autocon/22

Manu Thapar, CTO of Mastercard, explains Mastercard's ML-based fraud detection system and gives his POV on going 100% autonomous. Mastercard built a massive ML based system to detect and reduce fraud, running on AWS, using Kubernetes and serverless with EKS and Lambda. Manu explains the problem, Mastercard's solution and how it was deployed. Manu then gives his POV on Mastercard's goals for 100% autonomous operations and why autonomous systems are needed to meet the SLAs that global companies like Mastercard operate with.00:16: Massive scale of credit card transactions01:13: Mastercard's solution03:36:Rules-based vs ML based approaches05:58: Deployment on AWS with k8s and serverless09:11: Achieving performance & scalability with AWS10:30: Ensuring availability & resilience with well-architected design11:24: Assuring security & compliance - PCI, GDPR12:20: Closed loop analytics and reporting13:09: Customer impact on approval rates and fraud rates14:11: Principles for operating systems at scale15:47: Manu's POV on autonomous17:28: What % of systems can be autonomous Learn more about Sedai at https://bit.ly/3e3Cqlv#manuthapar #mastercard #antifraud #aiml #autonomouscloud #eks #kubernetes #aws #sedai #autocon22

This video covers the evolution of systems from mechanized to automatic to autonomous, and the five advantages of autonomous systems in this kickoff to the technical track of autocon/22.Key moments:0:00: autocon/22 welcome0:23: The evolution of systems: mechanized, automatic, autonomous0:54: Autonomous systems are everywhere1:28: How does autonomous help?2:15: How does autonomous relevant to production clouds?2:56: autocon/22 technical sessions listTo learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #lambda #kubernetes #sre #devops #aws #serverless #a8s4k8s #sedai #autocon22

Four engineering leaders with backgrounds at Paypal, Topcoder, eBay, Ironclad and Sedai discuss how to optimize cloud costs and performance with Autonomous Cloud, including how Autonomous complements Kubernetes by taking care of choosing configuration, and enables teams to focus on applications not infrastructure.Panelists:Sumesh Vadassery, Senior Director Engineering, PayPalKumar Rethi,CTO, TopCoder (ex-eBay)John Fielder, CISO and Head of Platform Engineering, IroncladBenjamin Thomas, Co-Founder & President, SedaiJohn Jamie, VP Marketing, Sedai (Moderator)Key moments:01:20: Evolution of Cloud Cost Management03:35: How PayPal saved 60% of compute costs by resizing the site using AI/ML05:51: PayPal can now focus on apps not infrastructure07:03: PayPal manages in a multi-cloud, multi business unit environment10:05: Why eBay needed to move from automation to autonomous14:44: Autonomous Complements Kubernetes16:35: How Salesforce used AI and autonomous to manage complex cloud platforms22:22: Ironclad's advice for implementing an autonomous platform on production clouds22:43: Autonomous helps with growing security workload24:14: Security, Platform, Data Eng are value centers not cost centers26:48: What is the need for autonomous when you have Kubernetes?28:23: Static rules become outdated rapidly29:44: Autonomous manages configuration when app behavior changes30:48: Who manages config for declarative platforms?31:32: Drivers for ROI for autonomous vs automated systems33:29: Autonomous can provide 50% gains in cost & performance over automated approachesTo learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #kubernetes #cloudcost #finops #eks #ecs #sre #devops #aws #sedai #autocon22, #sumeshvadassery, #paypal, #kumarrethi, #topcoder, #johnfielder, #ironclad, #benjaminthomas

