Journey to the Modern App with Containers, Microservices and Big Data

Editor's Notes

• #4: Data is growing at 40% YoY [http://www.emc.com/leadership/digital-universe/2014iview/executive-summary.htm] Big data is not just about Volume, but also Velocity
• #5: Our world is getting increasingly connected, and all those connections represent opportunities and challenges. Users now expect personalized experiences, and companies need to leverage data in real time to manage risk, create value and stay competitive.
• #6: Our world is getting increasingly connected, and all those connections represent opportunities and challenges. Users now expect personalized experiences, and companies need to leverage data in real time to manage risk, create value and stay competitive.
• #7: Mayo Clinic is bringing real-time clinical decision making to the point of care with a vertical application that allows physicians to find similar patients and explore what-if scenarios using outcome and intervention parameters. This allows Mayo to improve surgical outcomes by managing pre and post-operative care. One of the main challenges for Mayo Clinic was the rapid processing of historical clinical notes, radiology notes, and other unstructured data resources in order to deliver real-time, personalized clinical care recommendations. Big data architecture at Mayo consists of three layers: (i) data ingestion layer that reads data from real-time feeds from the EMR and archived data, (ii) big data analytics layer that does stream processing for analyzing the data, and (iii) data storage and retrieval that stores the information and knowledge that are generated through big data analytics and facilitate retrieval at the appropriate time for clinical use.
• #8: Uber connects riders with drive partners for an average one million trips a day, providing safe, reliable, convenient transportation at a variety of price points in more than 311 cities around the world. Key technologies: distributed logging via Apache Kafka and large-scale data processing via Apache Spark.
• #9: The common thread is that successful companies re using a new application architecture - one that scales and can process data in real time. -- Data and user growth: Users http://www.kpcb.com/internet-trends 2+ Billion smartphone users in 2014 3+ Billion internet users in 2016 Data http://www.edgelens.com/mobile-data-growth-to-hit-enterprise/#sthash.IMRP9gtP.EVQGv8YX.dpbs Cisco’s VNI predicts 10EB monthly in 2016. Global Mobile data traffic at 10 Exabytes/month in 2016 with 78% CAGR http://www.emc.com/leadership/digital-universe/2014iview/executive-summary.htm Enterprise data growth at 40% CAGR
• #10: Twitter started like a traditional app. Didn’t scale anymore On the right is microservices Agility Scalability Efficiency (Utilization)
• #11: Narrative: Containers, Big Data, Microservices, Open source software Each bring their own value. But greatest impact is if you use them together. Iike verizon did The power of Modern enterprise apps come from from its modular architecture, enabling enterprises to make updates quickly, and easily scale to process data or serve users. The first is microservices running in containers. Microservices enable teams to work self-sufficiently, and deliver new functionality and iterate improvements quickly. Containers simply packaging of the code, and can be spun up quickly - for example we may need to scale up services to deal with an influx of users. The second major component is big data services, as microservices don’t typically store data. Big Data services process information and retain the states of modern apps (e.g., user activity, sensor data, probability of an event). Big data services are commonly open source technologies like Spark, Cassandra, and Kafka, among others. Using open source technologies enable enterprises to use the latest technologies and not rebuild from scratch. Modern apps are distributed systems running on multiple servers, VMs, or cloud instances. And portions of the app might scale with a variety of factors - speed of data being ingested, volume of data stored, number of users being served, or type of analysis being performed.
