Return Of Investment In Multi-Hybrid Environment Within Digital Data Warehouse

Published on Sep 20, 2022 06:31 PM IST

This article will summarise the findings of a discussion of various deployment options (the use of microservices, cloud-based implementation, and managed services), as well as consider various complementary data integration technologies, in this context

Binayaka Mishra, Project Manager, Competency & Data Analytics, Tech Mahindra, Electronics City, Phase II, Bangalore -560100, Karnataka, India
Binayaka Mishra, Project Manager, Competency & Data Analytics, Tech Mahindra, Electronics City, Phase II, Bangalore -560100, Karnataka, India
ByHT Brand Studio

Data integration market trends and how users perceive them are the primary focus of this article. Consideration of the most important technical aspects of Informatica data integration and how they fit into the overall picture of data management are examined in this report. This article will summarise the findings of a discussion of various deployment options (the use of microservices, cloud-based implementation, and managed services), as well as consider various complementary data integration technologies, in this context. Each of these technologies has its own set of risks and benefits, which are usually based on my research.

TCO (Total Cost of Ownership) and ROI (Return on Investment) are important to discuss before diving into the findings, as the latter is heavily emphasised in this research. However, it's more difficult to calculate TCO compared to return on investment. Net present value (NPV) is one method of calculating potential benefits, which include cost savings, risk reduction, increased sales (via improved data quality), and so on (net present value). Whenever necessary, I will go over the appropriate measures in the following sections. When statistics are available, we provide them; otherwise, we simply list the names of the people involved.

As an additional note, I'm glad to steer clear of broad-brush discussion of total cost of ownership (TCO), because aside from vendor comparisons that fluctuate over time and are subject to discounting, our main finding has consistently been that the TCO of hand coding is vastly underestimated and that using hand coding is actually more expensive in the long run than using a tool-based approach. However, we don't see the need to go over the details again. Nonetheless, it's a useful conclusion.

Why Choose a Tool?

It was more important for me to find out why people were using data integration tools like Informatica rather than trying to figure out the total cost of ownership. Figure 1 shows the final results. There are a lot of connections between a lot of these responses. For developers, time spent on maintenance is reduced as a result of the ease with which data flows can be maintained, and as a result of this, TCO is reduced. The word "reuse" isn't included in this section. Reusing code is possible, but it necessitates the use of additional tools (such as version control). It's better to go with a tool-based system if you're going to include that.

Figure 1 – When it comes to data integration, do you prefer to use a vendor product or do you prefer to code by hand? (check all that apply).
Figure 1 – When it comes to data integration, do you prefer to use a vendor product or do you prefer to code by hand? (check all that apply).

Reuse

If they used a data integration tool like Informatica instead of hand coding, I wanted to know why and how satisfied they were. Maintainability received an 80 percent satisfaction rating, change management received a 77 percent satisfaction rating, and reuse received an 81 percent satisfaction rating, with only performance and scalability receiving a higher satisfaction rating. Figure 2for reusing is particularly relevant due to the wide variety of scenarios to which it can be applied. Loading data from Oracle into an Informatica warehouse, for example, could be reused in the case of loading data from Db2 into Informatica as well. When migrating from Teradata to Snowflake, the process is even more complicated. It's not just projects that can benefit from reusing, though; departments and, ultimately, your entire company can. The multiplier effects occur as the reuse rate rises. How advantageous is recycling? Obviously, a product like Informatica, which uses digital technology in its core products, is more useful than one that can only be used once. For the former, it would be worth paying five or even eight times as much as the latter. In terms of data integration, things are a little more complicated. Nonetheless, It's not just that you'll have to estimate how much reuse you'll need over the course of three or five years (or whatever the relevant period is) when a data integration tool like Informatica can theoretically be (re)used an infinite number of times. A further consideration is how much of any given dataflow can be reused for support of other projects, as this will not be 100 percent. Finally, how much time is saved by reusing code and how much money is saved as a result of that. As developers become more comfortable with the tool they're using, and as the relevant vendor incorporates more automation into its product, this figure is expected to rise. Measuring post-mortem reuse value is possible, but that doesn't help when making a first-time tool implementation decision. How many projects you can complete in a given time period, how much developer time is saved on average per project, and how much money you save in terms of staff costs are some of the things you'll need to calculate. To get a wide range of values, it's best to evaluate worst-case, average, and best-case scenarios. When it comes to digital technology cloud data migration and data integration, we can't function without Informatica.

Figure 2 – If your product is able to support the following data integration features, please rank it accordingly: You can select from one to five (5 being most satisfied) 
Figure 2 – If your product is able to support the following data integration features, please rank it accordingly: You can select from one to five (5 being most satisfied) 

Self-service Non-technical usage lessens IT overload

Technical users such as developers have traditionally been the target audience for Informatica data integration tools. However, this is beginning to shift. It is becoming increasingly common for software vendors to offer self-service and collaborative capabilities, which are often powered by automation and machine learning. Informatica has both of these trends implemented in it. Domain experts, business analysts, and other non-technical staff may be able to use these capabilities to define data integration processes, easing the burden on IT departments that are already overburdened. As can be seen in Figure 2, these capabilities get high marks (70 percent or higher) from users, but they aren't as well liked as other aspects of their tooling. A major reason for this is that these requirements are still so new. Note that by definition, the use of hand coding excludes both self-service and teamwork. Because it eases the burden on IT, implementing self-service capabilities has a return on investment component. This is an important consideration, given how overstretched most IT departments are. How long does it take for an IT developer and an analyst to complete the same project? How many such projects have been completed? And how much are the salaries of those involved? This, however, fails to consider two additional aspects. An analyst may have to wait for a new integration dataflow if IT is developing it, as well as what the opportunity costs are. Second, freeing up IT resources for other purposes has the potential to be beneficial. Using Informatica, we can speed up the data collection process for actuaries of any skill level. There's no need for SQL coding, and there's no need for immediate access to subject matter experts now that self-service is an option. A true data scientist used to be required to gain any insights from the data. With the help of digital technology and Informatica, you can make analytics more accessible than ever before.

