Snowflake Database

This section explains how to connect to a Snowflake Database.

Overall Sequence

To connect to a remote Database using DirectQuery you need to complete these steps:

1. Configure your remote Database in Atoti Market Risk's expected Database format
2. Set the Atoti Market Risk configuration properties
3. Deploy Atoti Market Risk in Horizontal Distribution

Database Schema

Required DirectQuery Properties

DirectQuery only supports scalar mode for the Sensitivities Cube. In addition, the Summary Cubes can’t be activated with DirectQuery, so they have to be disabled:

mr:
  enable:
    data-model:
      # Set the sensitivity cube to scalar mode
      scalar-sensitivities: true
    cubes:
      # Disable all summary cubes
      var-summary: false
      pnl-summary: false
      sensi-summary: false

Snowflake Configuration

Properties for enabling and connecting to a remote database can be defined using either the YAML or standard property formats. These properties can be used in .yaml and .properties files, or on the command line.

# Snowflake Database Connection Parameters
directquery:
  enabled: true
  database:
    type: snowflake
    snowflake:
      connectionString: YOUR_JDBC_SNOWFLAKE_CONNECTION_STRING
      username: YOUR_USERNAME
      password: YOUR_PASSWORD
      warehouse: YOUR_WAREHOUSE
      feedingWarehouse: YOUR_FEEDING_WAREHOUSE
      database: YOUR_DATABASE
      schema: YOUR_SCHEMA
      role: YOUR_ROLE
      arrayAggWrapperFunctionName: YOUR_SPECIFIC_ARRAY_AGGREGATE_FUNCTION
      additionalOptions:
        anSFSessionProperty: YOUR_EXTRA_PROPERTY 

The above properties populate the SnowflakeSpringProperties java class.

connectionString

Full property name: directquery.database.snowflake.connectionString

This is the JDBC Driver connection string which points to the Snowflake Database. for details on how to find your JDBC Connection String, see the JDBC Driver Snowflake documentation.

warehouse

Full property name: directquery.database.snowflake.warehouse

The warehouse is where our queries and aggregation will take place. You can view available warehouses through the SHOW WAREHOUSES command.

feedingWarehouse

Full property name: directquery.database.snowflake.feedingWarehouse

Special wareHouse used during the feeding.

database

Full property name: directquery.database.snowflake.database

The Snowflake Database you will be connecting to.

schema

Full property name: directquery.database.snowflake.schema

The Schema within the specified Database to use. This Schema should match Atoti Market Risk’s expected Database Structure either by having the same Table structure or through the use of views.

role

Full property name: directquery.database.snowflake.role

The Role with which the specified Database is used. This Role allows request and use of the specified Warehouse on the specified Schema.

arrayAggWrapperFunctionName

Full property name: directquery.database.snowflake.arrayAggWrapperFunctionName

Optional name of the user-defined function that can be used to wrap the ARRAY_AGG operations and improve performance. This function must be defined in Snowflake and be accessible to the role running the queries.

additionalOptions

Any connection option defined in the Snowflake documentation represented as a key/value pair.

Deployment Options

We provide two main options to run with DirectQuery:

Horizontal Distribution

In Horizontal Distribution you have access to the in-memory tools such as What-If, data updates and Sign-Off for the data loaded in-memory as well as access to a large number of historical dates.

When running in Horizontal Distribution you need to run three Nodes:

• Query node
• In-memory data node
• DirectQuery data node

mr.data-load.initial-business-dates

For both Combined in-memory query node and DirectQuery data node, set the mr.data-load.initial-business-dates property.

Given that the data nodes are distributed by AsOfDate, no two data nodes can contain the same Partition - meaning that one AsOfDate cannot be present in another node. So to prevent any issues the directQueryDatesToIncludeFilter bean is used to set which dates to include in the DirectQuery data node.

JVM With Query Node And In-Memory Data Node

This JVM consists of two distributed nodes; a query node and an in-memory data node.

Start the JVM by specifying the following parameters to the Atoti Market Risk application, either in a .properties file or through command-line arguments (add -D before each property).

See Activating and de-activating cubes for details about the enabling/disabling of cubes.

# Disable cubes
mr.enable.cubes.<cube>: false

# Set MR to run this JVM with a query node and in-memory data node
spring.profiles.active=dist-data-node,dist-query-node

JVM With DirectQuery Data Node

Once the JVM with the query and in-memory data node is running, you can start a second JVM with the DirectQuery data node with the following configuration properties either in a .properties file or through command-line arguments (add -D before each property).

# Set MR to run as a data node and to connect to the second JVM
spring.profiles.active=dist-data-node

# Use a different port from the second JVM
server.port=10011

# Use a different Content Server for this JVM
mr.application.content-service.db.url=jdbc:h2:mem:content_service;DB_CLOSE_DELAY=-1
mr.application.content-service.db.hibernate.hbm2ddl.auto=create

# Disable cubes
mr.enable.cubes.<cube>: false

# Enable and configure the Snowflake DirectQuery Database
directquery.enabled=true
directquery.database.type=snowflake
directquery.database.snowflake.username=database_username
directquery.database.snowflake.password=database_password
directquery.database.snowflake.connectionString=connection_string
directquery.database.snowflake.warehouse=snowflake_warehouse
directquery.database.snowflake.database=snowflake_database
directquery.database.snowflake.schema=snowflake_database_schema

Single JVM

You can run a single JVM consisting of only DirectQuery.

By running in a single JVM with DirectQuery only, you can now see the DirectQuery data in the VaR/ES cube.

Properties

To run the Single JVM node you only need to configure the Snowflake Properties and disable cubes and run the application as normal. There are no distributed nodes to configure when running in a single JVM.

Here is an example of the properties to use:

See Activating and de-activating cubes for details about the enabling/disabling of cubes.

# Disable cubes
mr.enable.cubes.<cube>: false

# Enable and configure the Snowflake DirectQuery Database
directquery.enabled=true
directquery.database.type=snowflake
directquery.database.snowflake.username=database_username
directquery.database.snowflake.password=database_password
directquery.database.snowflake.connectionString=connection_string
directquery.database.snowflake.warehouse=snowflake_warehouse
directquery.database.snowflake.database=snowflake_database
directquery.database.snowflake.schema=snowflake_database_schema

Single JVM Limitations

By running DirectQuery with a single JVM you will be running purely on DirectQuery. This will come at the compromise of query performance of Trade level queries while also not having access to in-memory only tools such as WhatIf and SignOff.

Reference Snowflake Database Schema

Atoti Market Risk ships with the SQL and data needed to initialize a Snowflake database that can be used for testing. The SQL scrip and reference data are located within the directory: mr-directquery/src/test/resources/databases/snowflake/.