What is the cardinality of a dimension?

Cardinality refers to the number of distinct values available in a column. For example, the cardinality of states in the US is 50 whereas the cardinality of the number of customers in the US could be in millions. The lower the cardinality, the more duplicated elements in a column. Thus, a column with the lowest possible cardinality would have the same value for every row.

For Kyvos Free, you can have a maximum of 50 Million Cardinality for each dimension.

How can I control the number of dimensions?

Reduce the number of dimensions by using any or all of the following options:

1. Combining dimensions: At the Register File level, whenever possible, combine dimensions with a fewer number of attributes (1-3) into one dimension to reduce the number of dimensions. For example, some dimensions can be merged at the register file level to reduce the number of dimensions at the semantic modellevel. 

2. Using multiple hierarchies: Consider using multiple hierarchies if you need two types of time data such as year-month-day and year-quarter-month-day for different purposes. Or if you need two types of location data such as division-region-district-location and state-county-city-location.

What are the supported dimension types and how to configure them?

Kyvos supports Time, Location, and Regular dimension types. The dimension type can be configured through the Dimension Properties panel as shown below.

What are the supported hierarchy types and how to create them?

• Predefined Time Hierarchy: By selecting the hierarchy from Kyvos' pre-defined hierarchy list (allowed only on hierarchies for Time Type Dimension).

  Open

  

• Custom Time Hierarchy: By selecting the Date Data Type for each level of the Time hierarchy individually from the Level Properties panel (allowed only on hierarchies of Time Type Dimension)
  NOTE: If Date Data Type is not configured on a level of the custom time hierarchy, all the levels below that level will not perform/behave as Date Data Type even if configured.

• Custom Calendar: If the Quarter, Month, etc. for your data model are not the same as defined in the Gregorian calendar  (allowed only on Custom Time Hierarchy of Time Type Dimension).
  
  

  Open

  

• Alternate Hierarchy: By selecting the Has Alternate Path checkbox on the Hierarchy Properties panel.

  Open

  

• Parent Child Hierarchy: By selecting the Has Parent Child Relation checkbox on the Hierarchy Properties panel.
  
  

  Open

  

• Unbalanced and Ragged Hierarchies: You can hide redundant levels using the Hide Member property from the Level properties panel.
  
  

  Open

  

How to create a Time dimension in Kyvos?

1. At the time of adding dimensions, choose the dimension Type as Time from the Dimension Properties pane on the right.

2. Choose the hierarchy and field by right-clicking the dimension and making a selection.

3. If this is a time hierarchy, choose Year-Month-Date, Year-Quarter-Month, Year-Quarter-Week-Date, or Year-Quarter-Month. 
   If there is not a pre-defined time hierarchy shown, set the field data type to a Date type such as year-quarter-month.

Kyvos allows you to use time hierarchies in two ways:

• Predefined: These are pre-created hierarchies in the system. You can select and add them to the semantic modelfrom the hierarchies list from Year-Month-Date, Year-Quarter-Month, Year-Quarter-Week-Date, or Year-Quarter-Month.

• Custom: In this type, you can define different date values for different columns. For example, year=2021 and month=09. In this case, you can drag individual columns and set up the field data type to each field such as year-quarter-month-day.

You can use the following to indicate time values:

• Year: YYYY -2020

• Quarter: Q1, Quarter1, 1 , 01

• Month: Jan, January 01, 1

• Week:  w1, 01, 1, Week1

• Day: 01, 1

If your data contains values like " 2020-Sept " for a month, it is not a true month value and you will need to convert the data to a supported format. 

How can I process newer data without a full build?

You can use the Incremental jobto update only the latest changes to an existing semantic model , without a full build. This option is faster than a full build.
For example, use it to update the semantic modelwith daily changes. You can also replace already processed data using the replace partition option. 
NOTE: You can run an incremental job only on a successfully built semantic model .

You can also define a schedule for an incremental build, such that the job is triggered automatically as per the defined schedule.

Open

Read more: Semantic Model Process Types

I want to load incremental data in a built semantic model, what should I know?

• Please refer some terminologies and details in the below table. The semantic model entities whose source tables have Incremental Data will be processed when an Incremental job is scheduled on the semantic model .

• There is no difference in the “Incremental data” definition for the dimension table and fact table.

• Complete processing will be skipped during the incremental job for the semantic modelentities whose source table does not have Incremental Data. 