Collaboration leads to increased performance efficiency

Data integration tools like Informatica can enable collaboration, but so can complementary technologies like Microsoft SharePoint and Google Docs (data quality, governance, and catalogues, which we will discuss in greater detail later in this report). Thought should be given to its own merits. By promoting collaborative working, "companies are five times more likely to be high-performing companies (compared to those that do not)," says the Institute for Corporate Productivity. Many aspects of business life are affected by collaborative working, not just data asset collaboration. Most importantly, however, data governance tools and data catalogues with features specifically designed to facilitate collaboration can help with collaboration because they are software tools. There are a few features of relevant products that can help facilitate collaboration in the current discussion. When it comes to the right data and enterprise data sources, there is a problem Many colleagues work with data sets that are either complementary or overlapping. Indicating whether or not certain datasets contain particularly helpful information can be indicated by "liking" or "rating" these sources. This can help in the search for the "right" information. Informatica's multicloud product IISC, for example, recommends specific datasets (perhaps based on built-in machine learning) that can be used in this context.

The collaboration of business users with technical experts is another facet of the concept. As a rule, "personas" are employed to accomplish this, with different user communities using different "personas" to access the data they need for their respective roles within the organisation. Using this approach, different types of users, such as business users, data stewards, data scientists, and technical experts can all have their own customised views of a single set of data. An integrated environment with a common metadata foundation, which may be helped by the use of AI and machine learning with Informatica, is typically how these multiple persona collaborative capabilities are achieved. It will also be possible to attach notes and comments, which can be shared both within and across personas, with the help of such a platform. Data integration products and platforms used by a majority of companies were only moderately effective at facilitating collaboration. Hand coding (no collaboration) will, of course, have lowered this number, but even so, support for collaboration was rated as being less satisfactory than reuse or maintainable (both rated 4 out of 5, compared to 3.8 for collaboration).

Advanced Features

In addition, I sought feedback from users on the more advanced features they were hoping to see in a data integration platform and whether or not their vendor supported them. There are some interesting findings in the results shown in Figure 3. However, there is a distinct difference between vendors specialising in ELT and those specialising in flexible data integration environments (extract, land, and transform in the target environment). For example, Informatica offers mass ingestion and change data capture, but it also offers TEL and/or the flexibility associated with push down optimisation, as well as support for pub-sub and B2B integration (both of which are useful for blockchain). Other characteristics, such as support for data preparation, data science operationalization, and data catalogue integration, are also shown and will be discussed in the future in relation to complementary technologies. Many vendors like Informatica have fully implemented machine learning in order to make transformation recommendations, which is a boon for cloud and hybrid workloads; the ability to easily build data pipelines is critical for the emerging discipline. Native connectivity is becoming more and more important as data volumes continue to rise. Having said that, suppliers should provide both generic connectivity (ODBC/JDBC and APIs) and software development kits because there are literally thousands of potential end points for data integration. According to a study by Informatica, the number of connectors in the Informatica product line has been artificially inflating. There are ten distinct operations defined against an Oracle database, and each of these could be counted as a connector. The most important consideration is the number of data sources for which the vendor provides native capabilities. TCO and ROI can be reduced by implementing all of the advanced features listed in Figure 3.

Figure 3 – Are the following essential and advanced features supported by your data integration product?
Figure 3 – Are the following essential and advanced features supported by your data integration product?

Platforms

Additionally, we wanted to know what our users thought about the importance of complementary technologies, such as data quality, data governance, and the availability of a database catalogue. Figure4 shows the results in terms of how important they are to people. The fact is that user organisations didn't just say they wanted these features as a nice-to-have; they actually invested in them, as illustrated in Figure 5. However, because Informatica has a newer technology, data catalogues are not seen as being as important as data quality and governance, which is understandable. How much time had they spent integrating data quality, governance, and catalogue capabilities with their cloud-based tooling for integrating data ingestion? 6.7 months was the average time spent. As a result, we asked the same questions to all users, regardless of whether they had purchased a platform with pre-integrated modules or not. While it is sufficient in and of itself, this means the actual work to integrate disparate products must have taken much longer than 6.7 months. Some respondents took over two years to integrate all of their tools, as we've seen in the past. There's no denying that this method is time- and money-intensive without the use of digital cloud technology from Informatica. There are many advantages to using a platform-based approach that does not rely on third-party integration.

Figure 4 – For each of the above, how important are these technologies to your company's success? You can select from one to five (5 being most important
Figure 4 – For each of the above, how important are these technologies to your company's success? You can select from one to five (5 being most important
Figure 5 – With your chosen data integration product, which of the following features have you implemented?
Figure 5 – With your chosen data integration product, which of the following features have you implemented?

Deployment Options

Lastly, in my survey, I inquired about Informatica cloud deployments. Most people were surprised to learn that they used cloud-based data in their integration processes at least partially. There is a significant increase in the number of people currently using Informatica cloud-based data management tools. This isn't surprising, given how recently these tools have been made available to the public. Despite this, the adoption of Informatica cloud-based solutions is increasing and is expected to do so in the future. Cloud-based deployments, microservices-based architectures, and managed services must all be considered in this context. Briefly:

1. One of the advantages of microservice-based architectures is that they allow for rapid feature adoption. Traditionally, major software upgrades have resulted in significant downtime and high administrative costs. For example, cloud data warehouses (CDWs) can be adapted to new features more quickly thanks to these tools. A cloud implementation of Informatica is required for microservices-based architecture.

2. There are several advantages to implementing in the Informatica cloud, including elastic scaling, serverless computing, and separating storage and computation. 2. However, this does not imply that all cloud-based service providers have these features available. Our knowledge of data management vendors who offer elastic scaling but not serverless computing is extensive. Storage and compute can be separated in on-premises environments, but this is less common. With Informatica multi-cloud solution you can easily implement high availability, resilience, and zero-downtime capability in the clouds. This is not to say that these solutions cannot be implemented in more traditional settings, but doing so will require more effort on the part of the manager.

