Azure Big Data Architecture
Building the Data Lake with Azure Data Factory and Data Lake AnalyticsThe cloud is changing the way applications are designed, including how data is processed and stored. Instead of a single general-purpose database that handles all of a solution's data, polyglot persistence solutions use multiple, specialized data stores, each optimized to provide specific capabilities. The perspective on data in the solution changes as a result. There are no longer multiple layers of business logic that read and write to a single data layer. Instead, solutions are designed around a data pipeline that describes how data flows through a solution, where it is processed, where it is stored, and how it is consumed by the next component in the pipeline.
Control and protect your data through privileged access management capabilities
Big data architectures and the data lake
Azure SQL Database Managed Instance
Differentiate Big Data vs Data Warehouse use cases for a cloud solution
Building a modern data warehouse
Big data solutions typically involve one or more of the following types of workload:
- Batch processing of big data sources at rest.
- Real-time processing of big data in motion.
- Interactive exploration of big data.
- Predictive analytics and machine learning.
Consider big data architectures when you need to:
- Store and process data in volumes too large for a traditional database.
- Transform unstructured data for analysis and reporting.
- Capture, process, and analyze unbounded streams of data in real time, or with low latency.
Components of a big data architecture
The following diagram shows the logical components that fit into a big data architecture. Individual solutions may not contain every item in this diagram.Most big data architectures include some or all of the following components:
Data sources. All big data solutions start with one or more data sources. Examples include:
Application data stores, such as relational databases.
Static files produced by applications, such as web server log files.
Real-time data sources, such as IoT devices.
Data storage. Data for batch processing operations is typically stored in a distributed file store that can hold high volumes of large files in various formats. This kind of store is often called a data lake. Options for implementing this storage include Azure Data Lake Store or blob containers in Azure Storage.
Batch processing. Because the data sets are so large, often a big data solution must process data files using long-running batch jobs to filter, aggregate, and otherwise prepare the data for analysis. Usually these jobs involve reading source files, processing them, and writing the output to new files. Options include running U-SQL jobs in Azure Data Lake Analytics, using Hive, Pig, or custom Map/Reduce jobs in an HDInsight Hadoop cluster, or using Java, Scala, or Python programs in an HDInsight Spark cluster.
Build hybrid data platform with Azure SQL Database and SQL Server
Real-time message ingestion. If the solution includes real-time sources, the architecture must include a way to capture and store real-time messages for stream processing. This might be a simple data store, where incoming messages are dropped into a folder for processing. However, many solutions need a message ingestion store to act as a buffer for messages, and to support scale-out processing, reliable delivery, and other message queuing semantics. This portion of a streaming architecture is often referred to as stream buffering. Options include Azure Event Hubs, Azure IoT Hub, and Kafka.
Stream processing. After capturing real-time messages, the solution must process them by filtering, aggregating, and otherwise preparing the data for analysis. The processed stream data is then written to an output sink. Azure Stream Analytics provides a managed stream processing service based on perpetually running SQL queries that operate on unbounded streams. You can also use open source Apache streaming technologies like Storm and Spark Streaming in an HDInsight cluster.
Analytical data store. Many big data solutions prepare data for analysis and then serve the processed data in a structured format that can be queried using analytical tools. The analytical data store used to serve these queries can be a Kimball-style relational data warehouse, as seen in most traditional business intelligence (BI) solutions. Alternatively, the data could be presented through a low-latency NoSQL technology such as HBase, or an interactive Hive database that provides a metadata abstraction over data files in the distributed data store. Azure SQL Data Warehouse provides a managed service for large-scale, cloud-based data warehousing. HDInsight supports Interactive Hive, HBase, and Spark SQL, which can also be used to serve data for analysis.
Microservices architecture with .NET Core Docker containers and Azure
Azure IoT Edge with examples | Laurent Ellerbach | Microsoft
Orchestration. Most big data solutions consist of repeated data processing operations, encapsulated in workflows, that transform source data, move data between multiple sources and sinks, load the processed data into an analytical data store, or push the results straight to a report or dashboard. To automate these workflows, you can use an orchestration technology such Azure Data Factory or Apache Oozie and Sqoop.