NOTE: If none of the source tables are found to have “Incremental Data”, the incremental job will be marked as skipped. Thus, the semantic model will continue to remain in the same state as before the incremental build.

I want to load incremental data in existing partitions/add new partitions, what should I know?

Kyvos allows the user to configure what must be done with the “Incremental Data". While scheduling the Incremental job job, you can choose from either of the Replace Partition options as detailed below.

• None: Doesn't replace the partition. Adds incremental data in already processed partitions.

• Auto: Identifies the partition automatically, and replaces data of partitions (existing data will be dropped from the semantic model and incremental/new data will be added).

NOTE: The Incremental Data will simply be added into the semantic model in the following scenarios:

• If partitioning strategy is configured on the semantic model : New partitions from the source table.

• If Partitioning strategy is NOT configured on the semantic model : Complete Incremental Data from the source table.

How can I connect to Kyvos semantic model s from Tableau and Excel?

Kyvos integrates seamlessly with Tableau (SQL and MDX connectivity) and Excel (MDX connectivity).

To connect Kyvos semantic models:

Tableau on Kyvos

Excel on Kyvos

What is partition strategy, how and why to set it up before full build?

Partitions are a flexible means of managing semantic models. They are internal structures used by Kyvos to manage and drop or update data in a semantic model. The base partition must be created on the Date type value.

For example, a semantic model that contains sales information can contain a partition of data for each past year and partitions for each quarter of the current year.

When new information is added, only the current quarter's partition needs to be processed. At the end of the year, the four quarters of accumulated partition data can be merged into a single partition for the year and a new partition created for the first quarter of the new year. 

To create partitions, the semantic model must have a date type dimension defined.

Kyvos provides options for getting recommendations based on a data profile job or queries executed on the semantic model to further optimize the partitioning strategy. 

Note that partitions are not visible to business users of the semantic model. However, users with semantic model modify roles assigned to them can configure, add, or drop partitions.

Read more:

• Create partitions

• Replace partitions

• Sliding window

• Drop partition from semantic model

What are the prerequisites to create a partition and/or sub-partition?

You can create semantic model partitions/sub-partitions from the Refine tab on the semantic model designer screen.
Following are the prerequisites:

• General

• Can only be created levels on a dimension hierarchy.

• Can only be created on levels of the first hierarchy of a dimension (NOTE: A dimension may have multiple hierarchies).

• Cannot be created on columns of Slowly Changing Type1 dimension.

• Base Partition

• Can only be created on a level of Time type hierarchy.

• Can be created on any level of Predefined Time hierarchy (Date Data Type is implicitly created in Predefined Time hierarchy).

• Cannot be created on levels on which the Date Data Type is not configured.

• Cannot be created on a level of a hierarchy that is marked as a Custom Calendar.

• Sub Partition

• Can be created on any level of hierarchy excluding the ones which are mentioned in the General section.

What are the recommendations to create semantic model partitions and/or sub-partitions?

It is recommended to create a partition/sub-partition on columns that are likely to be used as filters while browsing.

Example: If it is likely that Month, Region will be used as a filter in the queries. These columns should be configured in semantic model partition / sub-partition. This helps in the pruning aggregations and hence in achieving better query response time.

The total number of combinations of partitions and sub-partitions must not be huge. It is recommended to keep the combinations in thousands.

How does the partition size for the base partition works?

Kyvos assigns a partition number to each fact record based on static logic. These partition numbers are calculated as the time difference of the record with the base date “01-Jan-1970". Below are some examples for various Date Data Types

Year: Partition Number = Number of Years between Year of record and 1-Jan-1970/Partition Size

Example (Base Partition on Year, Partition Size = 2)

Day: Partition Number = Number of Days between Date of record and 1-Jan-1970/Partition Size

Example (Base Partition on Day, Partition Size = 15)

Thus, the partition numbers are a simple and static calculation depending upon:

• Date Data Type of base partition Level

• Partition Size configured on semantic model

Kyvos considers all the records of a partition number (Date Range) as an atomic unit and all the jobs like Drop Partition, Replace Partition etc. work accordingly.

What is the difference between Expert mode vs Smart mode aggregation strategy?

• Expert (user-driven) is the configuration-driven mode that allows you to tune your semantic model performance by setting up some key materialization values and the precompute threshold level. From the summary screen, you can see how many dimension or hierarchy materializations are selected. From there you can view details, make changes, or learn more by clicking the i symbol. You can also see the degree of materialization and the precompute threshold. For all of these items, you can see whether it is the default, inherited, or modified. Use this option if you want to have more control over the process.