• #12: Start with containerizing your current apps. You don’t need to rewrite. Big data services allow you to gain insights into your users, and improve your products and services Microservices sets you up for future agility and scalability
• #14: Containers by themselves are useful - they are lightweight, start faster, you achieve higher workload density. Difference: doesn’t include a full OS and doesn’t virtualize the hardware Ref verizon talk at mesoscon 50k containers
• #16: Traditional architecture - challenges: it’s hard to innovate and scale Micro services: application architecture is broken down into small pieces of functionality. Easier to organize an engineering team around those services, and scale the pieces independently. Each team can innovate on their service within the agreed upon API boundaries. Teams can work self-sufficiently, fewer organizational dependencies
• #17: Did this at Twitter, Airbnb: Example: who to follow, Airbnb matching guest and host using data Poll crowd: who is doing one of these? They are distributed, benefits of VMs don’t apply
• #18: The traditional approach of running modern enterprise apps organize (or silo) the infrastructure based on the technology being used. You might have a cluster for Spark, Kafka, Cassandra, and so on. This is because these services are distributed systems with their own scheduling logic to grow and shrink their footprint, so operators run them in different clusters to avoid resource conflicts. The result is extremely low utilization, and waste of infrastructure resources. The DC/OS approach turns the entire logical datacenter (on prem, cloud, or both) into one giant computer - it does this by running an agent on your existing servers and aggregates how the resource is being used. A DC/OS powered datacenter therefore runs like a single computer. Just as your personal computer wouldn’t ask you which core to run your web browser, this giant computer with DC/OS runs all modern apps services with no requirement for intervention by an operator. With DC/OS, all components of modern apps can be pooled: containerized microservices, big data services for analytics and data persistence. This approach effectively abstracts modern apps with the underlying infrastructure (physical, virtual or cloud), and has several benefits. 1: First, pooling drives higher utilization without impacting performance, and delivers capex savings. 2: Second is automation with dramatically simplified app development, rollout, and production operations - you can imagine the silos on the left being true for every phase of the software development lifecycle (dev, test, staging, prod) 3: Third is the ability to plug in datacenter-wide services - whether they be for CI/CD, data services, or services yet to come. 4: Lastly, this whole approach accelerates innovation because we dramatically lowers the technical expertise you need in your organization to use or experiment with new technologies (e.g., the next spark)
• #19: The traditional approach of running modern enterprise apps organize (or silo) the infrastructure based on the technology being used. You might have a cluster for Spark, Kafka, Cassandra, and so on. This is because these services are distributed systems with their own scheduling logic to grow and shrink their footprint, so operators run them in different clusters to avoid resource conflicts. The result is extremely low utilization, and waste of infrastructure resources. The DC/OS approach turns the entire logical datacenter (on prem, cloud, or both) into one giant computer - it does this by running an agent on your existing servers and aggregates how the resource is being used. A DC/OS powered datacenter therefore runs like a single computer. Just as your personal computer wouldn’t ask you which core to run your web browser, this giant computer with DC/OS runs all modern apps services with no requirement for intervention by an operator. With DC/OS, all components of modern apps can be pooled: containerized microservices, big data services for analytics and data persistence. This approach effectively abstracts modern apps with the underlying infrastructure (physical, virtual or cloud), and has several benefits. 1: First, pooling drives higher utilization without impacting performance, and delivers capex savings. 2: Second is automation with dramatically simplified app development, rollout, and production operations - you can imagine the silos on the left being true for every phase of the software development lifecycle (dev, test, staging, prod) 3: Third is the ability to plug in datacenter-wide services - whether they be for CI/CD, data services, or services yet to come. 4: Lastly, this whole approach accelerates innovation because we dramatically lowers the technical expertise you need in your organization to use or experiment with new technologies (e.g., the next spark)