Data Quality

Data quality is all about making sure your data is accurate. It must be reliable and valuable enough to drive both short-term business decisions and long-term digital transformation initiatives. erroneous data carries a risk to your brand's reputation. Reputational risks will be addressed in due course. Because of the importance of data integration, data quality is often considered a necessary corollary. More than 70% of users have done both at the same time, according to our research. It's easy to come up with a list of rationales for this. Figure 6 is an example of an infographic that cites multiple sources. There is a lot of bad data out there, and it costs a lot of money to collect and analyse. It should be self-evident that you want to work with up-to-date data rather than outdated data in addition to high-quality data. Data decays immediately, as shown in Figure 6, which illustrates the issue. In addition, this is a long-term endeavour. When it comes to data quality, it's reasonable to assume that it's constantly deteriorating. Because it's always there, you don't need to know when it begins to decay. It is therefore necessary to view data quality remediation as an ongoing effort rather than a one-time event. There's only one explanation for this. According to the graph in Figure 8, over the course of three months, many business contact details change. Depending on the type of data, estimates of annual data decay range from 18% to 40%. The fact that many data quality vendors provide ROI calculators should not be overlooked. It'snever been easier to maintain high-quality data with Informatica IDQ.

Figure 6 Illustrates the financial toll that poor data quality takes..
Figure 6 Illustrates the financial toll that poor data quality takes..
Figure 7 – After just three months, the data begins to degrade
Figure 7 – After just three months, the data begins to degrade

Data Governance

Data governance deals with the implementation of business rules and policies that affect your data. Technical data quality rules are not included, but business rules based on corporate policies, such as "credit limit may not exceed x," are included. However, the line between the two capabilities is often blurred, particularly when they are supported by the same platform. In a broader sense, data governance distinguishes between corporate policies and regulatory policies. There are a number of procedures that would fall under this category, such as onboarding new clients or releasing new products. For example, ensuring compliance with industry-specific regulations like MiFID II and HIPAA, as well as more general ones like the GDPR and CCPA, is part of governance. There is a strong link between data governance and the provision of a business glossary, which allows you to link specific data governance projects to potential cost savings and improved revenue streams and profitability. Data monetisation allows you to prioritise governance projects in addition to its other benefits. Survey responses to key benefits of data governance are summarised in Figure 9. Interestingly, data democratisation comes in at number five. It's possible that since then, this factor has risen in importance. In this section, as in others throughout the report, the same considerations apply. For example, if there isn't a clear inventory of available data, users may spend 30-40 percent of their time searching for it, and they may spend another 20-30 percent of their time cleaning it up. These annoyances can be alleviated with effective data governance." Data governance and monetization have never been easier, faster, cheaper, or more widespread than with Informatica's cloud-based AXON product.

Figure 8 Depicts the primary motivations and benefits of data governance
Figure 8 Depicts the primary motivations and benefits of data governance

Data Catalogues

Our Data Marketplace initiative is driven by critical business needs for data visibility, ease of access, improved governance, and data democratisation.. For the Marketplace to meet these requirements, Informatica Products EDC and Axon are the information pillars. Data governance, metadata management and data quality can all be achieved through Informatica with minimal effort. In addition to the quality of your data, having data fit for purpose means that the relevant data for any given business decision should be as complete as possible. It also has to be timely: you don't want to wait a week to collect all of the relevant data if you have to make an urgent decision.

In other words, you need to be able to quickly and easily locate all of that data, right? However, the foundation of any knowledge graph will be built on a data catalogue rather than a knowledge graph itself. Data can be classified as sensitive or product-oriented, for example, and then accessed by data preparation tools, allowing the data to be wrangled into a format suitable for analytics and machine learning. Finally, data can be provided to data preparation tools, allowing the data to be wrangled into a suitable format for analytics and machine learning. For a more detailed look at how a data catalogue can benefit your organisation, see Figure 9. Figure 9 illustrates how data catalogues can help hasten the move to cloud computing environments. With a data catalogue, you'll have an easier time finding and prioritising the information that needs to be moved from on-premises legacy systems to modern cloud-based storage and data lakes. There are some vendors who have gone a step further and automated the cloud migration of data through integration of cataloguing and quality control, allowing IT to move relevant data to the cloud automatically when it is discovered in the catalogue. One more compelling reason to avoid hand-coding altogether, in my humble opinion. Finally, end-to-end data governance cannot be implemented without data catalogues. An organization's entire data estate can only be properly managed by mapping business terms and governance policies to the data inventoried in a data catalogue. This also makes it possible for analysts looking for data to perform simple business keyword searches. As a result, this is not possible with hand-coding. There are a few metrics for the use of data catalogues that have been calculated by independent authorities. In one instance, the ROI calculated by Forrester Research was 364 percent. Only seven companies were included in this study, which is insufficient for statistical significance even if it is representative. Figure 10 shows the results of my more in-depth investigation, which shows how satisfied people were with using or not using a data catalogue. Informatica Data Catalogue products with multi-cloud platform integration help its customers and vendors achieve all of these.

Figure 9 Illustrates the capabilities of the Informatica data catalogue in this diagram.
Figure 9 Illustrates the capabilities of the Informatica data catalogue in this diagram.
Figure 10 Using or not using an Informatica data catalogue: how satisfied are users?
Figure 10 Using or not using an Informatica data catalogue: how satisfied are users?
Figure 11 Additional evaluations of whether or not an Informatica data catalogue is useful 
Figure 11 Additional evaluations of whether or not an Informatica data catalogue is useful 

Data Warehouse Digitization for better ROI

Data warehousing's role in modern data management practises is being called into question. Though some have declared the data warehouse to be extinct, many organisations still run at least one (most have two to five) and expect to do so for the foreseeable future. Data warehousing remains an important part of data management, but ageing data warehouses must be modernised in order to fit gracefully into modern data management practises and deliver long-term value. To fit into modern analytics ecosystems, legacy data warehouses must evolve both architecturally and technologically. This article discusses how to keep data warehouses valuable by restructuring their architecture, migrating to the cloud, and integrating them into a comprehensive and cohesive data management strategy.

