 Thank you, everybody, for the very warm welcome. Really appreciate that. All right, let's jump in, because I only have five minutes. This is my first time speaking on Observability Day. My name is Linsang. I come from Kerry, North Carolina. It's a very small town, so in case you guys don't know, I load a map on there. Like he said, I work at Solow. I'm the head of open source at a small company called Solow.io. And before I joined Solow, I worked at IBM. And on my last day at IBM, I took a screenshot of our corporate directory. I actually helped contribute to 207 patents to IBM. I wrote two books about Istio, Istio Explain, and Istio Ambient Explain. I'm also a CNCF ambassador, also one of the newest CNCF TOC member. I want to start to talk about ServiceMesh before we jump into observe the cost of ServiceMesh. So in 2017, the Istio project was launched with the sidecar approach, where every single service request would go through the sidecar to be able to provide automatically mutual TLS encryption upgrade, automatically traffic shifting, and observability for you. How many of you are familiar with ServiceMesh? Raise your hand. All right, many of you. So no need to explain this. Also back then in 2017, LinkedIn project launched the other notion of ServiceMesh without using sidecar, essentially using a node proxy. Now, about a little bit over a year and six months ago, Istio project launched this new thing called Ambient, which is the new data plane mode without needing to run sidecars. We've noticed there's a lot of user come to us and said, hey look, the sidecar is really complex to operate. Hey look, the sidecar is costing us 30% more than what I want to pay for CPU and memory. So we introduced a new mode to allow you to run ServiceMesh without sidecars. And Matt Klein, who is the founder of Envoy, which is the most popular proxy out there, particularly for sidecar, he actually agreed with our architecture. So how many of you know what a node agent is? In Kubernetes or on VM? So that's just a very common concept. It's agent running on the node. So in Istio Ambient, we introduced a component called ZilTrustTunnel, which essentially is a node agent. It's a shell for multiple apps under a particular node and can upgrade your connection to mutual tiers, can give your app cryptographic identity and upgrade the connection. So that's shell per node, which we call a secure overlay tunnel. How many of you know what a gateway is? All right. It's normally used to be able to control traffic into your cluster, for example, Kubernetes Gateway. So in Istio Ambient Mesh, we're bringing the concept of a gateway to inside of your Kubernetes cluster, where you would like to maybe apply traffic shifting or traffic routing, or maybe some rich layer seven authorization policy using the gateway, which is called Waypoint Proxy. All right. So what about cost? We just get to something related to observability today. So for this exercise, I'm setting up an environment using the application card online boutique app thanks to Google Cloud. It's a microservice application that has 12 microservices and it can generate loads automatically. And then for the exercise, I'm running using CNCF environment. I install Kubernetes. I have C-Advisor. I have Calico CNI. I have Grafana and Permissus installed. And for the exercise, I'm essentially running into four stages of testing. The first stage is, I'm just running the online boutique app without any psychos. And the second stage is using psychos. And the third stage is using the NodeAge and the Zotrast tunnel, which is a shared among multiple applications on that node. And the fourth stage is, what if I need to do some layer seven processing? So I install the Waypoint component we discussed earlier. So I'm going to reveal what I did. So essentially on the Grafana dashboard, I created my customer dashboard and I import that in. So I've shown you a snippet of that dashboard, but that is available online for you to download if you want to repeat what I've done. And this is the Grafana dashboard looks like for the four stages we talk about. So what you can see really, really interesting is on the right side, let's see, I'm not sure if it's the right side on you, but it's the right side for me. It's the memory. And then on the left side, it's the CPU. So as you can see throughout the four stages, particularly for memory, because typically when you go use cloud, you tend to pay based on memory usage. You can see on the sidecar is costing about 11 gigabytes of RAM. Well, the other stages is costing only about four gigabytes of RAM, somewhere between four to five gigabytes. So that's a big jump for the second stage, which is using sidecar and costs a lot more. And this is the sidecar and CPU memory stacked, as you can see the big jump also. By the way, this is only showing the three stage because the first stage wasn't in service much. So it doesn't have any sidecar or is still ambient cost. As you can see, again, the stage for memory, particularly for the sidecar, is costing a whole lot more. We also did a little bit of analysis with Google Cloud. So basically Google Cloud has a cloud-provising calculator, and then we run a test with 150 services and parts. And you can see the baseline is the one without service mesh, which pretty much costs the same as ambient without any layer seven processing. So essentially you could potentially run is your ambient without sidecar and get mutual TRS cryptographic identity, automatic certificate rotation for free with the seminal cost. Now, in summary, ambient introduce a new sidecarless data plane mode so you can run your application in the mesh without dragging the sidecar. And by leveraging Grifana permissives, the NodeExport and CAdvisor, we can compile CPU and memory in the dashboard and have a visualization of how the usage goes. The cost to get mutual TRS with ambient without changing your application, without restarting your deployment is pretty much free. And the cost to get mutual TRS with some of the layer seven functions is about 18% more for ambient. And if you need to use sidecar, it's about 36% more. So that wraps up my talk. Do I have any time for questions? All right, thank you all so much.