Notes from my Google Cloud Associate Cloud Engineer — Section 1
I passed the Associate Cloud Engineer exam. I successfully cleared this exam after 2 months of lecture and practices. In this article I want to share with you the first part of my notes that I took during the preparation steps.
The exam needs a lot of preparation. Before diving in the notes I recommend these useful resources :
- Architecting with Google Cloud Platform Specialization.
- Google Cloud Platform Associate Cloud Engineer Practice Test.
- Google Kubernetes Engine
Billing
- You can create budgets and attach them to project or a billing account, then define alert based on percentage.
- You can export billing information to BigQuery(dataSet) or Cloud Storage (bucket).
- Billing is accumulated from bottom to top ( billing accumulates all the resource consumption of a specific project).
Pricing calculator
BigQuery
- On-demand (better) : storage pricing & query pricing
- Flat-rate : storage pricing
Storage options
Cloud bigTable
- Size Petabyte
- Key value API ( you can have only one index..)
- No-Sql database
- For flat large data, analytics and big data application use ( IoT , machine learning ..)
- Manage scalability
- Heavy read/write query
- Low latency
- Single key lookup
- Structured
Cloud Datastore
- Size Terabyte
- AppEngine storage option
- kind -> Entity -> properties
- Persistant HashMap
- Structured
- NoSql database for application
- with transactions and SQL-like queries (GQL)
- Multi-regional / Regional
- You can have more than one index / you can set specific column as index in specific row.
Cloud Storage
- Size Petabyte +
- Like file in file system
- Structured & Unstructured
- Immutable object
- The only database that have unstructured objects
- Multi-regional / Regional
- The long-term storage location for data
- They are immutable, and new versions overwrite old unless you turn on versioning.
Cloud Spanner
- Size Petabyte
- The best of Relationnel and non-relationnel database
- Transactional / horizontale scalability
- Sql
- Entreprise grade security
- Multi-regional / Regional
Cloud Sql
- Size Terabyte
- Relationnel database
- Transactional
- More faster than cloud spanner
- Sql for application
- Scope is only regional ( you can’t make replicas within different region , but it is possible in different zone )
- Mysql / posture database
BigQuery
- Size Petabyte
- Sql
- Analytics target
- Warehouse
- ServerLess
- Scope global
2 — Load Balancing :
- Load balancing is not a hardware. It is some rules that you apply to traffic based on criteria like capacity and distance etc.
- Load balancer receives traffic and redirect it to different instances (the closet group of instance. If the closet doesn’t have enough capacity , it will redirect it to the next closet that have enough capacity)
- Load balancer as a server is much more flexible than a hardware load balancer
HTTP(S) Load Balancing
- If the selected backend Service don’t have the capacity to make a load balancing ( CPU utilization, request per seconds..) , the target proxy will choose an other back-end service to do the job.
- Cross-region http(s) load balancing will make the load balancing to the closet region which have enough capacity to support the charge
- Cross-content load balancing will make the load balancing to the region based onURL & header.
- Each back-end service have a collection of instance groups.
Network Load balancing (regional non proxied load balancer)
- the traffic is forwarding to a target pool that choose which health check it will perform on the instances.
- The advantage of the managed instances is the autoscaler, depending on traffic coming to the target pool you can spin up / spin down the instances and this changement will be recognized by the target pool
- Each project can have up a 50 target pool.
Best practices of Load Balancing :
- Make a firewall rules to restrict the traffic only for a specific LoadBalancer in the GCP network.
- Disabling external ip adresses.
Autoscaling :
Set an autoscaler per managed instance group to scale-up or scale-down based on traffic load ( zonal managed instance group / regional managed instance group) . The autoscaler make an horizontal scalability.
Autoscaling policies :
- Average CPU utilization : maintain cpu utilization among managed instance group to the target utilization ( for example it shouldn’t exceed or be very less from 70% ) .If you exceed 70% you should scale up instances
- You can use Stackdriver’s metrics.
- The autoscaler allows multiple policies in a one instance group.
- It collect instance based in policies in order to have the maximum number of instances.
- You can define custom metrics in Stackdriver and use them as base for the autoscaler
I hope you have found the above resources useful for the exam. Please let me know if I should add anything else to this list.