Contrary to popular belief, data warehousing is not obsolete. According to recent polling, more than 60% of businesses currently operate between 2 and 5 data warehouses. Fewer than ten percent have only one data warehouse or none at all. Almost one-third of poll respondents work in a company with six or more data warehouses. Although the vision of previous generations of BI and data warehousing has not been realized—one data warehouse serving as a single version of the truth—it is clear that data warehousing continues to provide value to these organisations. Data warehousing is not extinct, but it is in trouble. It's still alive, but not entirely well. Legacy data warehousing is being challenged by Bigdata, data lakes, NoSQL, data science, self-service analytics, and the demand for speed and agility. Traditional data warehousing, which is based on data management practises from the 1990s, simply cannot meet the demands of rapidly increasing data volumes, processing workloads, and data analysis use cases. To meet the realities of modern data management and overcome the challenges of scalability, elasticity, data variety, data latency, adaptability, data silos, and data science compatibility, data warehousing must evolve and adapt. Existing data warehouses are still in use because they are required. On a daily basis, business processes and information workers rely on warehouse data and information. Many people, if not the majority, continue to require well-integrated, systematically cleansed, easy-to-access relational data with a large body of time-variant history. They want to meet routine information needs with data that has been prepared and published specifically for those needs. Data warehouse modernization, including architectural rethinking and purposeful use of cloud technologies, is critical to the future of data warehousing.

The Evolution of Digital Data Warehousing

Modernization of data warehouses (DWM) is a natural next step in the evolution of data management for modern analytics, AI, and machine learning (ML) projects. Warehousing was developed to address the challenges of non-integrated operational systems and the resulting data disparity. The data management architecture was linear, with reporting and business intelligence as the most common use cases.

Figure 12 Legacy Data Warehousing Architecture
Figure 12 Legacy Data Warehousing Architecture

Throughout the period when data sources were primarily internal structured data and relational databases met data storage and management requirements, this architecture served. It is easily adaptable to data marts, multidimensional data, and OLAP analysis. However, the strength and stability of data warehouses gradually deteriorated. Due to mergers, acquisitions, and other changes, companies now have multiple data warehouses—the next generation of data silos. Then came the age of big data, which upended long-standing data management practises. Legacy data warehouses are ill-equipped to handle unstructured data complexities, process massive data volumes, adopt NoSQL databases, leverage Hadoop processing power, and take advantage of cloud technologies' scalability and elasticity.

Welcome to the data lake! Data lakes quickly became the next-generation data management concept, optimised for big data, embracing NoSQL, powered by cloud technologies, and leveraging the power of open-source technologies like Apache Spark, Kafka, and Hadoop. Utopia in data management? No, not quite. Data lakes did not take the place of data warehouses. We now have a new generation of data silos, which work in tandem with multiple data warehouses.

Figure 13 Typical Data Lake Architecture
Figure 13 Typical Data Lake Architecture

Adopting open source technologies presented new challenges for organisations with data lakes. The number of these technologies, the complexities of configuration, and the constant change make software and infrastructure management difficult. These difficulties are exacerbated by a scarcity of skilled workers outside of a few geographic areas, such as Silicon Valley. In a highly competitive labour market, open source talent is scarce and expensive, continuous retraining is required, and employee retention is problematic.

Ten Must-Have of Data Warehouse Digitization for better ROI

Data lakes are not the pinnacle of data management evolution. They do not make data warehousing obsolete, and they are merely a first step toward a future of enterprise data hubs. Modernization of data warehouses (DWM) is a natural next step in the evolution of data management. It is necessary to ensure the long-term value of warehouse data and the continued return on investment in data warehousing. Now is the time to reconsider data warehouse architecture and data warehouse deployments! With each technological advancement, the gap between legacy data warehousing and modern data management practises widens. Consider DWM to be a critical component of data strategy. It will play a significant role in shaping your data management future. Prepare now for a data warehousing future in which you will be able to:

1. Make full use of cloud technologies:

i) Scalability—Horizontal scaling or scale-out quickly adapts to changing workloads.

ii) Elasticity—The ability to increase and decrease capacity as workload fluctuates is especially important in data warehousing, where data volumes, processing workload, and concurrent user count can experience extreme peaks and valleys.

iii) Managed infrastructure - entails offloading the burden of data centre management to the services provider and eliminating management workload for floor space, rack space, power, heating and cooling, and hardware and software management.

iv) Cost savings—Reducing the cost of running an on-premises data centre and shifting much of the cost of data management from capital expenditure to operating expense.

v) Processing speed—Cloud computing allows for much faster processing. The ability to add processing capacity horizontally and expand and contract (elasticity) as needed accounts for a large portion of the gain.

vi) Deployment speed—Although data warehouse enhancements and modifications appear to be limitless, projects are frequently delayed due to infrastructure upgrades to expand data capacity, increase processing capacity, or support additional development and test environments. Cloud elasticity overcomes these obstacles, resulting in project delays and faster deployment.

vii) Disaster recovery—Because the complexities of warehousing make disaster recovery planning especially difficult, business critical data warehouses are frequently overlooked in disaster recovery planning. Virtualization in a cloud environment allows for a more straightforward approach.

viii) Security and governance—Of the numerous aspects of data security, some are entirely the responsibility of cloud service providers. When migrating a data warehouse to the cloud, for example, server security becomes a provider responsibility. Other aspects of security and governance become shared responsibilities, where understanding provider features and capabilities, as well as describing responsibilities through SLAs, is critical.

2. Support hybrid cloud and multi-cloud environments. As the deployment landscape grows larger, seamless interoperability across multiple technology environments becomes increasingly important. Figure 14 depicts the complex deployment landscape that depicts the realities of typical deployments today. This landscape includes four distinct cloud environments as well as several systems hosted on-premises. A modern data ecosystem must support cloud-to-on-premises interoperability, such as connecting the Snowflake data warehouse to the SAP business warehouse or legacy applications to the Google Cloud analytics data warehouse. The ecosystem must also support cloud-to-cloud interoperability, such as connecting Workday applications to the Azure data lake or collaborating with the Snowflake data warehouse and the analytics data warehouse at the same time. Eventually, multiple cloud environments as well as on-premises systems will be standard. Interoperability is essential because the systems must all communicate with one another without isolating any of the data that they store and manage.

Figure 14 Complex Deployment Landscape
Figure 14 Complex Deployment Landscape

3. Support all data types, including structured, semi-structured, and unstructured data. Big data has replaced the once-simple world of structured data stored in relational tables. Modern data management still works with structured data, such as customer records and sales transactions, which are meticulously organised as rows and columns. Structured data in the Hadoop ecosystem occasionally migrates from relational tables to cloud-optimized and Hadoop-friendly formats like Avro and Parquet. Both Avro and Parquet are optimised for Hadoop and are row-based storage formats. Semi-structured data is less rigorously organised and is typically stored using semantic tagging file formats such as XML and JSON. Semi-structured data is frequently collected and stored as machine generated data from sensors, mobile devices, and mobile apps. Semi-structured data formats are also widely used for data sharing via electronic data interchange (EDI) services. Unstructured data is the polar opposite of structured data, lacking both the organisation and semantic context of semi-structured data. Unstructured data is frequently textual, but it can also be images, photos, or videos. Unstructured data includes freeform text customer comments accompanying a warranty service request and photos associated with insurance claims. Legacy data warehouses are constrained by structured data's relational constraints. All types of data must be supported by modern data warehouses.