Modern architecture alternatives and their pros and cons are debated in this panel discussion from autocon/22. While the right choice is organization-specific driven by application needs and organizational capabilities, the case is made that serverless is underutilized at many companies given its operational advantages.Panel members:Rachit Lohani, CTO & SVP (Engineering and Product), PaylocitySiddharth Ram, Ex-CTO, Inflection & VP/Engineering Fellow, IntuitShridhar Pandey, Senior Product Manager, Lambda, AWSKenneth Nguyen, Co-Founder, TasqSalil Deshpande, General Partner, Uncorrelated Ventures (moderator)Key moments:00:00: Context for the choice between serverless and Kubernetes02:23: Why not run stateless and just use serverless and autonomous?05:52: Why Inflection moved from Kubernetes to serverless08:13: What about vendor lockin with serverless?08:57: Paylocity Heavily Uses Serverless10:25: The 5 Steps to Get to Autonomous11:02: Getting to Autonomous in Stateless vs Stateful12:40: Why Paylocity Converted Almost Completely to Serverless14:03:Kubernetes Helps with High Throughput, Cold Starts, Latency Sensitivity14:35: Rachit's POV on Managing Kubernetes15:47: Refactoring Kubernetes to Serverless can be costly16:44: Paylocity's default approach is serverless; exceptions need a rationale17:33: Seek fewer variations & avoid abstraction complexity19:14: Lambda team aims to support different schools of thought19:30: Long running apps that wait for calls are not best fit for Lambda21:29: Why Intuit uses Kubernetes and Lambda22:54: Siddharth does see use cases for Kubernetes23:11: Tasq used serverless to iterate quickly24:24: Tasq now looking at distributed compute for processing (ECS, EKS, Fargate)26:15: Tasq's experience with serverless28:30: Solving for customer experience and minimizing ops work30:55: The Panel Compares Kubernetes, Serverless Containers and Serverless Functions32:20: Trading off more cloud spend for lower ops burden 33:25: Kubernetes can make sense as scale grows34:14: Open multiple paths for dev choice35:11: Portability with serverless is getting easier35:55: Summary: Use high productivity approaches (e.g, serverless) as default, systems language (k8s) when neededTo learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #lambda #kubernetes #eks #ecs #sre #devops #aws #serverless #a8s4k8s #sedai #autocon22, #rachitlohani, #paylocity, #siddharthram, #inflection, #shridharpandey, #kennethnguyen, #tasq, #salildeshpande, #uncorrelatedventures

In this kickoff to the leadership track of autocon/22, the autonomous cloud conference, Sedai CEO Suresh Mathew explains the difference between an autonomous and automated system, why autonomous systems are safer, why the industry should "shift up", and the four ages of IT operational management approaches leading to the "a8s-ops" (autonomous operations) age we are moving into.0:00: What Tesla figured out: how to separate tasks for machines vs people0:25: Separating Machine vs Human Jobs0:48: Autonomous vs Automated Systems (read more at https://bit.ly/3RimrOS)2:42: It's Time to Shift Up (read more at https://bit.ly/3ATcAcW)3:36. Sedai autocon/22 Announcements5:19: Conference WelcomeTo learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #lambda #kubernetes #sre #devops #aws #serverless #a8s4k8s #sedai #autocon22

Combining serverless with autonomous is a path to NoOps. From a developer's perspective serverless expert Sam Williams of Complete Coding outlines the history of operations, the impact of serverless and the benefits of autonomous in a serverless environment.Key moments:00:06: Run a software company without Ops or SREs?01:20: History of Operations06:37: Moving to Serverless is a Massive Step Forward09:17: Using Sedai to Autonomously Optimize Lambda12:00: Cold Starts13:26: Provisioned Capacity14:28: Business Impact of NoOps17:05: Release Intelligence18:24: Deliver features 5-10x faster with serverless vs EC219:20: How does the expansion of serverless impact developers?Visit Complete Coding at completecoding.ioTo learn more about Sedai visit https://bit.ly/3e3Cqlv#noops #serverless #samwilliams #completecoding #autonomouscloud #lambda #sre #devops #aws #serverless #fabric #sedai #autocon22

Lambda extensions offer a way to easily integrate Lambda with your favorite monitoring, observability, security, and governance tools. Shridhar Pandey of AWS outlines why extensions were built, how they work, and how to get started with them in this video from autocon/22.Key moments:01:07: Why AWS built Lambda Extensions04:32:How Lambda Extensions Work08:38: Complete Lifecycle of Extensions vs Lambda11:55: How to Get Started with Lambda Extensions14:19: Q&A: Operational Burden Saved with Extensions#autonomouscloud #lambda #lambdaextensions #sre #devops #aws #serverless #sedai #autocon22 #shridharpandey

Existing Datadog users can now add autonomous management capabilities, improving the performance, cost and availability of their applications while avoiding the time & cost of traditional automation. In this session from autocon/22, Alex Sweetser of Sedai and Monica Cortazar of Datadog walk through the limitations of current approaches, describe top use cases for autonomous with Datadog, and show how to get setup and use Sedai with Datadog using the integration available in the Datadog marketplace. Key moments:01:32: Manual Remediation with Datadog02:15: Automated Remediation with Datadog03:19: Where automation fails03:47: Autonomous Approach with Datadog04:39: Autonomous performance use case for Datadog05:54: Release intelligence use case for Datadog06:31:Autonomous SLOs for Datadog07:06: Datadog & Sedai: Better Together08:17: Q&A: Managing from Datadog vs Sedai08:58: Q&A: Where do you add autonomous to Datadog09:27: Walkthrough of setup in Sedai10:32:Installing & Using The Datadog Integration Tile11:37: End of Q&ATo learn more about Sedai visit https://bit.ly/3e3CqlvTo install Sedai from the Datadog marketplace visit https://app.datadoghq.com/marketplace/app/sedai-sedai-license/overviewTo read the Datadog docs for Sedai visit https://docs.datadoghq.com/integrations/sedai_sedai/#autonomouscloud #datadog #lambda #kubernetes #sre #devops #aws #serverless #a8s4k8s #sedai #autocon22

