608B71FC-006A-4934-A643-7D9BA9340450Blog

Maintain Serverless

blog__author-img
Michał Czapracki
CEO at Scramjet, Data Streaming Expert.
23F2E8CD-3026-46A5-86CC-D13114F7176E425AE875-B1A1-4EA1-8529-075D08DA0BB1

10 May 2021

The speed at which data is generated, consumed, processed, and analysed by companies is increasing at a rapid pace. As a result, most of the technologies we use today rely on vast amounts of data and need on-demand data processing and analysis at an almost real-time pace.

It’s no wonder that cloud computing has become so popular because it gives companies instant access to their data and allows them to store vast amounts of data without the cost implication of having an on-site infrastructure to manage the data.

But recently, a new challenge reared its head. This challenge is to migrate data from legacy systems to the cloud. Here, companies have a variety of options that they can consider. In this post, we’ll look at some of these options in more detail and illustrate why stream processing may be the best option.

What is Data Migration

Generally, data migration is the transfer of existing historical data, like from legacy systems to a new storage system. Generally, nowadays, data migration refers to when companies migrate their data from on-premises infrastructure to the cloud or even between different cloud environments.

This migration, commonly referred to as cloud migration, is a migration of all a company’s data. This includes their data storage, databases, applications, datacenter, and business process migration. Although this process sounds reasonably simple as it entails the moving of data from one platform to another, it poses several challenges.

Different Approaches

Traditionally you basically had two approaches to data migration. The first one is moving all your data from your source environment to your target environment in one operation. Here, you typically also transfer the data within a relatively short time window.

It’s perfectly understandable why you would want to do this. Because your systems will be down and users will be unable to use it as long as the data is transferred to the target infrastructure, you’ll want to complete it as quickly as possible. That’s why it’s most often done during a weekend or holiday when the disruption will be minimized.

So, using this approach allows you to complete the migration as quickly as possible, and you eliminate the hassle of working with two different, often disparate, systems simultaneously during the transfer.

However, you should keep in mind that even smaller companies nowadays accumulate vast amounts of data which would take time to transfer. As a result, it’s often not the best choice for critical applications that should be available 24/7.

Your next approach would be to incrementally transfer your data from the source system to the target system. This process, known as iterative or trickle data migration, allows you to break the data transfer up into smaller pieces. This will enable you to minimize unexpected failures in the data transfer and eliminate any downtime resulting from it.

Unfortunately, this means you’ll use old systems and new systems in parallel while the data is being transferred. Here, you face a challenge because you’ll have to ensure that your data is synchronized between the two systems in real-time. This real-time synchronization is challenging because any changes in the source system must also update the target system.

Also, you’ll need more resources to keep two systems running. Besides this, this iterative strategy takes more time and is more expensive than transferring everything at once.

Ultimately, there must be a better solution.

A Better Way

And there is. With stream processing, you will be able to move your data to the target infrastructure without interrupting the correct working of your current infrastructure. This will, for instance, be when you move your data to new modern systems from legacy architecture.

In simple terms, you end up having a hybrid deployment where part of your applications and data sources will run in the cloud while others remain on premises for the time being until your data is transferred. Now, you might think that this sounds a lot like trickle data migration.

The big difference is that, with stream processing, you’ll have real-time synchronization between your cloud systems and your on-site systems. In this way, you’ll centralize all your data from your current systems and your cloud deployment into one centralized place. It gives you the added benefit of getting real-time insights from your data because it’s analysed in-stream.

In practice, this solution requires you to build a data pipeline where data will be extracted and loaded into the appropriate solution every time an event occurs. This, in effect, offers you the benefit of real-time data availability while eliminating the drawbacks of traditional methods of data migration.

Typically, you’ll have a choice between a variety of tools to implement event streaming to facilitate your data migration. These include:

  • Apache Flume. Apache Flume is a reliable system for collecting, aggregating, and moving large amounts of data. Its simple architecture is based on streaming data flows.
  • Apache Spark. Apache Spark is a distributed data processing system that’s used for big data processing workloads. It’s fast and a good option for large-scale data processing.
  • Apache Storm. Apache Storm is a distributed and real-time big data-processing system. It’s designed to process large amounts of data and offers fault tolerance and horizontal scaling. It’s probably the streaming framework with the highest ingestion rates.
  • Apache Kafka. Apache Kafka is an open-source distributed event streaming platform. It’s used by companies for building high-performance data pipelines, streaming analytics, data integration, and data availability in mission-critical applications.

Conclusion

Up to now, migrating your data to the cloud required you to choose between doing complete migrations or trickle data migrations. Both of these methods have pros and cons, which make them suitable for specific situations. But neither offered the best solution.

Using stream processing for your migration project now poses a far better solution. It gives you the ability to migrate your data over time while having a central repository of data that provides real-time synchronization between your data sources.

If you want to know more about how stream processing can help you with your data migration or our other services, visit our website and contact us for more information.

Project co-financed by the European Union from the European Regional Development Fund under the Knowledge Education Development Program. The project is carried out as a part of the competition of the National for Research and Development: Szybka Ścieżka.