• #20: Unlike traditional apps which are monolithic, Modern enterprise apps are built on microservices running in containers, big data services, and open source technology that’s always evolving. These microservices are built and rolled out by small teams organized around business objectives. Running and operating modern apps effectively requires a new approach to running the infrastructure. The DC/OS model of running infrastructure allows businesses to treat the entire datacenter (on premise or in the cloud) as one single form factor. Developers code against the datacenter (installing datacenter-wide apps services as needed), and operators issue datacenter-wide commands for administration and operations. What makes this possible is DC/OS’s unique ability to pool all modern app components - stateless apps, containers, and stateful big data. This abstracts developers and operators away from the underlying infrastructure - they’re driving policies at the datacenter level, and not thinking about individual containers or virtual machines. This approach is distinct from virtual machines, which consolidate monolithic apps into a single server, and operations are always at the VM level, not datacenter scale. Benefits of this approach: Data Agility - Processing lots of data - batch and in real time. Personalization, Predictive analytics, Anomaly detection, Internet of Things Developer Agility - Faster time to market of new services. Shipping code faster with CI/CD. Accelerated time to market. Top quartile companies have 3x productivity, 4x throughput, 5x quality Operational Agility - Hybrid cloud, container orchestration, quality of service at scale. Serving millions of users Container operations. Hybrid cloud. Ability to run legacy and modern application components ============ Every evolution in enterprise IT has been about improving What these enterprises have done is change how they build apps, and how they run infrastructure - making their businesses more competitive to meet today’s expectations on scale and speed. Modern enterprise apps are built on microservices running in containers, big data services, and open source technology that’s always evolving. DC/OS model of running infrastructure allows businesses to treat the entire datacenter (on premise or in the cloud) as one single form factor. Developers code against the datacenter (installing datacenter-wide apps services as needed), and operators issue datacenter-wide commands for administration and operations. What makes this possible is DC/OS’s unique ability to pool all modern app components - stateless apps, containers, and stateful big data. This abstracts developers and operators away from the underlying infrastructure - they’re driving policies at the datacenter level, and not thinking about individual containers or virtual machines. This approach is distinct from virtual machines, which consolidate monolithic apps into a single server, and operations are always at the VM level, not datacenter scale. Let’s spend a few minutes on modern apps and how DC/OS powers them. -------------- To stay competitive, enterprises need to act quickly and capture new value streams - this means building new business models, improving customer engagement & making real-time data-driven decisions. For CIOs and CTOs, this means managing unprecedented scale and speed, on top of the traditional challenges of efficiency, security, reliability. There are two aspects to unprecedented scale. The first is processing lots of data, both batch and in real-time. Enterprises are swimming in data, growing at 40% a year. Looking ahead, every major opportunity for businesses and the scientific community will requiring processing this big data and acting on insights. Examples include include IoT, anomaly detection, predictive analytics, and personalization. The second component to scale is users - serving millions of them through a variety of devices. A core requirement of personalization is scalability to capture, process, and react to users on an individual basis. Accelerated speed (Time to Market) is the expectation of today’s users and is key to staying competitive. Businesses are only as responsive as the software that powers them. According to a McKinsey study last year on software practices and enterprise performance, top quartile companies have 3x more productive developers, 4x more overall throughput, and 1/5 the defects in rolling out software compared to bottom quartile companies. Addressing these challenges means changing how you build and run software software. -- http://www.emc.com/leadership/digital-universe/2014iview/executive-summary.htm http://www.mckinsey.com/industries/high-tech/our-insights/the-perils-of-ignoring-software-development
• #21: Only way to run all these datacenter-scale services effectively is with a datacenter-scale approach Datacenter scale commands vs. individual containers or machines Need to think about operations of whole system This is the ops model that is enabled by DC/OS Unlike traditional apps which are monolithic, Modern enterprise apps are built on microservices running in containers, big data services, and open source technology that’s always evolving. These microservices are built and rolled out by small teams organized around business objectives. Running and operating modern apps effectively requires a new approach to running the infrastructure. The DC/OS model of running infrastructure allows businesses to treat the entire datacenter (on premise or in the cloud) as one single form factor. Developers code against the datacenter (installing datacenter-wide apps services as needed), and operators issue datacenter-wide commands for administration and operations. What makes this possible is DC/OS’s unique ability to pool all modern app components - stateless apps, containers, and stateful big data. This abstracts developers and operators away from the underlying infrastructure - they’re driving policies at the datacenter level, and not thinking about individual containers or virtual machines. This approach is distinct from virtual machines, which consolidate monolithic apps into a single server, and operations are always at the VM level, not datacenter scale. Benefits of this approach: Data Agility - Processing lots of data - batch and in real time. Personalization, Predictive analytics, Anomaly