Bigtable is designed for high-volume writes and storing large volumes of data. BigQuery is designed for large volumes of data, but it is not optimized for writes the way Bigtable is. Datastore and Firestore are not designed to scale to large data volumes the way Bigtable is.

Google Cloud Interconnect – Dedicated is the only option for a dedicated connection between a customer's data center and a Google data center.

Multizone/region clusters are available in GKE and are used to provide resiliency to an application, so option B is correct. Option A refers to instance groups that are a feature of Compute Engine, not directly of Kubernetes Engine. Option C is incorrect; regional deployments is a fictitious term. Load balancing distributes load and is part of Kubernetes by default. If load is not distributed across zones or regions, it does not help to add resiliency across data centers.

Pods are single instances of a running process in a cluster, so option C is correct. Pods run containers but are not sets of containers. Application code runs in containers that deployed in pods. Pods are not controllers, so they cannot manage communication with clients and Kubernetes services.

The only states a Kubernetes deployment can be in are progressing, completed, and failed.

The options for setting scopes are: Allow Default Access, Allow Full Access, and Set Access For Each API, so option B is correct. Option A is incorrect; it is missing Set Access For Each API. Option C is incorrect; it is missing Allow Default Access. Option D is incorrect; it is missing Allow Full Access.

App Engine provides three ways to split traffic: by IP address, by HTTP cookie, and by random selection.

VOIR IMAGE 1 :
--->
Think of app engine standard dot notation
instance number is the lowest level of addressing you can get. e.g. 1.version.service.app-id.appspot.com
https://cloud.google.com/appengine/docs/legacy/standard/python/how-requests-are-routed#targeted_routing

Shield VM is an advanced set of security controls that includes Integrity Monitoring, a check to ensure boot images have not been tampered with, which makes option B the right answer. Firewalls are used to control in ingress and egress of network traffic to a server or subnet. Project-wide SSH keys are used for authenticating users across servers within a project. Boot disk integrity check is a fictional feature.

Cloud Trace is a distributed tracing application that provides details on how long different parts of code run, so option C is correct. Option A is incorrect; monitoring is used to notify DevOps engineers when resources are not functioning as expected. Option B is incorrect; Logging is for collecting, storing, and viewing log data, and although log entries might help diagnose bottlenecks, it is not specifically designed for that. Option D is incorrect; Debug is used to generate snapshot and inject logpoints.

Specialized services are monitored by Google so users do not have to monitor them

Exporting from Dataproc exports data about the cluster configuration, which makes option C correct. Option A is incorrect; data in dataframes is not exported. Option B is incorrect; Spark does not have tables for persistently storing data like relational databases. Option D is incorrect; no data from Hadoop is exported.

Bigtable is designed to accept billions of rows of data. Collecting data from 100,000 sensors every 5 seconds will generate 6,000,000 data points every minute, or 8,640,000,000 data points per day. Spanner is a relational database and supports transactions, but they are not needed. Cloud SQL MySQL and Cloud SQL PostgreSQL would be difficult to scale to this level of read and write performance.

gcloud by default will retry a failed network operation and will wait a long time period before each retry. The time to wait is calculated using a truncated binary exponential back-off strategy.
Cloud DataProc is the managed Spark service. Cloud DataFlow is for stream and batch processing of data, BigQuery is for analytics, and Cloud Hadoop is not a GCP service.

You can find the number of vCPUs on the cluster listing in the Total Cores column or on the Details page in the Node Pool section in the size parameter, making option D correct. The Labels section does not have vCPU information.

The Deployment Manager is the name of the service for creating application resources using a YAML configuration file, so option B is correct. Option A is incorrect although you could use scripts with gcloud commands to deploy resources in Compute Engine. Options C and D are incorrect because those are fictitious names of products.

Shield VM is an advanced set of security controls that includes Integrity Monitoring, a check to ensure boot images have not been tampered with

When a user is granted iam.serviceAccountUser at the project level, that user can manage all service accounts in the project, so option A is correct

All Google regions have the same level of service level agreements, so reliability is the same