Soru

Zorluk: Çok zorDesigning Migration Plans and Data Transfer Strategies

A healthcare organization is migrating its core platform from an on-premises data center to Google Cloud. The migration scope includes 850 TB of archived historical imaging data and an active 12 TB PostgreSQL transactional database supporting critical clinical workflows. The organization has an existing 1 Gbps Dedicated Interconnect with an average baseline utilization of 70%. Security policies strictly mandate that all data must travel over private connectivity without traversing the public internet, and the maximum permissible downtime window for the transactional database during final cutover is 4 hours. Which data transfer and migration strategy should a Cloud Architect recommend to satisfy all business and technical constraints?

  1. A
    Stream the 850 TB imaging dataset over the Dedicated Interconnect using Storage Transfer Service while concurrently configuring continuous replication for the PostgreSQL database using Database Migration Service.
  2. B
    Provision multiple High Availability (HA) VPN tunnels over the public internet to burst total throughput up to 10 Gbps for transferring the 850 TB imaging dataset, while performing an offline database dump and restore over the Dedicated Interconnect during the maintenance window.
  3. Order a GCP Transfer Appliance to move the 850 TB imaging dataset offline, while using Database Migration Service over the Dedicated Interconnect with continuous change data capture (CDC) to synchronize the PostgreSQL database prior to cutover.Cevap
  4. D
    Order a GCP Transfer Appliance for the 850 TB imaging dataset, but migrate the PostgreSQL database to Cloud Spanner by exporting the database and running a full schema and bulk data import during the 4-hour cutover window.

Cevap

The optimal strategy is ordering a GCP Transfer Appliance to transport the 850 TB imaging dataset offline while utilizing Database Migration Service with continuous change data capture (CDC) over the Dedicated Interconnect for the 12 TB PostgreSQL database.
The solution combining GCP Transfer Appliance for the 850 TB unstructured dataset with Database Migration Service (DMS) continuous replication for the 12 TB database is the only strategy that satisfies all constraints. Transfer Appliance moves petabyte-scale data offline without saturating the limited 300 Mbps available bandwidth, while DMS over Dedicated Interconnect performs online CDC to keep downtime during cutover well under the 4-hour SLA without traversing the public internet.

Adım Adım Çözüm

1
Calculate available network bandwidth and time required for online data transfer of the unstructured dataset.
With 70% of the 1 Gbps link utilized, only 300 Mbps (0.3 Gbps0.3\text{ Gbps}) remains available. Transferring 850 TB=6.8×106 Gb850\text{ TB} = 6.8 \times 10^6\text{ Gb} at 0.3 Gbps0.3\text{ Gbps} takes approximately 6.8×1060.322.67×106 seconds262 days\frac{6.8 \times 10^6}{0.3} \approx 22.67 \times 10^6\text{ seconds} \approx 262\text{ days}. This mandates an offline physical appliance transfer.
Large multi-hundred terabyte datasets cannot be moved online over low-bandwidth connections within reasonable operational timeframes.
2
Evaluate connectivity and security constraints for the offline data transfer.
Transfer Appliance provides hardware-encrypted physical ingestion that avoids public internet routing, satisfying compliance requirements.
Security policy strictly forbids sending sensitive healthcare data over the public internet.
3
Design the database migration path to achieve a cutover window under 4 hours.
Using Database Migration Service (DMS) over the Dedicated Interconnect enables continuous initial snapshot sync followed by Change Data Capture (CDC). During cutover, only the remaining minimal replication lag must be processed, completing cutover in minutes.
Offline dumps of a 12 TB database take many hours to export and restore, exceeding the 4-hour maximum downtime SLA.

Anahtar Kavram

Selecting appropriate GCP data transfer mechanisms based on data volume, bandwidth constraints, security policies, and downtime SLAs.
Bu soruyu puanla