Rakuten Symphony / RobinIO - IT Press Tour #48 Jan. 2023

Transcript

1. Symworld Cloud: The Distributed Stateful Edge Platform Accelerate your journey towards the cloud-native future Partha Seetala President, Cloud BU, Rakuten Symphony (Formerly Founder & CEO, Robin.io (acquired by Rakuten Symphony in April 2022))
2. 2 Why Distributed Stateful Edge? High volume of data generated at 10,000s of geographically dispersed locations Why is Distributed Stateful Edge needed? If data is not filtered, processed and acted on closest to where it is generated, then it results in: • Poor user experience • Increased cost incurred by transferring data over the network • Poor end-to-end performance
3. 3 Challenges managing Distributed Stateful Edge What are the challenges? • How to manage the infrastructure at 10,000s of locations (installing, upgrading, OS, Platform software)? • How to deploy, heal and manage applications on all these servers? • How to get a birds-eye view into the health, and performance of infrastructure and applications at all these locations? • How to ensure security at this scale? • How to contain high energy costs? High volume of data generated at 10,000s of geographically dispersed locations
4. 4 4 The Distributed Stateful Edge Cloud serving Stateful, 5G and Edge applications with highest performance and at scale Symworld Cloud Data-intensive applications App-aware Storage end Data management for Kubernetes 5G RAN, Core, Edge, Private 5G Cloud-native deployment of RAN, Core, Edge apps Metal-to-Service automation Manage baremetal servers, network elements and network services at scale
5. 5 Symworld Cloud Offers 3 Products Fully-integrated enhanced-Kubernetes platform, optimized for running Storage- and Network- intensive services with special focus on zero-touch deployment and industry leading operational automation at the Edge Symworld CNP Cloud Native Platform 1 Highest performing cloud-native storage stack for Kubernetes with special focus on storage and application-aware data management, where volumes + applications can be manipulated as a whole, as one single logical group. Symworld CNS Cloud Native Storage 2 Highly scalable Infrastructure and Service/Application orchestrator to manage baremetal servers, and Applications across 100K+ servers and 10K+ clusters, datacenters Symworld Orchestrator Baremetal and Service Automation 3
6. PRODUCT 1 Cloud Native Platform (CNP) Fully-integrated Kubernetes platform optimized for running Storage- and Network- intensive services with special focus on zero-touch deployment and operational automation at scale at the Edge
7. 7 Symworld Cloud Native Platform Architecture Enhanced Networking App-aware Storage Application Workflow Manager Kubernetes Advanced Scheduler Observability Monitoring Advanced Declarative Workload Placement • Policy-based modeling with no hunting or hardcoding • Highly granular NUMA-awareness with Affinity + Anti-affinity, CPU pinning, GPU slicing,HugePages, networking and much more • No K8S expertise needed to operate -- intuitive, easy to use GUI (low-code/no- code) approach Advanced networking • Persistent IPs & Multiple IPs • Overlays/underlays: Calico, OVS, VLAN, SR-IOV, DPDK • Multiple NICs • Dual-stack IPv4/IPv6 • Network policies High-performance App- aware Storage • Automates Snapshots, Clones, QoS, Replication, Backup, Data rebalancing, Tiering, Thin- provisioning, Encryption, Compression • Ties resources and roles that make up your databases and pipelines, as well as your application configuration, networking, meta-data and lifecycle states Sherlock Advanced Visibility & Diagnostics • Resource dependency analysis • Multi-domain failure analysis • Impacts of cluster, networking, server, storage etc. Runs on Baremetal & VM (on-premises and in the cloud) Cloud Native Platform (CNP)
8. 8 Use Cases served by CNP Deploy Stateful Apps on Kubernetes • Highest performing storage and data management for any Kubernetes on-premises or in any Cloud • Automates all aspects of critical day-2 operations to run Databases, Big Data, Timeseries and Message queue services in production on Kubernetes Deploy and Manage Apps at the Edge • Deploy complex storage and network-intensive apps to Edge clusters with ease • Scalable automation of rollout, upgrade, and auto heal apps at Edge locations Common platform to run Containers and VMs • Run Containers and Virtual Machines on Kubernetes directly on Baremetal • Removes operations and resource silos – consistent and common lifecycle