In this session AWS container hero Dijeesh Padinharethil covers autoscaling in Kubernetes, cluster autoscaling, scaling tools and event-driven autoscaling.Key moments:00:53: Why Kubernetes?03:13:Autoscaling in Kubernetes05:11: Cluster Autoscaling06:49: Cluster Autoscaler08:25: Challenges with Cluster Autoscaler10:47: Karpenter Autoscaler14:05: Karpenter Challenges14:53: Event Driven Autoscaling in Kubernetes16:02: What SREs are looking for18:05: Relationship to Fargate?19:58: Keda as an autoscaler?21:20: Recommendations outside Keda and KarpenterTo learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #autoscaler #hpa #vpa #keda #karpenter #kubernetes #sre #devops #aws #a8s4k8s #sedai

Kubernetes utilization is typically poor with only 20-45% of requested resources used. Kubernetes optimizations must meet application demands and minimizing idle resources. This video covers the potential to optimize at the pods and node/cluster level. The capabilities and limitations of Kubernetes HPA, VPA and Cluster Autoscaler are covered here.Key moments:00:24: Resource wastage common 01:22: Kubernetes utilization is typically poor and 35% of cloud spend is wasted02:23: How to balance application performance demand and minimizing idle resources02:41: Autoscaling is a major pillar to manage performance and cost02:54: Kubernetes offers scaling at pod or 03:32: Horizontal Pod Autoscaling (HPA)05:00: How Cluster Autoscaler Works06:06: Vertical Pod AutoScaler (VPA)07:46: When to use HPA vs VPA08:50: Kubernetes Autoscalers have serious limitations11:50: Cluster Autoscaler has limitations also13:31: Final Thoughts: scaling is key but admin overhead is high14:24: Autonomous can solve the issues15:04: Q&A - Why can't we use HPA and VPA16:32: Q&A - How configure HPA and HPA for stateful workloads?17:28: Q&A - Are there other tools that could help in HPA & VPA configuration? e.g,., Keda18:45: Q&A - How does compliance/change management work with autonomous systems?20:37: Q&A: Watchouts for performance management for K8s deployments22:16: Q&A: Differences between k8s providers To learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #kubernetes #hpa #vpa #autoscaling #sre #devops #aws #a8s4k8s #sedai #autocon22

Engineering & cloud infrastructure experts from Paylocity, Palo Alto Networks, fabric and Norwest Venture Partners discuss the drivers behind the growth of autonomous cloud management in this panel discussion from autocon/22. The participants are all looking to innovate faster and invest more in applications that drive customer outcomes and are using autonomous to help achieve those goals focusing on well defined use cases that free up their teams for higher value work. Panel participants:Rachit Lohani, CTO & SVP Engineering and Product, PaylocitySuresh Sangiah, SVP Engineering, Palo Alto NetworksPrakash Muppirala, EVP Platform, fabricMatt Howard, General Partner, Norwest Venture Partners (moderator)Key moments:00:38: Tech professionals are increasingly oversubscribed01:23: Delivering on promises to customers02:20: AI/ML has become practical03:33: The era of Collaborative Autonomy04:03: Palo Alto Network's approach to Autonomous05:23: Palo Alto Network's use cases for Autonomous06:13: How did Palo Alto Networks sell Autonomy internally?07:45: Paylocity's view on Autonomy vs their next generation service concept09:34: Importance of automation to create time for innovation and avoid low value toil09:59: Paylocity on Autonomy vs Automation (read more at https://bit.ly/3RimrOS)10:57: How Paylocity built trust with Autonomy12:33: Impact of Autonomous systems on productivity at Paylocity12:59: fabric's life before & after Autonomous15:37: Palo Alto's results from Autonomous & bottlenecks18:42: Did autonomy make people at Palo Alto feel threatened?20:28: Autonomous use cases at Paylocity22:37: How Paylocity uses KPIs to monitor performance24:38: How fabric avoided harm when rolling out autonomous with a hybrid model27:14: Palo Alto Network's view on capabilities to maintain internally28:13: Palo Alto Network's Roadmap for Autonomy29:15: The ROI of Autonomy at Paylocity30:26: Important Tasks for Humans Change Over Time31:44: fabric's advice for others adopting autonomousTo learn more about Sedai's autonomous cloud management platform visit https://bit.ly/3e3Cqlv#autonomouscloud #lambda #kubernetes #eks #ecs #sre #devops #aws #serverless #a8s4k8s #sedai #autocon22

