here we explain, how data moves into a system in data architecture. it breaks down how companies collect, process, and store data from different sources. whether you're new to data management or looking to expand your knowledge, this chapter will help you understand how to manage data in different situations.

batch processing vs real-time processing
batch processing pros and cons
real-time processing pros and cons
choosing between batch and real-time processing
data governance

what is data ingestion?

data ingestion is how we bring data into a system so it can be stored and analyzed. it includes methods like etl (extract, transform, load) and elt (extract, load, transform), along with a newer method called reverse etl. data can be processed all at once (batch) or as it comes in (real-time), depending on what the business needs. good data management ensures the information is accurate and easy to access.

why data ingestion matters?

data ingestion is very important for helping businesses to manage and use their data well. it organizes the data, makes it easy to access, and prepares it for analysis, which helps in making better decisions and smooth operations. here's why it matters:

better decisions: it collects data from different sources, giving businesses a complete view to make smart choices.
saves time: it simplifies the process of gathering data, which reduces manual work and minimizes errors.
quick insights: it allows fast analysis of incoming data, helping businesses react quickly to changes.
grows with you: it can handle new data sources and larger amounts of data as businesses grow.
keeps data clean and safe: it ensures the data is accurate, consistent, and secure while following rules.

what is etl?

etl stands for extract, transform, load. it is a process where data is taken from various sources, modified and cleaned, and then stored in a destination, such as a data warehouse.

remember: etl = "early transformation leads": this means that the transformation of data occurs before it is loaded into the final destination.

advantages of etl

now, let's look at the benefits of etl

etl works well for smaller datasets with simpler changes.
it provides better control over data quality since data is cleaned before loading.
it improves data security by only loading the necessary, cleaned data.
it is usually more efficient for relational databases.

drawbacks of etl

here are some downsides of etl:

the transformation process can be slow and use a lot of resources, which may affect overall performance.
if there's an error, the data has to be re-extracted from the source, causing extra delays.
traditional etl tools might struggle with large amounts of data.
some etl tools may not support many different data types.

what is elt?

elt stands for extract, load, transform. in this process, data is first loaded into the destination system without any changes. after loading, the data is transformed. you can also remove unnecessary data during extraction.

remember: elt = "every load transforms" means that the data is transformed only after it has been loaded into the system.

advantages of elt

now, let's look at the benefits of elt.

good for data lakes and large amounts of unorganized data.
allows changes to the data after it's been loaded.
uses modern processing power for better performance.
makes changes faster using batch processing.
works with many types of data and tools.

disadvantages of elt

this section highlights the challenges of using elt.

requires more storage since raw data is kept.
can be slower if the transformation process is complex.
may lead to data quality issues if raw data is not managed well.
needs powerful systems to handle large data loads effectively.

etl vs elt

extract-transform-load (etl) was the main way to move data into a relational data warehouse. recently, extract-load-transform (elt) has become more popular, especially for data lakes.

both etl and elt have their strengths. etl is good for keeping data quality and security, especially with smaller datasets. elt is more flexible and works better for larger, unstructured data in data lakes.

choosing between etl and elt depends on your specific data needs. it's not simply one or the other; the goal is to find the best fit for your data processing.

reverse etl

reverse etl is about moving data from a data warehouse to other systems so that the data can be used for everyday tasks. traditionally, data in a data warehouse is used mainly for analysis and planning. now, many companies also use this data for operational analytics and daily operations.

for example, customer data can be cleaned in the data warehouse and then sent to systems like salesforce. this ensures that all teams have access to the same information, making it easier to identify customers who might be at risk of leaving.

in the data warehouse, companies create key metrics to better understand their customers, such as:

lifetime value: the total profit expected from a customer over time.
product qualified lead: a potential customer who has shown interest in a product.
propensity score: the chance a customer will buy.

these metrics help with decision-making. by using reverse etl, businesses can provide personalized experiences in real-time, enhancing customer satisfaction and improving overall outcomes.

batch processing vs real-time processing

in extract, transform, load(etl) and extract, load, transform(etl), there are two main options for when and how often to extract data: batch processing and real-time processing. here's a closer look at each.

batch processing

batch processing is a method for handling large amounts of data all at once. in this approach, similar transactions from a source system are grouped together, or "batched," and processed at regular intervals, such as daily or monthly. the system then runs a job to copy this entire batch to a destination, like a data lake or warehouse. this usually occurs during off-peak hours, which means times when the system has fewer users, making it easier to manage without slowing down.

for example, your electric bill is processed monthly, where the utility company collects your usage data and generates your bill at the end of the month.

real-time processing

real-time processing means working with data as it arrives, so you get immediate insights. when new information is available, it starts a process that quickly sends the data to where it needs to go.

for example, banks can instantly alert customers about suspicious transactions to help prevent fraud. similarly, traffic apps like waze use real-time data to update traffic conditions and suggest the best routes to take.

real-time processing updates the target system immediately, making sure reports and queries show the most up-to-date information. this helps businesses quickly spot issues that need immediate attention.

while traditional data warehouses mainly used batch processing, real-time processing is now more common, especially in data lakes capable of handling millions of events per second. each method has its own advantages and challenges in data warehousing.

batch processing pros and cons

batch processing handles large amounts of data at once, making it efficient but slower to access data. here are some pros and cons.

pros of batch processing

these points show why batch processing is a good option for handling large amounts of data efficiently without affecting the system too much.

efficiency: processes many items together, which is faster than handling them one by one.
scheduled tasks: runs during off-peak hours to avoid disrupting regular work.
lower risk: if something goes wrong, it can be easily retried.

cons of batch processing

these points highlight why batch processing may not be a good choice when quick data access is needed, resulting in delays in obtaining information.

delays in data availability: data can take time to be ready because it's processed in groups.
resource underutilization: can waste resources if not managed well.
not real-time: not suitable for applications that need immediate updates.

real-time processing pros and cons

real-time processing keeps data updated all the time for quick insights. it gives you timely information but requires more resources. here are the pros and cons.

pros of real-time processing

these benefits show why real-time processing is important for businesses needing quick and reliable information for decisions.

immediate insights: provides up-to-date information for fast decision-making.
continuous updates: great for systems needing constant data updates.
flexibility: easily adapts to changing business needs.

cons of real-time processing

these drawbacks explain why real-time processing can be challenging for businesses, as it demands more resources and can lead to higher costs.

higher resource demand: uses more system resources continuously.
increased failure risk: there's a higher chance of system failures, which can make fixing errors more complicated.
data consistency challenges: keeping data consistent can be tough with constant updates.
higher costs: more expensive due to ongoing operations.

choosing between batch and real-time processing

when choosing between batch and real-time processing, think about your data type, processing needs, and how much delay you can handle. batch processing is good for systems that can wait a bit, while real-time processing is best for situations needing immediate access.

data governance

data governance is about managing data in an organization. it sets rules for how data is collected, stored, secured, transformed, and reported. it makes sure the company follows laws and checks that the data is accurate and of good quality. this means ensuring data is cleaned and changed properly.

a good governance framework defines who is responsible for managing and using data. one way to do this is by creating a data governance center of excellence (coe). this coe helps develop policies and standards and clarifies roles and decision-making for data activities.

it's important to spend time creating a data governance framework and building your coe before starting a data warehouse project. many projects fail because they don't pay enough attention to data governance.

Save or share this article