management of Containers and Virtual Machines Common Platform for Hybrid & Multi-Cloud • Consistent storage experience across different environments • Easy migration and portability of apps along with their data across different cloud environments Automate day-2 operations • Bring cloud agility to your baremetal infrastructure • Centrally manage 10s of thousands of Kubernetes clusters from edge to cloud to private data centers • Create repeatable and robust workflows to automate all Network and IT operations Cloud Native Platform (CNP) Deploy 4G/5G RAN, Core and NFV at Scale • Bring cloud agility to deploy 4G and 5G RAN (DU, CU-CP and CU-DP) on COTS hardware in minutes • Centralized orchestration of 10s of thousands of cell-sites and base- band units
9. PRODUCT 2 Cloud Native Storage (CNS) Highest performing cloud-native storage stack for Kubernetes optimized for storage and application-aware data management
10. 10 Cloud-Native Storage • Lightweight operator installs on any Kubernetes • Discovers and pools Local Disks (SSD, HDD, NVMe), Cloud Disks, SAN High Availability • Strictly consistent Replication • Auto resync on fall-behind • Fast Failover with split-brain avoidance Advanced Policies • Compression, Encryption Compute + Storage Locality, Affinity, Anti-affinity Placement policies • Node, Rack, Datacenter, Zone-aware replication Symworld CNS: Day 0 Capabilities Robust data persistency for Stateful Apps Always-on Availability with self-healing Advanced data placement for Stateful apps Cloud Native Storage (CNS)
11. 11 Symworld CNS: Day 1+ Capabilities Predictable & Stable Performance Elastically scale capacity over time Observability to enhance maintainability Data management to ensure business continuity Capacity Management • Add/Remove disks to increase/decrease capacity • Grow and shrink volume size • Replace faulty disks Observability • Map storage allocations from App->Nodes->Disks • Get detailed storage metrics Performance • Baremetal performance (2-6x faster than others) • Live data rebalancing to avoid IO hotspots • Cap IOPs to avoid noisy neighbor challenges Data Management • Snapshot, Rollback • Backup, Restore • Business Continuity and App Portability Cloud Native Storage (CNS)
12. 12 Symworld CNS: Data Management Capabilities CNS Kubernetes app2-snap2 app1-snap2 app1-snap1 APP 1 APP 2 snapshot rollback Snapshots CNS Kubernetes app2-snap2 app1-snap2 app1-snap1 APP 1 APP 2 Off-cluster Local Backup Target Off-site Remote/Cloud Backup Target backup restore snapshot Backups CNS PRODUCTION app1-snap2 K8S Cluster 1 app1-snap1 APP 1 snapshot UAT APP1 Clone1 clone K8S Cluster 2 CNS TEST/DEV APP1 Clone2 clone Clones CLOUD OBJECT STORE (S3, GCS, AzureBlob, MinIO) On-prem Snapshot 1 3 months ago Snapshot 2 3 days ago Snapshot 3 yesterday CNS APP GCP GKE CNS APP AWS EKS CNS APP Kubernetes Multi-Cloud Portability Simplified Operations 1-click via GUI Single CLI command Triggered via API Ease of operations Cloud Native Storage (CNS)
13. PRODUCT 3 Symworld Orchestrator Large scale automation and orchestration platform of Infrastructure resources and Network Services
14. 14 Symworld Orchestrator Symworld Orchestrator 10’s of SymWorld Cloud Clusters 10,000’s of SymWorld Cloud Clusters 1,000s of Symworld Cloud Clusters OSS Network Function-as-a-Service (Deploy, Heal, Monitor, Scale, Upgrade) Edge 1 Edge 3 Edge N BBU 1 BBU N Edge 2 NF1 NF4 NF2 BareMetal-as-a-Service (BIOS, OS, FPGA, Robin K8S Install, Monitor, Upgrade) NEAR EDGE FAR EDGE CENTRAL Symworld Orchestrator MOPs Manager (Highly scalable and extensible MOPs Management across 100s of thousands of Network Elements) Centralized Management of 10,000s of RAN, Edge & Core clusters Policy based Closed-loop Automation Define & Execute Complex DAG based Workflows across RAN, Edge, Core NF3
15. 15 Symworld Cloud: The Distributed Stateful Edge Platform • Highly-optimized Kubernetes platform for running Storage- and Network- intensive applications: 1. Built-in advanced storage and data management stack 2. Built-in advanced networking 3. Built-in advanced compute (CPU + GPU) for low-latency/high-throughput applications 4. Built-in advanced scheduling and observability • One common platform to run apps as both CNFs (in containers) and VNFs (in virtual machines) • Automation and Orchestration at 3 layers: 1. Infrastructure (baremetal servers) 2. Platform (K8S) 3. Application Services (CNFs, VNFs) • Centralized infrastructure-to-application observability and policy-based closed-loop operations