AWS section 6 dynamoDB

Low latency NoSQL non relational database

Has tables, items and attributes

supports document AND key-value data models

supported document formats are JSON, HTML and XML

partition key 

Combinatino of partition key_sort key (composite key)

Strongly consistent and eventually consistent.

Strongly consistent gives most recent stuff.

• With IAM policies
• Fine grained access control can happen using IAM condition paramter

dynamodb:leadingKeys. Allows users to access only the items where the partition key value matches their user ID.

Useful for video games where you cant view other peoples data

Both provide in memory caching and both are supported for dynamo db

DAX is optimised for use with dynamo db. DAX only supports write through caching strategy. Doesnt support lazy loading.

Elasticache should be used for RDS.

Between application and database

Lazy loading only caches the data when its requested


Write through adds data in the cache when data is written to the database

• Pros
• Data in the cache never gets stale
• Users are more tolerant of additional latency updating data than retrieving it

• Cons
• Wirte penalty. You're writing the cache AND the database each time

If a node fails and a new one is spun up, data is missing until added or updated in database

Wasted resource if most of the data is never read.

Lazy loading:

• Pros:
• - Only requested data cached.
• - node failures not fatal

• Cons:
• - Cache miss penalty exists
• - Each cache miss results in a request for data from the cache in 3 trips (Initial request for data from the cache, query of db for data, writing data to cache)
• -stale data
• If your data only gets updated when theres a cache miss then it can get stale. Missing data.cache churn. Most data not read. Can add TTL to fix this

add a ttl. time to live

Remember how we said that the partition key was unique.

Well that might not be the case cos they could have the same partition key like a user posting to different forums.

They enable fast data fast queries on specific data columns.

• Local secondary indexes
• Must be created when you create teh table
• Same partition key as the table
• Different sort key

• Global secondary inex. 
• Can create any time - at table creation or after
• Different partition key, 
• different sort key.

A query operation finds items in a table using only the primary key attribute

You provide the primary key name and a distinct value to search for.

• A scan operation examins every item in the table. 
• By default returns all data attributes

Query is more efficient

use the projection expression parameter.

In ascending order.

To reverse we use a scanIndexFoward parameter

set a smaller page size for fewer read operations

Isolate scan operations to specific tables and separate them from mission critical traffic.

Try parallel scans

Avoid using scan operations if you can.

• You pay for what you use. For unknown workloads and unpreditable application traffic. 
• You want a pay-per-use model.

reserved capacity is the other one

Provides in memory cachin for dynamo db tables

Improve response times for eventually consitent reads only

You point your API calls to the DAX clustrer instead of the table

If the item you query is in the cache then DAX will give it back. If not it will perform an eventually constent getitem operation to dynamo db table

no

Something that is unique like a phone number or an email address

You exceeded the maximum allowed provisioned throughput for a table or for one or more global secondary indexes

Time ordered sequence of item level modification in dynamodb tables

Any time there is a change, it is logged.

Data is stored for 24 hours only

Can be used an event source for lambda so you can create applications which take actions based on events in your dynamoDB table.

yes. It can get triggered by the dynamo db which is a log

So you can have a lambda function that checks for transactions.

retrying requests using progressive longer waits

requests too high

bactchgetitem

In capacity units

1 write capacity unit = 1 x 1kB write per second

• 1x Read capacity unit = 1 x strongly consistent read of 4kb per second
• or
• 2 x Eventually consistent reads of 4kB per second

• or 
• 2 eventually consistent reads of 4kb per second.

A sequence of values like unique dates or times

To find items in a able based on the primary key attribute

Use a projection expression parameter