Global Azure Bootcamp 2017 - Performance and Health Management for Modern Applications

Editor's Notes

• #4: 3 person persona journey 1: Developer 2: IT Ops 3: Network Admin
• #5: Persona 1: Developer
• #6: With the cloud and mobile app stores, the cost and time to enter the market has reduced dramatically. Business models are getting copied and it’s very easy to build an app or service that is similar to yours, and it’s hard to keep your differentiation. Technology and requirements are changing continuously, and you need to be on top of these changes to evolve your app to be successful With rapid continuous delivery cycles, e.g. we ship every week, some services are shipping daily, and the faster you move, there are more chances for quality to go down Given all this, success requires data-driven decision making. You need the right data about your customers, about your live site, and all the aspects of your app, so you can make decisions quickly and adapt to be successful.
• #7: With this premise, let’s look at the modern ALM cycle This is a continuous cycle
• #8: Plan: Prioritize investments Dev/Test: Lack of collaboration between Devs and Ops Release: Quickly detect, diagnose, triage issues Monitor/Learn: Contextual info People doing the planning are not exposed to real world data that is coming out from customers and usage of their apps to validate and make good prioritization decisions Developers don’t have full visibility to what happens in prod, because ops have a different set of tools and processes, and this leads to lack of collaboration between them Customers usually become aware of issues before us, and we don’t have enough info to decide whether an issue requires a hotfix or can wait until the next version Once you have info that something is failing, you need the information that will help you resolve the issue in the fastest way Application Insights is a service that will help you solve these problems through your application management lifecycle.
• #9: Telemetry from server backend, middleware, web, browser Telemetry process with Machine Leaning Portal diagnose, display
• #10: Intelligent APM Instant Analytics Integrated DevOps
• #11: Ingestion: How do we get the data in Exploration: How do we work with the data collected Correlation: How do we visualize the data
• #12: What sources of telemetry are collected by Application Insights? Outside-in User behavior Application behavior Custom tracing and events Infra performance
• #13: Built-in analytics Deep insights 360 view
• #16: Persona 2: IT Ops
• #17: What system is hosing what service Monitor distributed service Know change impact To track dependencies, IT infrastructure has relied on conventional techniques such as endless spreadsheets and extensive audits. But the proliferation of virtualized data centers, cloud, and micro services, has made it increasingly difficult to track dependencies using these archaic methods.
• #18: The increased complexity of modern applications has changed how we how we define “applications”: As complex, multi-tier, business-service systems that may span multiple datacenters and cloud hosting environments. With this complexity comes increased challenges for the teams supporting those applications, yet IT operations still faces a set of tools focused on individual aspects of application infrastructure, from infrastructure monitoring solutions to tools that provide code-level analysis. To manage these complex applications, and ensure they meet their SLAs , IT Ops needs an end-to-end view that ties together the different application components and infrastructure services required that make up the critical business applications they support.
• #19: What if… Auto discovery Processes, Services Dynamic map Issues across dependancies
• #20: Discovery Automatically build a common reference map of dependencies across servers, processes, and 3rd party services. Incident Management View cascading alerts, failed connections, load balancing issues, and rogue clients. Migration Assurance Ensure nothing is left behind, identify connectivity failures, view computer and process inventory, and identify systems for decommissioning.
• #21: Service Map presents a view of your servers as you think of them - as interconnected systems that deliver services and rely on other technologies. Service Map discovers and maps server and process dependencies in real-time, without any predefinition, and visualizes application components, service dependencies, and supporting infrastructure configuration. This helps you eliminate the guesswork of problem isolation, identify surprise connections and broken links in your environment, and perform Azure migrations knowing that critical systems and endpoints won’t be left behind. The Service Map public preview supports Windows and Linux guests, in any cloud and on-prem.
• #24: Hop-by-Hop Performance monitoring to identify the real-world performance response times across the entire application system. This allows you to identify where the bottleneck is occurring, as well as clues into how to resolve the problem. Users can configure events and alerts to notify you before performance issues are experienced broadly. Integration with dashboarding and visualization tools. Giving you the flexibility and data to provide a single pane of glass to monitor the health of your application systems
• #25: Accelerate troubleshooting
• #26: Discover: Build application and server dependency maps in minutes rather than weeks or months No need for costly manual cataloging of servers, services, and applications Assess: Identify critical servers and applications Identify and fix existing configuration issues before migration Save on post-migration cost by identifying under-utilized servers Improve application performance by identifying over-utilized or under-provisioned VMs. Validate: Verify proper migration with Comparison Reports Quickly catch any performance issues in the new environment No need to install new tools to monitor cloud resources
• #35: Persona 3: Network Admin
• #36: - Monitors network links and these links can be between datacenters, office sites, clouds and even different tiers of an application. It measure packet loss and network latency on these links. In other words it checks whether the packets that are being sent from one end are reaching the other end and how much time is it taking for packets to reach other end <Click> So lets say you have two networks – network a and network b; NPM keeps a track of availability of the links and quality of connectivity <Click> And we do this using synthetic transactions. i.e. we have OMS agents installed on two ends of the link and these agents periodically exchange packets. We’ll see more on how solution works in a moment.
• #37: How many of you have used OMS before? Ok just to make sure that everyone is on same page here is a brief description of OMS architecture. Service has two parts – a cloud service and agents that can be installed on machines; these machine can be on premise or VM in the cloud. The agents collect different types of monitoring data and upload this data to the cloud service that indexes and aggregates this data and then this data is presented to the user on a web console. And here are some of the features of NPM.
• #38: NPM solution OMS Agent and PowerShell script Agent gets Intelligence Pack Detects subnet Learns of other OMS Agents TCP probe between agents Displayed in OMS OMS solution gallery is like a market place and user can pick their payment plan and choose the solution that they want. There are dozens on solution available there for different monitoring and managements requirements. There are solutions for SQL, active directory, update management, change tracking etc.
• #40: App Ins: Proactive detect Diagnose Service Map: Map dependencies Root cause Migrations NPM: Connections Network performance Learns baseline thresholds