is File Systems and 32 files when the copy destination is a Cloud location.
Maximum: 32
The available range for the number of files is 1 - 10 threads.
Default: 8 threads
Maximum: 10
|
Copyright © 2016
|
|