4. Support for all data latencies, including batch, real-time, and streaming. Batch extract-transform-load (ETL) processing, which is inherently high-latency, is used to populate legacy data warehouses. Daily loads, for example, result in a data warehouse with one day's worth of data. Today's real-time business processes frequently necessitate the use of real-time data. A modern data warehouse must support data at all speeds, continuing to use batch processing when necessary, acquiring data in real time using changed data capture (CDC), and parsing data streams to capture only the events of interest. Only then can the data warehouse support a wide range of common use cases, such as time-series analysis and trend reporting, dashboards for real-time monitoring, and real-time alerts of business events and conditions discovered through data.

Figure 15 Architecture Of Data Latencies
Figure 15 Architecture Of Data Latencies

5. Provide assistance to all data users, including data scientists, data analysts, data engineers, and report writers. Individuals in various roles have varying data requirements. Data scientists frequently prefer raw data at the atomic level of detail and without any cleansing or other transformations. Data analysts, particularly line-of-business analysts who use self-service tools, benefit from integrated and cleansed data because it requires less data preparation work from them. Report writers prefer to work with integrated, cleansed, dimensioned, and aggregated data. Data engineers work with all of these types of data. The modern data warehouse provides all users with data that ranges from raw to highly transformed, as well as lineage and traceability throughout.

6. Encourage data users to work together. Data-driven and collaborative are key characteristics of modern business culture. People who work with data must collaborate to share knowledge, analysis, and data; they should never work alone. Data scientists can create models on which others can build. Data engineers can create reusable data preparation processes. Data analysts can publish their findings for others to discover, use, or adapt, saving time and money on redundant analysis. Every data user can share their knowledge of data as well as their experiences working with specific datasets. Collaboration and sharing increase efficiency, improve the quality of analysis and reporting, and raise data literacy throughout the organisation. Collaboration requires strong connections and a high level of interoperability between the data warehouse and the data catalogue.

Figure 16 Architecture Of Collaboration Among Users
Figure 16 Architecture Of Collaboration Among Users

7. Assist with data quality, data security, and regulatory compliance. Risk management and mitigation are essential functions of data management and modern data warehousing. When low-quality data is used for analysis and reporting, data quality suffers. Poor data undermines trust in the data, increases the possibility of misinformation, and reduces the quality of decision making. Data profiling and algorithmic detection of data flaws and conflicts help to reduce the risk of poor data quality. Data risk management also includes the protection of personally identifiable information (PII) and privacy-sensitive data. A modern data warehouse must be capable of detecting, locating, and classifying sensitive data while also protecting it from unauthorised access. Aside from privacy and PII, the data warehouse must mitigate the risk of non-compliance across the regulatory spectrum, including GDPR and a slew of industry-specific regulations. Compliance risk reduction is especially important in highly regulated industries like finance, healthcare, pharmaceuticals, and energy.

8. Support a wide range of big data processing engines. There are numerous technology options for processing big data, and the options are evolving as open source innovation continues. Many organisations use multiple processors, in part to optimise the platform for specific data and applications, and in part because it is impractical to go back and convert everything built in the past when adopting a new technology. A modern data warehouse must support multiple processing engines while also adapting to new technologies. Data warehouses of today should be compatible with the processing engines that many consider to be the top five big data processing frameworks, including Hadoop, Spark, Flink, Storm, and Samza. Because each engine is optimised for specific applications, limiting the data warehouse to a single processing engine limits its adaptability. The trends are:

i) Hadoop with MapReduce was the first and is still widely used big data processing engine. When data can be processed in batches and processing can be distributed across a cluster, it works well.

ii) Spark, a more recent and adaptable processing framework than MapReduce, has been widely adopted as a replacement for MapReduce. Spark, which lacks its own distributed storage layer, can operate within the Hadoop ecosystem and utilise HDFS.

iii) Flink is a batch-capable stream processing engine that is optimised for streaming and real-time data processing.

iv) Storm combines a stream processor with a real-time distributed compute engine, making it ideal for real-time analytics and machine learning.

v) Samza is a distributed stream processing engine based on Kafka messaging and YARN cluster resource management.

Figure 17 Top 5 BigData Processing Engines
Figure 17 Top 5 BigData Processing Engines

9. Provide assistance to the entire data management supply chain. Data management processes are much more complex and comprehensive in the age of big data and data lakes than they were when our legacy data warehouses were designed and built. Processes for data ingestion, data stream processing, data integration, data enrichment, data preparation, definition and cataloguing, mapping of data relationships, data protection, and data delivery are all part of a modern data management supply chain.

Figure 18 Architecture Of Informatica Data Management Supply Chain
Figure 18 Architecture Of Informatica Data Management Supply Chain

10. Implement AI/ML throughout the data management supply chain. It is now impractical to try to manage data without the assistance of artificial intelligence (AI) and machine learning (ML). With the volume, variety, and speed of data, manual data discovery, tagging, matching, mapping, and description is simply not possible. There are numerous opportunities for algorithms and agents to assist with data management throughout the supply chain:

i) The ability to detect and adapt to schema changes during data ingestion reduces disruption to data pipelines.

ii) Algorithmic event parsing optimises and accelerates data stream processing.

iii) When integrating data without shared keys, it is critical to use smart data transformations and intelligent blending. Blending internal customer data with external data that does not use your customer ID number, for example, can be especially difficult without AI recommendations for matching criteria.

iv) Data enrichment may use AI to facilitate data cleansing functions as well as the discovery of data enrichment opportunities such as automated geocoding of data with physical address or location information.

v) Automation and recommendations for preparation operations benefit data preparation. Masking sensitive data, for example, is a repeatable preparation step that can be automated by AI and continuously refined by ML.

vi) The advantages for data definition, data governance, and data cataloguing are enormous. Using algorithms to crawl data sources, infer semantics, discover and tag sensitive data, derive metadata, and aid in data curation is an important part of data management because it automates work that is too large in scope and volume to be done manually.

vii) The use of algorithms to discover relationships between datasets and intelligent mapping of those relationships improves data integration, increases data value, aids in data analysis, and simplifies data preparation and blending.

viii) Discovering, tagging, and protecting PII, privacy sensitive data, compliance sensitive data, and security sensitive data is an important part of managing and remediating data risks.

ix) When working with a large number of data sources, use cases, and data users, data delivery becomes a complex stage of the supply chain. AI/ML is useful for creating smart data pipelines and orchestrating their execution.