• Smart (system-driven) is query-based and allows you to filter or analyze historical queries and work with data profile information. You can set the materialization level and specify whether to use fresh aggregates or to potentially add additional aggregates in addition to any existing ones. Use this option if you want to get quick insights with a low learning curve. You will be prompted to regenerate aggregates after changes to the semantic model design.

How to create a Slowly Changing Type1 dimension? What are the prerequisites?

On the Dimension Properties pane, select the Is Slowly Changing Type 1 Dimension checkbox.

Prerequisites:At least one of the following rules must be fulfilled:

• The column that relates the dimension table and fact table (join key column) is the lowest level in the dimension hierarchy.
  OR

• The column that relates the dimension table and fact table (join key column) is one of the attributes in the dimension.
  OR

• The values of the lowest level of hierarchy/or any attribute possess a 1-1 relation with the join key column values.

NOTE: Kyvos compares the join key column value to identify if an existing record is updated as part of SCD1.

Login Failure

Unable to connect to database- Check if the Postgres is up on Kyvos Manager

• If not then start it from the KM.

Could not connect to BI server- Check if the BI Server is up and running from Kyvos Manager. If not, then start it.

If the BI server is not coming up, then check any port is being used by another process. End that process or contact Kyvos Support to configure other ports for the BI server.

Could not retrieve the users list

• Check if the BI server and Postgres are running and accepting connections on their ports.

• If not start the services accordingly.

• If LDAP is enabled then check whether Admins are able to login or not and check whether the LDAP user is listed in the user list (Kyvos Menu > Setup > Users).

Semantic model Job is Failed

Dependency on External Services

1. S3

1. 503 SlowDown Error.

2. EMR File Inconsistency Issues.

3. Access Rights issues. (403 Access denied)

4. Service Down. (Region failure/ Availability Zone

2. EMR

1. Error while scaling the EMR cluster. (Out of memory for example).

2. Incorrect configuration (For example glue not enabled properly).

3. Connectivity issue between EMR and Kyvos. (Different subnet).

4. Incorrect JDBC URL Specified for Hive.

5. Incorrect History Server URL Given for Spark. (behavior).

6. EMR is terminated while running the job.

7. EMR is terminated simply which is synched with Kyvos. How product behaves in that case.

8. Configurations are changed in EMR. (In Running EMR for ex scaling is changed.)

9. If EMR has no rights to access S3. (Monitor Behavior).

10. EMR doesn’t have rights on glue tables. (Rights are revoked, also on RF).

11. Cluster is resized while running.

12. Single EMR is shared across multiple Kyvos distributions.

13. Spot Instances are not available, or decommissioned. What is the behavior in that case ?

14. Spot Instances With Map Reduce.

15. In Non Glue Based Deployment, if EMR is terminated, how product behaves in that case.

3. Dynamo DB

1. Capacity Of Dynamo DB is reached.

2. Same Dynamo DB is shared across multiple kyvos deployment.

3. Dynamo DB Metadata Table is deleted.

4. EC2

1. Accidently BI Server/ KM/ QE node is shut down by anyone.

2. Attached disks are removed from the node, for any reason.

3. If attached volumes are not sufficient to job the semantic model, what is the product behavior in that case ?

4. If the user attached new volumes because the earlier job was failed due to lack of disk space, is he able to job the semantic model again successfully.

5. GLUE

1. BI Server doesn’t have Rights on glue tables.

2. Multiple kyvos deployments using the same Glue databases.

3. Table is modified in source after the registered file is created.

6. SECURITY

• IAM Role Doesn’t have sufficient privileges on:

• GLUE

• S3

• EMR

• IAM Role is removed during the semantic model job process.

• Snowflake

                                          i.    Access rights on Database and Tables on which semantic model is built.

                                         ii.    Credits are consumed.

                                        iii.    Access to create a stage on the database which is mentioned on the connection.

7. Miscellaneous

• AWS Lambda time out.

1.1.1   Semantic model job is failing with File Not Found Exception

• Check on the cluster for the presence of the file which is shown in the error snippet.

• Maybe there was an ETL job running that deleted the data file temporarily.

• Check on the cluster when did this file changed.

1.1.2   Semantic model job failed with out of memory error/ Java heap

• Increase the memory for the mapper and reducer for the semantic model (if you are using MapReduce to job the semantic model).

