DynamoDB

4 min read

Related: NoSQL
IS NOT A RELATIONAL DB
Flagship NoSQL, a key value DB. Serverless.
HA in 3 AZ, can be scaled massively (to global resilience)
Supports millions requests per second, trillion rows, 100TB storage
Single digit (0-9) latency
Cost saving: Standard or Infrequent Access Table Class
Backups points in time when enabled.

PK – Partition Key…
SK – Sort Key
high-cardinality = more unique

Capacity is set by units, can be Read (RCU) and Write (WCU). Pay for resources consumed.
Max Item, row.. 400KB

Consistency Model

Has leader mode and storage mode architecture.
This affects eventual and strong reads. Eventual are 50% price, but small chance of stale data from not updated node. Is application can tolerate non-current data, than consider eventual. If it’s something important, like stock level or medical research,

Modes

On-Demand - unknown, unpredictable, low admin
On-Demand - price per million R or W units.
Provisioned … uses RCU and WCU set on a per table basis
Every operation consumes at least 1 RCU/WCU(*)
1 RCU is 1 x 4KB read operation per second (*) (up to per RCU 4 kb)
1 WRU is 1 x 1KB write operation per second
Every table has RCU and WCU burst pool (300 second)
Try using it infrequently, by using it to much may lead to a throttle

WCU and RCU Calculation

Managing settings on DynamoDB provisioned capacity tables - Amazon DynamoDB

WCU

  1. If you need to store 10 ITEMS per second … 2.5K average size per ITEM
    1. Calculate WCU per item … ROUND UP (ITEM SIZE / 1 KB) (3)
      1. Multiply by average number per second (30)
        1. WCU required (30)

RCU

  1. If you need to retrieve 10 ITEMS per second .. 2 5K average size
    1. Calculate RCU per item… ROUND UP (ITEM SIZE / 4 KB) (1)
      1. Multiply by average read ops per second (10)
        1. = Strongly Consistent RCU Required (10)
          1. (50% of strongly consistent) = Eventually Consistent RCU Required 5)

Indexes

Local secondary indexes (LSI)

Only when strong consistency is required
Provides alt view for a table.
Both can have attributes, ALL, KEYS_ONLY and INCLUDE
Must be created when the table is created
Max 5 LSI

Global secondary indexes (GSI)

Use as default
Give alt PK and SK, own capacity
Can be created any time,
Default limit 20.
Have their own RCU and WCU allocation.
Eventually consistent.

Query

You can only query for PK and PK and SK.
It always better to pull more needed data in single operation for better RCU economy

Scan

More flexible (can query other columns) but consumes whole table
So better to use Indexes instead scan…

Backups

On demand

Point in time

Must be enabled
35d recovery window with 1 second granularity,

Event Driven

DynamoDB Streams

Stream is a list of item changes in the table, has 24 hour rolling window, powered by Data Streams.
Data recorded: INSERTS, UPDATES and DELETES
Has different view types:

  • Keys only – state of the item’s PK and SK.
  • New Image – state of the item after change.
  • Old Image – state of the item before change.
  • New and old image – state of the item before and after.

Triggers

Powered by Lambda. When data is changed in a stream, lambda is invoked to respond to that.

Global Tables

You create tables in multiple regions and then link them from a single table. At the same time all tables become R/W replicas, so a Global Table formed.

  • Last writer wins rules is used for conflict resolutions.
  • R/W can occur to any region.
  • Sub-second replication.
  • Strongly consistency reads only in the same region where write has happened.
  • Global application must tolerate eventual consistency.

Provides global High availability (HA) and Disaster Recovery (DR)

Accelerator (DAX)

  • There is DAX SDK for better integration of caching into an application.
    • Abstracts the overhead to a single entity combined DDB and DAX.
  • Has Query cache and Item cache.
  • Write-Through is supported, Data is written to DDB then to DAX.
  • Primary node supports write and Replicas which Read, HA is available (Failover and new Primary).
  • Scale UP and OUT (More resources or more caches)
  • Deployed in VPC.
    Eventual consistency here, again.

TTL

If enabled you can specify per item, date and time when data marked expired and then deleted. Value specified in a picked attribute “Column”
Stream of TTL deletions can be enabled.

Graph View

Backlinks