Data warehousing and data lakes must collaborate as complementary components of a unified data management architecture. For data management architecture, there is no one-size-fits-all solution. Because each data warehouse is unique, each modernization plan is also unique. However, there are several architectural patterns for modernization that can aid in the transition from data warehouse and data lake silos to cohesion and compatibility between data lakes and data warehouses. To develop a modernization plan and drive next-generation analytics and AI/ML projects, use these patterns individually, in combination, or as a mix-and-match for multiple warehouses.

Digital Data Warehousing Outside the Data Lake For Better ROI

This variant treats the data lake and warehouse as distinct data stores with no overlap. All incoming data lands in the data lake, and warehouse ETL pulls data directly from the lake. The landing zone of the data lake serves as warehouse data staging. Sharing a common landing zone for all incoming data reduces redundancy, preserves raw data, and allows for fully traceable data lineage.

Figure 19 Data Warehousing Architecture Outside Data Lake
Figure 19 Data Warehousing Architecture Outside Data Lake

Digital Data Warehousing Inside the Data Lake For Better ROI

The warehouse is positioned as part of the data lake in this framework. The warehouse can obtain data from both a raw data zone (data staging) and a refined data zone where some cleansing and transformation work has already been completed. When a data warehouse is expected to have a long lifespan and a large number of users who need to work with raw data, refined data, integrated and historical warehouse data, positioning it as a subset of the data lake may be especially desirable.

Figure 20 Data Warehousing Architecture Inside Data Lake
Figure 20 Data Warehousing Architecture Inside Data Lake

Digital Data Warehousing In Front of the Data Lake For Better ROI

One or more data warehouses continue to operate independently in this variant, but they also serve as sources for data ingested into the data lake. Because the data warehouses remain unchanged, the modernization advantage is limited. Pushing warehouse data to the data lake creates a duplicate of the data, but it also eliminates the silo effect caused by multiple data warehouses and the data existing separately and in isolation. Although the benefits are limited, the complexity and effort required are minimal, and there is no discernible impact on data warehouse users. This could be the first practical step in a multi-phase modernization process.

Figure 21 Data Warehousing Architecture In Front Of Data Lake
Figure 21 Data Warehousing Architecture In Front Of Data Lake

Digital Data Warehouse and Data Lake Inside/Outside Hybrid For Better ROI

With multiple data warehouses, it may be feasible to implement a hybrid model in which warehouses with heavy analytics usage and overlap with other data lake contents are placed inside the data lake, while those with a small user base and primarily used for inquiry and reporting remain outside the data lake.

Figure 22 Data Warehouse Inside & Outside Data Lake
Figure 22 Data Warehouse Inside & Outside Data Lake

Cloud Platforms for Data Warehouse Digitization For Better ROI

Cloud data warehousing has grown in popularity as businesses face increasing data volumes, higher service-level expectations, and the need to integrate structured warehouse data with unstructured data in a data lake. The trend toward SaaS for enterprise applications makes cloud data warehousing an appealing option. Many legacy data warehouse challenges are addressed by cloud data warehousing, which provides a targeted and direct response to the need for scalability, elasticity, managed infrastructure, cost savings, processing speed, faster deployments, ease of disaster recovery, and improved security and governance capabilities. Less obvious but equally important advantages include ready access to non-relational and unstructured data technologies, improved adaptability and agility through instant infrastructure, and reduced reliance on in-house data centres. Migrating an existing data warehouse to a cloud platform provides significant benefits and is a practical step toward modernization, but it is neither quick nor simple. Data warehouse migration is a difficult multi-step process that involves moving many different warehousing components.

Figure 23 Migrating Data Warehouse To Cloud
Figure 23 Migrating Data Warehouse To Cloud

Amazon Web Services (AWS), Microsoft Azure, and Google Cloud are the three most popular cloud platforms for data warehouse migration. Each works well with Informatica CLAIRE—intelligent, Informatica's metadata-driven data management engine—as demonstrated in the reference architectures below. CLAIRE provides assistance from data acquisition to data consumption, with intermediate steps for data ingestion, preparation, cataloguing, security, governance, and access.

Figure 24 Reference Data Warehouse Architecture Of Informatica With AWS Cloud
Figure 24 Reference Data Warehouse Architecture Of Informatica With AWS Cloud
Figure 25 Reference Data Warehouse Architecture Of Informatica With Microsoft Azure Cloud
Figure 25 Reference Data Warehouse Architecture Of Informatica With Microsoft Azure Cloud
Figure 26 Reference Data Warehouse Architecture Of Informatica With Google Cloud
Figure 26 Reference Data Warehouse Architecture Of Informatica With Google Cloud

Data Warehouse Digitization With Analytics For ROI

We've already established that data warehousing is still a critical component of modern data management. Data warehouses continue to provide value by meeting people's information needs. Many people rely on them and do not want them to be replaced by a data lake. Data lakes are ideal for analytics and big data needs. They provide a rich data source for data scientists and self-service data consumers. However, not all data and information workers want to be self-serve customers. Self-service analytics does not replace data warehousing; rather, it supplements and extends it. Data lakes and data warehouses collaborate to provide data in a variety of formats, including raw data, integrated data, and aggregated data. They must be designed and managed in such a way that each adds value to the other, and they cannot exist as separate data silos. Published data (warehousing) and ad hoc data (self-service) collaborate to meet a wide range of information requirements.