Lambda architecture
When working with very large data sets, it can take a long time to run the sort of queries that clients need. These queries can't be performed in real time, and often require algorithms such as MapReduce that operate in parallel across the entire data set. The results are then stored separately from the raw data and used for querying.One drawback to this approach is that it introduces latency — if processing takes a few hours, a query may return results that are several hours old. Ideally, you would like to get some results in real time (perhaps with some loss of accuracy), and combine these results with the results from the batch analytics.
The lambda architecture, first proposed by Nathan Marz, addresses this problem by creating two paths for data flow. All data coming into the system goes through these two paths:
A batch layer (cold path) stores all of the incoming data in its raw form and performs batch processing on the data. The result of this processing is stored as a batch view.
A speed layer (hot path) analyzes data in real time. This layer is designed for low latency, at the expense of accuracy.
The batch layer feeds into a serving layer that indexes the batch view for efficient querying. The speed layer updates the serving layer with incremental updates based on the most recent data.
Data that flows into the hot path is constrained by latency requirements imposed by the speed layer, so that it can be processed as quickly as possible. Often, this requires a tradeoff of some level of accuracy in favor of data that is ready as quickly as possible. For example, consider an IoT scenario where a large number of temperature sensors are sending telemetry data. The speed layer may be used to process a sliding time window of the incoming data.
Data flowing into the cold path, on the other hand, is not subject to the same low latency requirements. This allows for high accuracy computation across large data sets, which can be very time intensive.
Building resilient microservices with .NET Core and Azure Kubernetes Service
Eventually, the hot and cold paths converge at the analytics client application. If the client needs to display timely, yet potentially less accurate data in real time, it will acquire its result from the hot path. Otherwise, it will select results from the cold path to display less timely but more accurate data. In other words, the hot path has data for a relatively small window of time, after which the results can be updated with more accurate data from the cold path.
The raw data stored at the batch layer is immutable. Incoming data is always appended to the existing data, and the previous data is never overwritten. Any changes to the value of a particular datum are stored as a new timestamped event record. This allows for recomputation at any point in time across the history of the data collected. The ability to recompute the batch view from the original raw data is important, because it allows for new views to be created as the system evolves.
Kappa architecture
A drawback to the lambda architecture is its complexity. Processing logic appears in two different places — the cold and hot paths — using different frameworks. This leads to duplicate computation logic and the complexity of managing the architecture for both paths.The kappa architecture was proposed by Jay Kreps as an alternative to the lambda architecture. It has the same basic goals as the lambda architecture, but with an important distinction: All data flows through a single path, using a stream processing system.
There are some similarities to the lambda architecture's batch layer, in that the event data is immutable and all of it is collected, instead of a subset. The data is ingested as a stream of events into a distributed and fault tolerant unified log. These events are ordered, and the current state of an event is changed only by a new event being appended. Similar to a lambda architecture's speed layer, all event processing is performed on the input stream and persisted as a real-time view.
If you need to recompute the entire data set (equivalent to what the batch layer does in lambda), you simply replay the stream, typically using parallelism to complete the computation in a timely fashion.
Internet of Things (IoT)
From a practical viewpoint, Internet of Things (IoT) represents any device that is connected to the Internet. This includes your PC, mobile phone, smart watch, smart thermostat, smart refrigerator, connected automobile, heart monitoring implants, and anything else that connects to the Internet and sends or receives data. The number of connected devices grows every day, as does the amount of data collected from them. Often this data is being collected in highly constrained, sometimes high-latency environments. In other cases, data is sent from low-latency environments by thousands or millions of devices, requiring the ability to rapidly ingest the data and process accordingly. Therefore, proper planning is required to handle these constraints and unique requirements.AI for an intelligent cloud and intelligent edge: Discover, deploy, and manage with Azure ML services
AI for an intelligent cloud and intelligent edge: Discover, deploy, and manage with Azure ML services from James Serra
The cloud gateway ingests device events at the cloud boundary, using a reliable, low latency messaging system.