Siddharth RAM, Ex-CTO Inflection/GoodHire & Intuit Fellow, explains why FCIs (Failed Customer Interactions) are a better and more customer-centric way to monitor a customer's experience with your application than traditional time-based availability metrics. At Goodhire, FCIs were reduced from 3.2% to ~0% through a range of initiatives including adopting an autonomous system.Chapters:00:22: Why Monitor Failed Customer Interactions (FCIs)?00:59: Defining Engineering & Operational Excellence03:02: The Problems of Traditional Availability04:21: Availability Doesn't Work for Distributed Modern Systems05:49: Problems of Time Based Metrics06:48: Customer Centric Metrics07:19: What is an FCI?09:30: It's easy to measure FCIs in your observability platform10:36: FCI Case Study: GoodHire11:54: How to Implement FCIs?14:16: How GoodHire Used Sedai16:38: Q&A: At what level should you measure FCIs?18:11: Q&A: How did you reinvest productivity gains?To learn more about Sedai visit https://bit.ly/3e3Cqlv#fcis #failedcustomerinteractions #availability #siddharthram #autonomouscloud #aws #serverless #sedai #autocon22

Observability is a building block for autonomous systems. This session covers the problem with too many metrics, how Palo Alto Networks solved this problem and Sedai's approach to metric prioritization.Key moments:00:22: Picking the right metrics01:18: What is Autonomous Cloud Management?02:15: Sedai Architecture03:14: The Metrics Overload Problem04:53: Most metrics data is redundant, irrelevant and noisy06:12: Only some metrics identify root cause07:17: Overview of Palo Alto Networks07:58: SASE is a new approach to security09:33: Palo Alto Network's monitoring challenge11:17.:Golden metrics were great but noisy11:28: Palo Alto Network's Monitoring Approach15:21: The right monitoring approach is a journey15:50: Working in a cycle to improve monitoring & remediation16:39: Solving for multiple monitoring providers and multiple resources17:45:Data hierarchy in an Autonomous System19:07:Four ways to classify metrics19:30: Sedai's Three Categories of Metrics: Input, Secondary & Primary20:57: Building a correlation chain between metrics22:25: Getting to a small list of curated metrics22:41: Time-shifting exposes true relationships23:30: Outcome of Sedai's three tier approach: only 3% of metrics need continuous tracking24:39: Q&A: Don't you lose hidden signals by narrowing what's tracked?To learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #sre #devops #aws #serverless #sedai #autocon22 #paloalonetworks #sase #tsahipeleg

Managing Kubernetes with current automated approaches makes it almost impossible to achieve optimal cost, performance and availability. Autonomous offers a path out of this complexity. Autonomous optimization and availability are outlined, the architecture used by Sedai to provide this capability, and how safety is addressed.Key moments:00:20: Why autonomous when we can automate everything?03:10: Automation for Kubernetes04:00: Elements to manage availability & optimization08:14: Constraints & variables to manage13:21: Complexity of k8s: 18,000 variables to optimize 100 microservices14:58: Autonomous Approach15:24: Autonomous vs Automation17:35: Core Tenets of an Autonomous System20:05: Autonomous Availability Management22:36: Autonomous Optimization23:43: A Better Way to Manage Kubernetes27:31: Sedai Architecture28:26: How does Sedai ensure safety before taking action? 29:30: How do you identify & populate metrics in Sedai?To learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #kubernetes #sre #devops #aws #a8s4k8s #sedai #autocon22

Cold starts are the #1 performance problem for Lambda users. Autonomous concurrency provides a solution with fewer cold starts than warmup strategies, and lower cost than provisioned concurrency. Sedai VP of ML, Nikhil Gopinath explains the underlying challenge of cold starts and presents the autonomous concurrency solution and compares it to traditional solutions.Key moments:00:49: Understanding concurrency and cold starts04:20: Cold starts by the numbers05:56: Concurrent requests and cold starts08:02: Provisioned concurrency09:41: Introducing autonomous concurrency11:53: Difference from warmup plugins14:39: Difference from provisioned concurrency16:25: Comparing Cost of Different Approaches18:00: How Autonomous Concurrency Works19:19: How Autonomous Concurrency Looks20:14: Component Details21:32: Q&A: Working alongside other optimizations22:17: Q&A: Interference with other extensions?To learn more about Sedai visit https://bit.ly/3e3Cqlv#coldstarts #lambda #autonomouscloud #sre #devops #aws #serverless #sedai #autocon22