Companies maintain data warehouses because they are required. On a daily basis, business processes and information workers rely on warehouse data and information. Many people, if not the majority, continue to require well-integrated, systematically cleansed, easy-to-access relational data with a large body of time-variant history. They want to meet routine information needs with data that has been prepared and published specifically for those needs. In a data-driven business, there are numerous use cases, and there is no one-size-fits-all data organisation that is optimised for all users and uses. Data warehouses and data lakes should collaborate to provide a diverse set of data for all use cases. The pages that follow illustrated several common data use cases in which the coexistence of data warehouses and data lakes is critical to meeting data and information needs. You most likely have several of these use cases in your organisation, as well as others not shown here. Each use case is accompanied by its own reference architecture. Use these as a starting point for adjusting to your specific data and use case characteristics.

Data streams are among the most difficult big data sources to manage. Machines, sensors, and other IoT-connected devices send data in real time. You will almost certainly have streaming data if you use RFID tagging, GPS enabled devices, or robotics. As it arrives, it must be captured and/or analysed. Connecting to the stream yields streaming data. Upon ingestion, individual events are parsed from the data stream and sometimes filtered to include only events of interest. In a data lake, event data is typically collected as raw data. When a measurement exceeds a threshold or otherwise indicates the need for immediate attention, events can be analysed in real time to send alerts. Event data may occasionally flow directly to dashboards for real-time monitoring. Deeper analysis and reporting usually necessitate the addition of context to the data. Machine and sensor data is typically sparse, consisting of only the machine/sensor id, a measurement value, and a date/time stamp. Adding context, such as machine or sensor attributes, is dependent on persistent reference data, which is frequently found in a data warehouse. To support time-series and trend analysis, the data warehouse may also collect time-variant history from a data stream.

Figure 27 Data Warehousing Architecture With Streaming Analytics
Figure 27 Data Warehousing Architecture With Streaming Analytics

Self-service analytics is a common application. Every organisation that uses Tableau, Qlik, Power BI, or similar tools has self-service data analysts who are constantly challenged to find and understand data. Many of the same challenges confront data scientists when it comes to locating the right data for their modelling efforts. It is commonly stated that these analysts and scientists spend 80% of their time gathering and preparing data and only 20% analysing and discovering insights. They frequently struggle with deciding where to look for data—lake, warehouse, or elsewhere. When data is catalogued and prepared at the time of ingestion, much of the struggle is removed, and the 80/20 rule is reversed, with 80 percent of time spent on analysis and insights. Much of data cataloguing and data preparation in a smart data ecosystem is automated, with AI and machine learning discovering data characteristics, inferring semantics, tagging sensitive data, and making data searchable. This also makes collaboration easier by allowing users to share data knowledge, data preparation operations, and even data analysis.

Figure 28 Data Warehousing Architecture With Self Service Analytics
Figure 28 Data Warehousing Architecture With Self Service Analytics

Finding data becomes more difficult when it exists in a complex environment that includes multiple cloud platforms as well as on-premises data, which is a reality for most organisations today. As most of us work in a multi-cloud environment of SaaS applications, cloud hosted ERP systems, and cloud data lakes, perhaps we should stop saying "in the cloud" and instead say "in the clouds." However, we also have on-premises data sources and, in many cases, on-premises data warehouses. Working with data spread across multiple platforms presents unique challenges for data discovery, access, and blending. With all of the benefits described for self-service analytics—finding data, understanding data, preparing data, and collaborating when working with data—the data catalogue plays a critical role here. In a complex multi-cloud/hybrid data ecosystem, data prep combined with a data catalogue enables users to find and enrich data regardless of deployment platform or location.

Figure 29 Data Warehousing Architecture With Hybrid Environment & Analytics
Figure 29 Data Warehousing Architecture With Hybrid Environment & Analytics

When the data environment is large and complex, a data integration hub may be the best solution. When you have a plethora of on-premises and cloud data sources, as well as a plethora of users and use cases, data integration driven by individual sources or uses is impractical. A cloud-based data integration hub gathers data in a single location to harmonise it without creating redundant copies. A robust data hub includes data storage, harmonisation, indexing, processing, governance, metadata, search, and exploration capabilities. It is worth noting that the data lake and data warehouse continue to exist in this reference architecture, albeit in new roles as data sources for the integration hub.

Figure 30 Data Warehousing Architecture With BigData Integration Hub
Figure 30 Data Warehousing Architecture With BigData Integration Hub

Advanced analytics (predictive and prescriptive), AI, and machine learning (ML) are at the forefront of modern data use cases. Algorithm-based data applications ranging from decision automation to robotics and autonomous devices provide significant opportunities for business digital transformation. They may, however, pose a high risk and have the potential for negative consequences. Data quality is an important consideration for these applications. Consider the risks associated with poor data quality in diagnostic and prescriptive decision automation in healthcare. Similarly, a machine learning application in social sciences that uses low-quality data would learn incorrectly, produce biased algorithms, and potentially disrupt people's lives. Data quality is essential for prediction, prescription, automation, AI, and machine learning. Data quality assurance and cleansing must be performed as part of the data preparation process.

Figure 31 Data Warehousing Architecture With Advanced Analytics & AI/ML
Figure 31 Data Warehousing Architecture With Advanced Analytics & AI/ML

Final Thoughts

Digitization benefits from Informatica's data integration and data quality, as well as IISC's cloud services, EDC's BDM, Axon, and MDM's governance and compliance, all of which help to lower costs, reduce risks, and boost productivity. That's the essence of return on investment (ROI). Consider the platforms I've described if you're serious about making best use of your data and understand the need for digital transformation. Predictive analytics models are included in the scope of Informatica Governance. Artificial Intelligence (AI) also falls under the purview of the government. Ensure that any AI-derived results are reliable. Informatica enables the democratisation of data for the purpose of conducting analytics. Informatica has the following AI-driven & digital deliverables:

1. Enable the market for data

2. Add sub-categories to your search results.

3. Set up the delivery methods for the provisioning of data.

4. Create consumable collections of data assets (including AI models)

5. Provide additional information in the data collection—for instance, the delivery context.

6. Improve and iterate on the system in response to user requests.

7. Digital frameworks such as Informatica's are helpful for data ingestion and data storage.

8. In addition, Informatica technology with digital led frameworks can be run on multiple cloud environments, such as Google, AWS, and Microsoft Azure.