Devices might send events directly to the cloud gateway, or through a field gateway. A field gateway is a specialized device or software, usually collocated with the devices, that receives events and forwards them to the cloud gateway. The field gateway might also preprocess the raw device events, performing functions such as filtering, aggregation, or protocol transformation.
After ingestion, events go through one or more stream processors that can route the data (for example, to storage) or perform analytics and other processing.
The following are some common types of processing. (This list is certainly not exhaustive.)
Writing event data to cold storage, for archiving or batch analytics.
Hot path analytics, analyzing the event stream in (near) real time, to detect anomalies, recognize patterns over rolling time windows, or trigger alerts when a specific condition occurs in the stream.
Handling special types of nontelemetry messages from devices, such as notifications and alarms.
Machine learning.
The boxes that are shaded gray show components of an IoT system that are not directly related to event streaming, but are included here for completeness.
The device registry is a database of the provisioned devices, including the device IDs and usually device metadata, such as location.
The provisioning API is a common external interface for provisioning and registering new devices.
Some IoT solutions allow command and control messages to be sent to devices.
Relevant Azure services:
- Azure IoT Hub
- Azure Event Hubs
- Azure Stream Analytics
Microsoft and Hortonworks Delivers the Modern Data Architecture for Big Data
Advanced analytics goes beyond the historical reporting and data aggregation of traditional business intelligence (BI), and uses mathematical, probabilistic, and statistical modeling techniques to enable predictive processing and automated decision making.
Advanced analytics solutions typically involve the following workloads:
- Interactive data exploration and visualization
- Machine Learning model training
- Real-time or batch predictive processing
Data storage. Advanced analytics solutions require data to train machine learning models. Data scientists typically need to explore the data to identify its predictive features and the statistical relationships between them and the values they predict (known as a label). The predicted label can be a quantitative value, like the financial value of something in the future or the duration of a flight delay in minutes. Or it might represent a categorical class, like "true" or "false," "flight delay" or "no flight delay," or categories like "low risk," "medium risk," or "high risk."
Batch processing. To train a machine learning model, you typically need to process a large volume of training data. Training the model can take some time (on the order of minutes to hours). This training can be performed using scripts written in languages such as Python or R, and can be scaled out to reduce training time using distributed processing platforms like Apache Spark hosted in HDInsight or a Docker container.
Real-time message ingestion. In production, many advanced analytics feed real-time data streams to a predictive model that has been published as a web service. The incoming data stream is typically captured in some form of queue and a stream processing engine pulls the data from this queue and applies the prediction to the input data in near real time.
Stream processing. Once you have a trained model, prediction (or scoring) is typically a very fast operation (on the order of milliseconds) for a given set of features. After capturing real-time messages, the relevant feature values can be passed to the predictive service to generate a predicted label.
Analytical data store. In some cases, the predicted label values are written to the analytical data store for reporting and future analysis.
Analysis and reporting. As the name suggests, advanced analytics solutions usually produce some sort of report or analytical feed that includes predicted data values. Often, predicted label values are used to populate real-time dashboards.
Orchestration. Although the initial data exploration and modeling is performed interactively by data scientists, many advanced analytics solutions periodically re-train models with new data — continually refining the accuracy of the models. This retraining can be automated using an orchestrated workflow.
Fundamentals of Kubernetes on Microsoft Azure
Machine learning
Machine learning is a mathematical modeling technique used to train a predictive model. The general principle is to apply a statistical algorithm to a large dataset of historical data to uncover relationships between the fields it contains.Machine learning modeling is usually performed by data scientists, who need to thoroughly explore and prepare the data before training a model. This exploration and preparation typically involves a great deal of interactive data analysis and visualization — usually using languages such as Python and R in interactive tools and environments that are specifically designed for this task.