Most serverless functions are unoptimized resulting in unnecessary latency and/or cost. Serverless functions can be optimized for performance, cost or a balanced strategy without human effort using autonomous systems. Nikhil Gopinath, VP ML at Sedai, outlines these challenges and the autonomous solution, and also reviews key results from fabric which cut latency by 48% using an autonomous system.Key moments:01:53: Optimization Challenges for Serverless03:36: Memory Settings09:36: Optimize for Performance & Cost11:18: The default for most Lambdas12:20: The best of both worlds13:06: Real time use cases15:47: Batch & backend use cases17:51: Challenges for backend optimization20:32: Under the hood of autonomous system22:24: How fabric reduced latency by 50%25:15: fabric's challenges for managing serverless27:02: fabric application architecture27:59: fabric uses 10,000 services30:11: Scaling up & down autonomously in production32:03: Optimizing for latency vs cost32:45: Matching supply to demand34:43: Will autonomous cause downtime?36:08: Does duration always go down when you incTo learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #lambda #sre #devops #aws #serverless #fabric #sedai #autocon22

Prakash Muppirala, EVP Platform at fabric, explains how fabric achieved a 48% reduction in latency and a 6.7x gain in their customers/SRE ratio by implementing an autonomous cloud platform. Prakash explains the importance of latency in ecommerce, fabric's unique challenges, and their path to autonomous cloud management with Sedai.Key chapters:01:31: Why latency is critical in e-commerce04:01: Need for trust and scalability for fabric's headless commerce service04:49: fabric's environment has 10,000 services, uses Lambda, ECS, EKS05:23: Key challenges at fabric09:11:Three key challenges: cost vs latency, matching supply to demand and availability12:05: Evaluating solutions from manual through autonomous13:38: Autonomous was faster, better, cheaper than manual or semi-automated solutions14:16: fabric's journey to autonomous15:48: QoQ growth in autonomous and improvements in latency, cost16:42: Scaling up & down autonomously17:50: Preventing availability issues18:57: Creating a feedback loop with developers19:37: Improving SRE productivity with a hybrid model (people and systems)20:25: Step-by-step phases to be 100% autonomous21:05: fabric's experience with autonomous22:17: Q&A: What was the hard ROI?24:09: Q&A: What is the role of the SRE with autonomous systems?To learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #lambda #kubernetes #eks #ecs #sre #devops #aws #serverless #a8s4k8s #sedai #autocon22

How can autonomous improve cost, availability and productivity? Experiences from RingCentral, Uber, Cato, fabric, Belcorp, Freshworks, and Norwest Venture Partners are brought together here in this panel discussion from autocon/22.Panelists:Kira Makagon, Chief Innovation Officer, RingCentralMatt Howard, General Partner, Norwest Venture PartnersSuresh Mathew, Founder & CEO, SedaiDean Nelson. CEO, Cato and Ex-Head Uber Compute (Moderator)Key moments:00:00: Why automation is no longer enough01:47: Autonomy negates the tradeoffs of the past between quality, speed and cost03:51: RingCentral's need to innovate drives Autonomous adoption05:56: Autonomous is helping RingCentral improve productivity, cost and reliability07:30: RingCentral's team saw opportunity to reduce toil08:06: How Autonomous impacts RingCentral Dev & SRE productivity09:22: Why Norwest has made multiple investments in autonomous11:51: Great technologies elevate people's careers15:04: The Enterprise Business Case for Autonomous17:52: fabric has 180,000 autonomous operations a year18:51: Belcorp saw 75% reduction in latency20:39: Autonomous improved fabric performance21:22: Humans can pass boring tasks to autonomous systems22:30: How does Sedai avoid errors?23:34: Microservice complexity is beyond human scale24:37: Augment people & help them scale25:38: RingCentral's POV on Performance29:40: Slow is the new down31:09: Progress from automation to autonomous32:17: Time to value is fast vs general purpose technologies35:38: RingCentral's POV on the full value of autonomous36:26: Do No Harm & Practical Autonomy37:25: Autonomous started in 2003, but adoption was slow38:40: Wrapup on the value of autonomyTo learn more about Sedai visit https://bit.ly/3e3Cqlv#autonomouscloud #lambda #kubernetes #eks #ecs #sre #devops #aws #serverless #a8s4k8s #sedai #autocon22, #kiramakagon, #ringcentral, #matthoward, #norwestventurepartners, #nvp, #deannelson, #cato