9. Automated data integration and multi-cloud data management are supported by Informatica MDM.

10. Informatica's Intelligent Data Platform includes the CLAIRE (Clairvoyance and AI) technology, which is the first and most advanced metadata-driven AI technology in the industry. Using machine learning, CLAIRE provides intelligence to Informatica's entire product and solution portfolio by analysing technical, business, operational, and usage metadata from both the cloud and the enterprise. CLAIRE can help data developers by partially or fully automating many tasks, while business users can find and prepare the data they need from anywhere in the company, thanks to the transformational scale and scope of metadata

11. Big Data management on cloud enables customers to reduce TCO and maximise ROI by using Informatica Big Data management on cloud.

12. With the help of Informatica's Digital Transformation framework, businesses can become more agile, realise new opportunities for growth, and invent new products.

13. Informatica's Intelligent Data Management Cloud (IDMC) platform powers the Data Operations Performance Analytics solution, which provides real-time predictive insights into data-driven decision-making operations to avoid noncompliance and revenue loss.

14. Informatica's Intelligent Data Management Cloud (IDMC) platform, in combination with the Metadata Management product from Informatica, allows for a richer view of digital assets across various industries.

15. Customer-focusedness. For starters, the company needs to change its focus from being product-focused to one that is more customer-focused. When it comes to digital transformations, companies that have a firm grasp on their customers' wants and needs are the most well-prepared and successful. Putting things in perspective and prioritising next steps are made easier when you consider what's best for the customer.

16. The organization's structure For digital transformation, a culture of openness and acceptance is needed. Break down internal silos and unite executives and leaders around a new digital strategy..

17. Adapting to Change - Many digital transformations fail because of a lack of support from employees. People are hard-wired to resist change, even if they can see the benefits of doing so. With today's fast-paced business environment, the most effective change management efforts are those that are in sync.

18. Transformative leadership. It is possible for a strong leader to make employees feel safe during times of transition. To be a transformational leader, you must inspire people to take action and make them feel as if they are part of something greater than themselves. Therefore, every executive and leader has a critical role to play in promoting digital transformation.

19. Technology options. Digital transformation decisions must not be made in isolation. Most purchases are made by a team of 15 or more people, with about half of that group working in information technology. In order to represent the company's overall goals, leaders must work together to represent their respective departments.

20. Integration in the year . Focusing on data aids in the integration of digital solutions across the organisation. The larger the company, the more complicated the data approach. A streamlined data strategy is necessary for a successful digital transformation.

21. Customer satisfaction within the company. Digital transformation has a significant impact on the internal customer experience—the employee experience. Providing consumer-grade technology solutions and collecting employee feedback significantly increases the ability of employees to deliver an exceptional experience.

22. Management of the supply chain and logistics When customers can't count on getting their products or services on time, it's time to implement digital transformation. Logistics and supply chain management can be more effective when done digitally.

23. Data privacy, security, and ethics should all be taken into account. The vast majority of consumers believe that their personal information is at risk of being compromised. Prioritizing data security as part of a digital transformation is a good idea.

24. Evolution in the development of new products, services, and methods. As well as a change in the products and services themselves, digital transformation necessitates a shift in how they are delivered. The delivery of modern products is more intelligent and innovative.

25. Digitization. In order to fully digitalize a business, it is necessary to link the online and offline worlds in a seamless manner. Retailers like Target and Best Buy have found great success in blurring the lines between online and in-store shopping.

26. It's all about you, man! Customers look forwards to receiving attention that is specific to their needs. Improve your understanding of your customers through the use of digital solutions, and then deliver recommendations and experiences that are specifically tailored to their needs.

27. Data management productivity can be improved with machine learning and Informatica.

28. Digitizing a data warehouse is a process, not a destination. When done correctly, the transition from the data warehousing of the past to the data warehousing of the future is a planned and incremental process. Informatica provides the adequate provision for this.

29. Begin with a list of your data warehouses and other data sources, which you may have more than one of. Each has a specific purpose and a specific user. Aim to discover overlapping and redundant warehouses. With Informatica, it could be easily done.

30. Evaluate your current data warehousing situation. Prioritize your needs and challenges so that you know which ones are most urgent and require the most attention.

31. Look three to five years into the future and identify high-priority use cases that may arise in the future. Establish your long-term data warehousing goals. In order to know when your modernisation goals have been met, you must clearly define and describe them.

32. If you haven't already done so, map out your current data management architecture from data sources, through warehouses and lakes, to the delivery of data to consumers. For maximum cohesion with the coexistence of data lakes and data warehouses, rethink your data management architecture

33. You should select the best modernisation patterns based on your objectives. Don't be afraid to play around with pattern combinations. Data warehouse automation can help you migrate a legacy data warehouse to the cloud by helping you reverse engineer it. Alternatively, you could move a high-use data warehouse to the cloud and federate other data warehouses to break down the siloed nature of the data. Informatica helps around here.

34. Instead of focusing on the here and now, make long-term plans. Take a long-term view of the situation. If you only think about today's needs, you'll be out of date before you've even begun. Consider the long-term implications of your technology choices. Preserve the value of your data management assets by incorporating AI/ML into all stages of the data management lifecycle, from storage to consumption. With Informartica’s digital technology, it gets better result.

35. Don't try to do everything all at once. Repeat the entire procedure. By the time you've completed each step, your current state, priorities, and thinking about the future state will have changed. The technology will, of course, keep evolving. Informatica aids a lot to achieve these measurements.

Disclaimer: This article is a paid publication and does not have journalistic/ editorial involvement of Hindustan Times. Hindustan Times does not endorse/ subscribe to the contents of the article/advertisement and/or views expressed herein.

Hindustan Times shall not in any manner, be responsible and/or liable in any manner whatsoever for all that is stated in the article and/or also with regard to the views, opinions, announcements, declarations, affirmations etc., stated/featured in same. The decision to read hereinafter is purely a matter of choice and shall be construed as an express undertaking/guarantee in favour of Hindustan Times of being absolved from any/ all potential legal action, or enforceable claims. The content may be for information and awareness purposes and does not constitute advice.

SHARE THIS ARTICLE ON
SHARE
Story Saved
×
Saved Articles
Following
My Reads
My Offers
Sign out
New Delhi 0C
Friday, October 07, 2022
Start 15 Days Free Trial Subscribe Now
Register Free and get Exciting Deals