Overview of Sampling

How Sampling Works

Initial Data

When a dataset is first created, a background job begins to generate a sample using the first set of rows of the dataset. This initial data sample is usually very quick to generate, so that you can get to work right away on your transformations.

• The default sample is the initial sample.
• By default, each sample is 10 MB in size or the entire dataset if it's smaller.  
• If your source of data is a directory containing multiple files, the initial sample for the combined dataset is generated from the first set of rows in the first filename listed in the directory.

  • If the matching file is a multi-sheet Excel file, the sample is taken from the first sheet in the file.If you are wrangling a dataset with parameters, the initial sample loaded in the Transformer page is taken from the first matching dataset.

Generating samples

Additional samples can be generated from the context panel on the right side of the Transformer page. Sample jobs are independent job executions. When a sample job succeeds or fails, a notification is displayed for you.

As you develop your recipe, you might need to take new samples of the data. For example, you might need to focus on the mismatched or invalid values that appear in a single column. Through the Transformer page, you can specify the type of sample that you wish to create and initiate the job to create the sample. This sampling job occurs in the background.You can create a new sample at any time. When a sample is created, it is stored within your storage directory on the backend datastore.For more information on creating samples, see Samples Panel.

Sampling methods

Depending on the type of sample you select, it may be generated based on one of the following methods, in increasing order of time to create:

1. on a specified set of rows (firstrows)

2. on a quick scan across the dataset 

   Tip: Quick scan samples are executed in the Trifacta Photon running environment.

3. on a full scan of the entire dataset 

   Tip: Full scan samples are executed in the cluster running environment.

Sampling mechanics

When a non-initial sample is executed for a single dataset-recipe combination, the following steps occur:

1. All of the steps of the recipe are executed on the dataset on the backend cluster, up to the currently selected recipe step.
2. The generated sample is executed on the current state of the dataset.

When your flow contains multiple datasets and flows, all of the preceding steps leading up to the currently selected step of the recipe are executed, which can mean:

• The number of datasets that must be accessed increases.
• The number of recipe steps that must be executed on the backend increases.
• The time to process the sampling job increases.

Implications:

• When the sample is displayed in the Transformer page, all steps after the one from which it was executed are computed in the web browser. So, if you have a lengthy series of steps or complex operations after the step where you generated a sample, you can cause performance issues of the Transformer page, including the occasional browser crash. Try generating a new sample later in your flow for better performance.
• If you have added an expensive transformation step, such as a complex union or join, you can improve performance of the Transformer page by generating and using a new sample after the transformation step.

Important notes on sampling

• When sampling from compressed data, the data is uncompressed and then expanded. As a result, the sample size reflects the uncompressed data.
• Changes to preceding steps that alter the number of rows or columns in your dataset can invalidate the current sample, which means that the sample is no longer a valid representation of the state of the dataset in the recipe. In this case,  Designer Cloud Enterprise Edition automatically switches you back to the most recently collected sample that is currently valid. Details are below.

Parameterization of samples

Any parameters that are associated with your dataset can be applied to sampling:

• Parameters: Subsequent samples generated from the Transformer page are sampled across all datasets matched by parameter values.

• Variables: You can apply override values to the defaults for your dataset's variables at sample execution time. In this manner, you can draw your samples from specific sources files within your dataset with parameters.

Samples management

After you have created a sample, you cannot delete it through the application.

Choosing Samples

After you have collected multiple samples of multiple types on your dataset, you can choose the proper sample to use for your current task, based on:

1. How well each sample represents the underlying dataset. Does the current sample reflect the likely statistics and outliers of the entire dataset at scale? 
2. How well each sample supports your next recipe step. If you're developing steps for managing bad data or outliers, for example, you may need to choose a different sample. 

Limitations

• Some advanced sampling options are available only with execution across a scan of the full dataset.
• Undo/redo do not change the sample state, even if the sample becomes invalid. 

• When a new sample is generated, any Sort transformations that have been applied previously must be re-applied. Depending on the type of output, sort order may not be preserved.

• Samples taken from a dataset with parameters are limited to a maximum of 50 files when executed on the Trifacta Photon running environment. You can modify parameters as they apply to sampling jobs. See Samples Panel.

Sample Invalidation

With each step that is added or modified to your recipe, Designer Cloud Enterprise Edition checks to see if the current sample is valid. Samples are valid based on the state of your flow and recipe at the step when the sample was collected. If you add steps before the step where it was created, the currently active sample can be invalidated. For example, if you change the source of data, then the sample in the Transformer page no longer applies, and a new sample must be displayed.

• The Transformer page reverts to displaying the most recently collected sample that is currently valid. 

• You can generate a new sample of the same type through the Samples panel. If no sample is valid, you must generate a new sample before you can open the dataset.

• A sample that is invalidated is listed under the Unavailable tab. It cannot be selected for use. If subsequent steps make it valid again, it re-appears in the Available tab.

Best Practices

Sampling checkpointing

All steps between the step in your current sample and the currently displayed step must be computed in the browser.  As you build more complex recipes, it's a good idea to create samples at various steps in your recipe, particularly after you have executed a complex step. This type of sample checkpointing can improve overall performance.

For example, as soon as you load a new recipe, you should take a sample, which can speed up the process of loading. 

Sample Types

Designer Cloud Enterprise Edition currently supports the following sampling methods.

Initial Data Samples

These samples are collected automatically when you first load a new dataset into the Transformer page. These sample contain the first 10 MB of data from the first file or table in the dataset. 

First rows samples

This sample is taken from the first set of rows in the transformed dataset based on the current cursor location in the recipe. The first N rows in the dataset are collected based on the recipe steps up to the configured sample size.

• This sample may span multiple datasets and files, depending on how the recipe is constructed.
• The first rows sample is different from the initial sample, which is gathered without reference to any recipe steps.  

These samples are fast to generate. These samples may load faster in the application than samples of other types.

Random samples

Random selection of a subset of rows in the dataset. These samples are comparatively fast to generate.You can apply quick scan or full scan to determine the scope of the sample.

Filter-based samples

Find specific values in one or more columns. For the matching set of values, a random sample is generated.

You must define your filter in the Filter textbox.

Anomaly-based samples

Find mismatched or missing data or both in one or more columns.

You specify one or more columns and whether the anomaly is:

1. mismatched
2. missing
3. either of the above

Optionally, you can define an additional filter on other columns.

Stratified samples

Find all unique values within a column and create a sample that contains the unique values, up to the sample size limit. The distribution of the column values in the sample reflects the distribution of the column values in the dataset. Sampled values are sorted by frequency, relative to the specified column.

Optionally, you can apply a filter to this one.

Cluster-based samples

Cluster sampling collects contiguous rows in the dataset that correspond to a random selection from the unique values in a column. All rows corresponding to the selected unique values appear in the sample, up to the maximum sample size. This sampling is useful for time-series analysis and advanced aggregations.

Optionally, you can apply an advanced filter to the column.