In environments with strict security requirements, SFTP is the most common protocol for loading data into a gbase database cluster. The latest versions of GBase 8a introduce configurable parameters that can significantly boost SFTP loading throughput. This guide explains how to assess your current load speed and apply the new parameters for optimization.
When to Consider Tuning SFTP Load Performance
If your current loading speed already meets business requirements, keep the default settings. When speed is lower than expected, first check the per‑node load rates via the information_schema.load_status table. Compare the aggregate cluster load speed against the raw SFTP transfer speed (e.g., a simple sftp get test on a single node). If the cluster’s total load speed is well below 30% of the raw SFTP speed, the parameters described below may help.
Key Parameters
gbase_use_ssh_sftp
-
Default:
0(legacy SFTP loading method) -
1: Enable the new high‑performance loading mechanism - Scope: Session‑level
SET gbase_use_ssh_sftp = 1;
gbase_use_ssh_sftp_algorithm
- Default: Empty (auto‑negotiation of encryption algorithm)
- Usage: Specify the encryption algorithm(s) for SFTP transfer, separated by commas
- Scope: Session‑level
SET gbase_use_ssh_sftp_algorithm = 'aes128-ctr';
Important: The chosen algorithm must be supported by the SSH daemon on the data source side, otherwise the connection will fail. Select a faster algorithm from the list of those supported by the server’s sshd.
Best Practices
- Validate the performance gain in a test environment or via session‑level settings first.
- Once the improvement is confirmed, write the parameters into the configuration file for global effect.
- The new method will increase CPU, network, and disk consumption; ensure sufficient resources.
- An incorrectly configured encryption algorithm will cause SFTP negotiation to fail immediately — choose carefully.
By applying these settings in your gbase database environment, you can overcome SFTP loading bottlenecks while maintaining secure data transfer. Always test first and monitor resource usage after rolling out to production.
Top comments (0)