Red Hat Openshift on Microsoft Azure
There are two broad categories of machine learning:
Supervised learning. Supervised learning is the most common approach taken by machine learning. In a supervised learning model, the source data consists of a set of feature data fields that have a mathematical relationship with one or more label data fields. During the training phase of the machine learning process, the data set includes both features and known labels, and an algorithm is applied to fit a function that operates on the features to calculate the corresponding label predictions. Typically, a subset of the training dataset is held back and used to validate the performance of the trained model. Once the model has been trained, it can be deployed into production, and used to predict unknown values.
Unsupervised learning. In an unsupervised learning model, the training data does not include known label values. Instead, the algorithm makes its predictions based on its first exposure to the data. The most common form of unsupervised learning is clustering, where the algorithm determines the best way to split the data into a specified number of clusters based on statistical similarities in the features. In clustering, the predicted outcome is the cluster number to which the input features belong. While they can sometimes be used directly to generate useful predictions, such as using clustering to identify groups of users in a database of customers, unsupervised learning approaches are more often used to identify which data is most useful to provide to a supervised learning algorithm in training a model.
Relevant Azure services:
- Azure Machine Learning
- Machine Learning Server (R Server) on HDInsight
- Deep learning
Openshift 3.10 & Container solutions for Blockchain, IoT and Data Science
Relevant Azure services:
- Deep Learning Virtual Machine
- Apache Spark on HDInsight
- Artificial intelligence
Democratizing Data Science on Kubernetes
Model deployment
The predictive services that support AI applications may leverage custom machine learning models, or off-the-shelf cognitive services that provide access to pretrained models. The process of deploying custom models into production is known as operationalization, where the same AI models that are trained and tested within the processing environment are serialized and made available to external applications and services for batch or self-service predictions. To use the predictive capability of the model, it is deserialized and loaded using the same machine learning library that contains the algorithm that was used to train the model in the first place. This library provides predictive functions (often called score or predict) that take the model and features as input and return the prediction. This logic is then wrapped in a function that an application can call directly or can be exposed as a web service.
Relevant Azure services:
- Azure Machine Learning
- Machine Learning Server (R Server) on HDInsight
- Choosing a cognitive services technology
- Choosing a machine learning technology
More Information:
https://docs.microsoft.com/en-us/azure/architecture/reference-architectures/https://docs.microsoft.com/en-us/azure/architecture/reference-architectures/data/enterprise-bi-sqldw
https://hyperonomy.com/category/microsoft-azure/
https://www.pluralsight.com/courses/azure-architecture-getting-started
https://docs.microsoft.com/en-us/azure/architecture/
https://wwt.com/all-blog/kubernetes-101/
https://docs.microsoft.com/en-us/learn/browse/?roles=solution-architect&products=azure
https://azure.microsoft.com/en-us/solutions/?cdn=disable
https://nagarajbhairaji.blogspot.com/2017/12/SmartHotel360-demoapp-Microsoft-Connect2017.html
https://mountainss.wordpress.com/2018/06/24/a-great-microservices-e-book-about-architecture-for-containerized-dotnet-apps-docker-kubernetes-containers/
https://github.com/dotnet-architecture/eShopModernizing/wiki/04.-How-to-deploy-your-Windows-Containers-based-apps-into-Kubernetes-in-Azure-Container-Service-(Including-CI-CD)
https://blogs.vmware.com/euc/2017/10/deploying-using-vmware-horizon-cloud-microsoft-azure-new-video-series.html
https://techzone.vmware.com/quick-start-tutorial-vmware-horizon-cloud-microsoft-azure/components-architecture
https://azurestack.blog/2016/10/dive-into-microsoft-azure-stack-architecture-part-1/
https://azurestack.blog/2016/10/dive-into-microsoft-azure-stack-architecture-part-2/
https://azurestack.blog/2016/10/adding-and-using-os-gallery-items-to-azure-stack-tp2/
Posts
0 reacties:
Post a Comment