• Property’s value to change:

map.memory.mb and mapreduce.reduce.memory.mb- In the case of Spark increase the value for the property spark.executor.memory.

Semantic model job is slow

Semantic model job stuck at indexing job

• Check the available resources on the Resource Manager.

• Check the semantic model job status on Resource Manager, It can be in Accepted, New, or Submitted Application.

• Wait for the other jobs to complete or decrease the value for mapper and reducer memory in semantic model properties.

Register File Issue

Unable to preview getting File_Name Execution failed- Try changing the execution engine for the file either MapReduce or Spark.

• To do so, add

build.execution.engine property in the Register File property.

• By default, the SPARK SQL context is used to preview the files so try changing to MapReduce.

• Restart the Spark SQL Context by API call or by restarting BI Server.

Connection error: Browser is not able to connect to server or the server is down.- This comes when you are registering the file from HCatalog.

• Check the JDBC URL at the connection level if it is correct.

• Check if the table is accessible from a beeline or hive shell.

Worksheet/Dashboard Preview Fail

Could not serve query as SQL operation asynchronous executor is stopped / Failed to execute query through semantic model layout

• The SQL context running on the BI server node is stooped.

• Restart the SQL context through REST API or restarting the BI server will automatically start the SQL context.

Could not execute the query as no query engine is active in the allocated or default segment- Check the state of the query engine on Activity Monitor.

• If it is in an initializing state then wait until they come up.

• If they are down then start them from Kyvos Manager and wait till they come up.

Please job the semantic model before browsing.- This happens when the semantic model is not built.

• Process the semantic model or enable raw data querying on that semantic model.

Failed to execute SQL query through semantic model layout, Connection URL not provided for external engine HIVE- Check the value of hiveserver2.jdbc.url property in Hadoop Connection. (Navigate to Kyvos Menu > Setup > Connections > HadoopConnection).

• If the correct JDBC URL is set or not.

• If yes try to access Hive through a beeline with the same JDBC URL.

Kyvos Services not Starting

Web Portal is not coming up after starting- Check the ports used by the web portal are being used by other processes.

• Default ports used by Web Portal are 6605, 8081, 8009, and 8005.

• Make sure the Kyvos Manager Agent is running on the Web Portal’s node.

BI Server is not coming up after starting- Check the ports used by BI Server such as 6602, 45460, 45450

• Make sure the Kyvos Manager Agent is running on the Web Portal’s node.

Query Engines are not coming up (YARN Mode)- Check the Resource Manager if the application for Query Engines is I which state.

• Check the available resource on the Resource Manager

• Check the available resources for the YARN Queue

Postgres showing down status- Check the port 45421 (used by Postgres) whether it is being used by another process.

• Check the Kyvos Manager agent is up and running.

Kyvos Manager showing click to upgrade, what shall I do?- This happens when you start the Kyvos manager from the old version’s directory.

• Make sure to start the correct Kyvos Manager from the correct directory.

Data Integrity Concerns

1.1.3   Measure values aren’t matching with expected results/other BI software

• Check the filter conditions on register files

• Check the SQL query on register files

• Check the number of records in the semantic model job summary

• Ensure that the source table had valid records when the semantic model was scheduled for build

• No data/ETL load was under process when the semantic model job was running

• In the Kyvos register files, validate that the column data types are correct as per the source data

• Check the summary function of measure(s) in the semantic model design

• Are your records falling beyond the configured semantic model sliding window strategy?

• Check the join condition/column of your dimension and fact tables in semantic model DRD

• Check if the discrepancy is because aggregates are shown as “unknown”

NOTE: NULL or blank does not contribute to aggregations in Kyvos, is your BI tool doing otherwise?

1.1.4   Dimension records are missing in viewer filter/browsing results

• In the raw data of dimension table, check the values in all the columns of dimension for the missing record, are all column values valid as per column data type?

• Ensure that the missing record(s) were present in the source table when the semantic model was scheduled for build.

• Are you missing record(s) which has NULL/blank value(s) in any column?

• If it is because of NULL/blank values, then set appropriate configurations as mentioned in the valid values table.

1.1.5   How does Kyvos handle special values like NULL, invalid value, and blanks?

Refer to the following table that shows how various special values are treated with respect to the data type of source column and the semantic model entity which is created using the source column.

Also, see  Refine > Advanced Properties > Data Cleansing on the semantic model designer screen.