Please enable JavaScript.

Coggle requires JavaScript to display documents.

GCP_Storage, Compare, Compare - Coggle Diagram

- - - - can setup a schedule
      - support incremental transfer (only changed objects)
      - reliable and fault tolerance
      - can filter files by name. path, etc
    - - less 1TB - gustil or API
      - more than 1 TB, transferring from separate cloud
    - - recommended - >20 TB or transfer takes more than a week
      - Can use customer managed keys
      - copy and ship - physical transfer
      - need to lease specific devices in Google
  - - - Storage class
        
        Standart - short term storage or frequently accessed, none - min storage duration
        
        Nearline - best for backups, and data accessed less than once a month, min 30 days
        
        Coldline - disaster recovery and access less than once per quarter, min 90 day
        
        Archive - less that once a year, min - year,
        
        Availability is bigger for Standart and less for Archive, none - min duration
      - Access control - fire grained or unified
      - where to store
        
        Dual Regional (for HA and low latency)
        
        Multi-regional - HA across large areas
        
        Regional
      - Observations
        
        inexpensive
        
        serverless and autoscalling
        
        can have folders
        
        store large objects as a key-value approach, threats objects as a units
        
        URI is a key
        
        "Object Content" is a object
        
        store objects
        
        when need to update - replace object fully
        
        No needs to say storage
        
        provide rest API access to file. CLI, programmatic connetctio
        
        Store any types of file
        
        max object 5 TB, but we can store any amount of it
        
        low latency, first bytes within 10 milliseconds
      - name - unique globally
  - - - Server side
        
        Customer managed (from KMS) - need to add Service Account to touch KMS
        
        Customer suppliers - supply key for any operation, but bucket doesn't store the key
        
        Google managed (default)
      - Client side - do before upload, GCP is not involved here
    - - identify objects by set of parameters: age, created time... - can be set
      - might be moved to another class or deleted. It is possible only for down strategy: `
      - automatic setup
    - - enabled on bucket level
      - live object - current object, once we delete it - it is "non-current object"
      - Prevent deleting
      - older versions unique identifier: object_key + a generation number
      - reduce costs - delete old versions
    - - custom metadata
      - non-editable: size, last mod, encryption type
      - fixed key - predefined by GCP, like content type, cache-controll, content-disposition(download or show in browser)
      - metadata - set of KV parameters on object
    - - you can not remove objects or reduce retention period
      - no options to delete the bucket
      - data retention policy
      - bucket lock - no options to modify. (locking retention policy) - can be add any time
    - - ACL - object level control, need to set Who and What Level to the single object (fine-grained)
      - IAM - bucket level control (uniform)
      - Sign URL
        
        used Service Account
        
        provide limited time access to the object by SA. need to create a sign URL
- - - - zonal
      - regional (replicatied in multiple zones) - costly
      - network storage, more durable
      - attached to VM, can be re-attach to another machine
      - very flexible: can increase size, still high performance
      - provisioned capacity
      - Snapshot supported
        
        it's backup
        
        can schedule
        
        retention policy is allowed
        
        can be regional or multi-regional
        
        can be incremental
        
        reduce performance
      - types
        
        Balance - SSD
        
        Balance -cost-performance
        
        SSD - (SSD) - Good for bigdata - really expensive
        
        High performance
        
        Standard - HDD - bad for random IO - cheap
        
        BigData - cost effective
      - Scenarios
        
        increase durability - go to regional PD
        
        disaster recovery - hourly backups
        
        Improve perf - increase size or add more Persis. Disks(PD), increases vCPU
        
        delete old images - setup retention policy
      - Attache to running VM: attach, format, mount
        
        Resize: In the Cloud Platform Console, increase the size of the persistent disk and use the resize2fs command in Linux.
    - - lifecycle related GCE , ephermal storage
      - Encryption key - only google managed
      - go to bigger if need to improve performance
      - Pros and cons
        
        good for temporary information (+)
        
        Ephemeral Storage (-)
        
        very fast (+)
        
        can not detach (-)
      - Snapshot NOT supported
      - multiple discs can be attached to the VM
        
        Persistance - data
        
        Boot - OS
- - - - disk type:SSD HDD
      - mode: standart and High Scale