You produce 20,000 CSV files/hour (<4 KB each) from on-prem to GCP over a 50 Mbps link with 200 ms latency using an SFTP VM. Volume may double seasonally. Which two actions should you take? (Choose two.)

Redesigning ingestion to use gsutil for parallel uploads is effective because gsutil is optimized for high-performance transfers to Cloud Storage, leveraging parallelization and resumable uploads. This significantly improves throughput compared to SFTP, which is often single-threaded and less efficient for many small files. Bundling files into TAR archives reduces the overhead associated with establishing a new connection for each small file. Each SFTP transfer incurs latency and handshake overhead. By combining 1,000 files, the number of individual transfers decreases dramatically, improving the effective data transfer rate, especially over high-latency links. Per-file compression is unlikely to increase throughput for already small files (<4KB) and might even add CPU overhead without significant size reduction. Increasing bandwidth alone won't solve the high-latency issue or the inefficiency of SFTP for many small files. Storage Transfer Service is designed for large-scale, one-time or recurring transfers, but exposing an S3-compatible endpoint on-prem adds complexity and may not be the most direct solution for continuous, high-volume small file ingestion compared to gsutil or bundling.