detection, Internet of Things Developer Agility - Faster time to market of new services. Shipping code faster with CI/CD. Accelerated time to market. Top quartile companies have 3x productivity, 4x throughput, 5x quality Operational Agility - Hybrid cloud, container orchestration, quality of service at scale. Serving millions of users Container operations. Hybrid cloud. Ability to run legacy and modern application components ============ Every evolution in enterprise IT has been about improving What these enterprises have done is change how they build apps, and how they run infrastructure - making their businesses more competitive to meet today’s expectations on scale and speed. Modern enterprise apps are built on microservices running in containers, big data services, and open source technology that’s always evolving. DC/OS model of running infrastructure allows businesses to treat the entire datacenter (on premise or in the cloud) as one single form factor. Developers code against the datacenter (installing datacenter-wide apps services as needed), and operators issue datacenter-wide commands for administration and operations. What makes this possible is DC/OS’s unique ability to pool all modern app components - stateless apps, containers, and stateful big data. This abstracts developers and operators away from the underlying infrastructure - they’re driving policies at the datacenter level, and not thinking about individual containers or virtual machines. This approach is distinct from virtual machines, which consolidate monolithic apps into a single server, and operations are always at the VM level, not datacenter scale. Let’s spend a few minutes on modern apps and how DC/OS powers them. -------------- To stay competitive, enterprises need to act quickly and capture new value streams - this means building new business models, improving customer engagement & making real-time data-driven decisions. For CIOs and CTOs, this means managing unprecedented scale and speed, on top of the traditional challenges of efficiency, security, reliability. There are two aspects to unprecedented scale. The first is processing lots of data, both batch and in real-time. Enterprises are swimming in data, growing at 40% a year. Looking ahead, every major opportunity for businesses and the scientific community will requiring processing this big data and acting on insights. Examples include include IoT, anomaly detection, predictive analytics, and personalization. The second component to scale is users - serving millions of them through a variety of devices. A core requirement of personalization is scalability to capture, process, and react to users on an individual basis. Accelerated speed (Time to Market) is the expectation of today’s users and is key to staying competitive. Businesses are only as responsive as the software that powers them. According to a McKinsey study last year on software practices and enterprise performance, top quartile companies have 3x more productive developers, 4x more overall throughput, and 1/5 the defects in rolling out software compared to bottom quartile companies. Addressing these challenges means changing how you build and run software software. -- http://www.emc.com/leadership/digital-universe/2014iview/executive-summary.htm http://www.mckinsey.com/industries/high-tech/our-insights/the-perils-of-ignoring-software-development
• #25: The traditional approach of running modern enterprise apps organize (or silo) the infrastructure based on the technology being used. You might have a cluster for Spark, Kafka, Cassandra, and so on. This is because these services are distributed systems with their own scheduling logic to grow and shrink their footprint, so operators run them in different clusters to avoid resource conflicts. The result is extremely low utilization, and waste of infrastructure resources. The DC/OS approach turns the entire logical datacenter (on prem, cloud, or both) into one giant computer - it does this by running an agent on your existing servers and aggregates how the resource is being used. A DC/OS powered datacenter therefore runs like a single computer. Just as your personal computer wouldn’t ask you which core to run your web browser, this giant computer with DC/OS runs all modern apps services with no requirement for intervention by an operator. With DC/OS, all components of modern apps can be pooled: containerized microservices, big data services for analytics and data persistence. This approach effectively abstracts modern apps with the underlying infrastructure (physical, virtual or cloud), and has several benefits. 1: First, pooling drives higher utilization without impacting performance, and delivers capex savings. 2: Second is automation with dramatically simplified app development, rollout, and production operations - you can imagine the silos on the left being true for every phase of the software development lifecycle (dev, test, staging, prod) 3: Third is the ability to plug in datacenter-wide services - whether they be for CI/CD, data services, or services yet to come. 4: Lastly, this whole approach accelerates innovation because we dramatically lowers the technical expertise you need in your organization to use or experiment with new technologies (e.g., the next spark)
• #27: DC/OS has a strong community to build and enhance DC/OS as the definitive platform for running modern enterprise apps. Two benefits are that are more and more ISVs whose products run elastically on DC/OS, and secondly you can be confident that the DC-scale OS running your infrastructure has a strong community behind its development.