Containers Life Beyond Microservices Sushil Kumar Robin Systems
Containers: Life Beyond Microservices? Sushil Kumar Robin Systems
Containers and Application Types Stateful Stateless Data Apps Enterprise Apps Containers Benefit the Entire Application Spectrum Why Containers? › › Break out of dependency › Lightweight, high-performance “virtualization” hell › Multi-tenancy with bare-metal performance Build, ship, deploy › Higher consolidation density than VMs anywhere › Simplify ops by eliminating OS or VM sprawl
Containers & Data Apps: What is Missing Apps on Bare Metal (1995 -2005) A ❶ Need 10 -100 x more performance/scale from Storage › The whole point is to pack 10 -100 x more containers than VMs on your hardware B › Milliseconds response time to start/stop containers vs. minutes for VMs › Each app in each container can bind to multiple data volumes – HDD, SSD, Flash Traditional storage systems can’t handle this scale, deployment agility or diversity Apps on Virtualized HW (2005 -2015) A B C D Apps Containerized (2015 -2025) A B C D E F G H I J K L ❷ Need End-to-End performance isolation and QOS control › Containers only provide CPU and Memory isolation – what about IO and Storage? ❸ Data lifecycle management in multi-tenant infrastructure is hard › There are no DBAs to tune queries and write scripts for snapshots/clones
Robin Containerization Platform Software LXC CONTAINERS (Full OS , VM like manageability) DOCKER CONTAINERS node. js ROBIN APP-AWARE FABRIC CONTROLLER CONTAINERIZED VIRTUAL CLUSTER node. js ROBIN APP-AWARE FABRIC CONTROLLER ROBIN CONTAINER-AWARE, STORAGE AND DATA LAYER ROBIN APP-AWARE FABRIC CONTROLLER Containerized Agile Compute Cloud-extended Compute & Storage Container-aware Scale-out Storage
Born for Containers Scale-Out Storage LXC CONTAINERS (Full OS , VM like manageability) DOCKER CONTAINERS › › node. js ROBIN APP-AWARE FABRIC CONTROLLER CONTAINERIZED VIRTUAL CLUSTER Supports 100 s of thousands of variable-sized volumes Containerized Agile Sub-second creation and teardownnode. js of data volumes Compute Very fast creation and deletion of snapshots and clones Per volume data protection, encryption and compression ROBIN APP-AWARE FABRIC CONTROLLER ROBIN CONTAINER-AWARE, STORAGE AND DATA LAYER ROBIN APP-AWARE FABRIC CONTROLLER Cloud-extended Compute & Storage Container-aware Scale-out Storage
Decoupled Compute and Storage LXC CONTAINERS (Full OS , VM like manageability) DOCKER CONTAINERS node. js ROBIN APP-AWARE FABRIC CONTROLLER › › CONTAINERIZED VIRTUAL CLUSTER Scale Storage independent of Compute Container portability without data loss Cut storage by 2 x via resilientnode. js storage layer Containerized Agile Share data between clusters Compute ROBIN APP-AWARE FABRIC CONTROLLER Cloud-extended Compute & Storage ROBIN CONTAINER-AWARE, STORAGE AND DATA LAYER Container-aware Scale-out Storage
App-to-Spindle Performance Qo. S LXC CONTAINERS (Full OS , VM like manageability) DOCKER CONTAINERS node. js ROBIN APP-AWARE FABRIC CONTROLLER CONTAINERIZED VIRTUAL CLUSTER › Control resources & enforce priorities at each layer: Containerized Agile compute, network, storage Compute › Track IO from each container to all volumes node. js ROBIN APP-AWARE FABRIC CONTROLLER ROBIN CONTAINER-AWARE, STORAGE AND DATA LAYER ROBIN APP-AWARE FABRIC CONTROLLER Cloud-extended Compute & Storage Container-aware Scale-out Storage
App-to-Volume Data Management LXC CONTAINERS (Full OS , VM like manageability) DOCKER CONTAINERS node. js CONTAINERIZED VIRTUAL CLUSTER node. js › Application selects optimal media for each volume – SSD, or HDD Containerized › Containerized 1 -click Agilecluster Agileapplication Compute deployment, distributed Compute snapshots & clones ROBIN APP-AWARE FABRIC CONTROLLER › Automatic failover, data protection. Cloud-extended Compute & Storage ROBIN CONTAINER-AWARE, STORAGE AND DATA LAYER Container-aware Scale-out Storage
Why Robin? Developers IT › Easily deploy and manage multicontainer, full application stacks › Run containerized applications in production with VM-like manageability › Accelerate application & database clusters deployment, from days to minutes › Consolidate applications on bare metal to avoid hypervisor perf-overhead › Snapshot/clone entire application stack in a single click › Optimize data capacity & performance application-driven storage › Deliver Predictable performance CPU, IOPS, Network › Reduce VM sprawl & software licensing costs - deploy multiple applications per machine › Simplify root cause analysis – App to spindle views
Demo Life of Application
What would it take to…. . ❶ Deploy, failover, or move the entire application stack, including database, in a single click? ❷ Guarantee IOPS for each application? Cap IOPS of each container to avoid “noisy neighbor” situation? ❸ Time travel across application states for iterative testing?
Win Amazon Fire TV Stick with Voice Remote! Visit Our Booth #G 8
Thank you!
- Slides: 13