Storage Systems Flashcards
3 types of Storage Systems
Cache, Persistent, Object
Example of cache in GC
Memory Store - managed Redis service
1 benefit of cache
low latency - sub-millisecond access
3 problems of cache
- volatile - lost when machine shuts down
- more expensive than SSD or HDD
- can get out of sync with the system of truth (persistent storage)
Cache - quick definition
in memory data store
Persistent Storage - quick definition
durable block storage
Where can you use persistent storage?
Can be attached to VMs in Compute Engine and Kubernetes Engine.
Where is persistent storage located?
On the network. They are not attached to the physical servers hosting your VM. They exist independently of VMs.
Can a VM have locally attached persistent storage?
Yes. VM can have local SSD, but it is volatile.
Types of persistent storage
SSD and HDD
Differences between SSD and HDD
HDD have higher latency, but lower cost.
Network attached SSD are 40/20 times faster (R/W) than HDD. Locally attached SSD are 200/150 times faster.
Max size of SSD/HDD
64TB
4 facts of persistent storage
- can create file systems on them
- data is automatically encrypted
- size can be increased while mounted to VM
- can be mounted in read-only mode on multiple VMs at once
Is persistent storage zonal or regional?
Both. Regional replicates data across different zones, but is more expensive than purely zonal storage.
What is Object Storage good for?
large volumes of data that is shared widely
3 storage data models
- object
- relational
- NoSQL
GC app example of object model storage
Cloud Store
How are object model objects stored?
Atomically. Cannot read parts of an object. Must copy to server, make changes and then copy object back to object storage system. Used when you don’t need fine-grained access to data within the object while it is in the object store
3 GC app examples of Relational model storage
- Cloud SQL
- Cloud Spanner
- Big Query
3 facts of Relational model storage
- supports frequent queries and updates to data
- allows for consistent view of data
- supports database transactions (Cloud SQL and Cloud Spanner)
3 GC app examples of NoSQL model storage
- Cloud Datastore
- Cloud Firestore
- Bigtable
Benefits/Limitations of 3 types of Storage systems.
- cache is fastest but most expensive and volatile.
- persistent is used for things that need block storage. SSD are faster but more expensive.
- object storage used for large volumes of data for long periods of time.
5 things to consider when planning object storage
- frequency of read/write
- consistency
- transaction support
- cost
- latency
Planning storage - frequency of read/write
Best for structured data that is frequently accessed
Cloud SQL
Planning storage - frequency of read/write
Best for global database that supports relational read/writes
Cloud Spanner
Planning storage - frequency of read/write
Best for writing data at high rates and in large volumes
Bigtable
Planning storage - frequency of read/write
Best for writing files and downloading them in their entirety
Cloud Store
Planning Storage - Consistency Strongest consistency (2)
Cloud SQL
Cloud Spanner
Planning Storage - Consistency
Good for unstructured data
Datastore
Planning Storage - Transaction Support
3 apps that support transactions
Cloud SQL
Cloud Spanner
Datastore
Planning Storage - Latency
Fastest
Bigtable
Planning Storage - Latency
Globally consistent and scalable
Cloud Spanner
