Containerized Storage

20
CONTAINERIZED STORAGE Kapil Jain, Verizon Labs Narendra Narang, Red Hat With special thanks to: Subhajit Mukherjee and Rehan Tahir at Verizon Labs Jeff Darcy, Ben England, Shyam Ranganathan and Sayandeb Saha at Red Hat Tuesday, January 19 th 2016

Transcript of Containerized Storage

Page 1: Containerized Storage

CONTAINERIZED STORAGE

Kapil Jain, Verizon Labs Narendra Narang, Red Hat

With special thanks to:Subhajit Mukherjee and Rehan Tahir at Verizon LabsJeff Darcy, Ben England, Shyam Ranganathan and Sayandeb Saha at Red Hat

Tuesday, January 19th 2016

Page 2: Containerized Storage

Brief Bio: Narendra N. Narang - Sr. Cloud Storage Solutions Architect at Red

Hat - At Bloomberg for 11 years, where primary

focus was on storage and emerging technologies

- 17 years experience working in various information technology roles at New York financial institutions.

Page 3: Containerized Storage

Brief Bio: Kapil Jain - Verizon Cloud platform team (storage) - Verizon Terremark - Verizon labs - Dell Equallogic SAN management team - Progress Relational Database SQL Engine Team

Page 4: Containerized Storage

Agenda - Strategic Business initiatives - Technology Stack - FSaaS - Migrating to the cloud - Architecture - State of the Union - Business benefits - Proposed futures - The Red Hat Engagement

Page 5: Containerized Storage

Verizon Labs- Center of excellence and

innovation - Platform for Internet Services.- Charter for the team

Page 6: Containerized Storage

Strategic Business Initiatives - Cloud platform for internal applications

created on commodity hardware, open-source software

- Resilient, scalable and highly available platform.

- Services for enterprise applications - Migration path for existing applications - Achieve substantial cost savings over

existing alternatives

Page 7: Containerized Storage

Key Technologies & Vendors

- CoreOS – open-source lightweight OS ideal for container environment

- EMC – Elastic Cloud Storage (ECS) for the object store

- MESOSPHERE – mesos orchestration tool for deploying apps in a cluster-computing platform

- Red Hat – elastic, distributed, replicated , shared file system.

Page 8: Containerized Storage

FSAAS- Software Defined Storage (SDS)- Containerized micro-services.- Highly Available- Disaster tolerance- High Performance- Support Structure

Page 9: Containerized Storage

Performance & Costs Scaling of SDS Architecture

Number of Storage Nodes

Rea

ds /

Writ

es T

hrou

ghpu

t (m

Bps

)

Software

Defi

ned S

cale-

out S

torag

e

( Glus

terFS)

Traditional Enterprise NAS Storage

Re-architecture of high performance NFS to scale-out achieves lower costs

Number of Storage Nodes

Tota

l Sto

rage

Cos

ts ($

)Trad

itiona

l Ente

rprise

NAS S

torag

e

Software DefinedScale-out Storage( GlusterFS)

Storage Performance Scalability Storage Costs Scalability

Page 10: Containerized Storage

Migrating to the CloudMigrating from a traditional NAS environment to a SDS environment

- Characterization and segmentation of different NAS workloads on based on performance requirements

- Mapping those workloads to software-defined-storage ( GlusterFS)

- Comparing and validating the reliability of Red Hat Gluster Storage running on commodity off-the-shelf hardware with a traditional enterprise class NAS appliance

- Aligning dynamic scaling of the persistent/storage layer (shared filesystem) with the dynamic scaling requirements of the application workload.

To meet our objectives and overcome these challenges, we engaged Red Hat.

Page 11: Containerized Storage

ETH/IB

app app app

Converged Computing Architecture: Mixed App and Storage Compute Workloads

App-only servers

Converged servers

app

app +

Storage-only servers

Storage stack imposes only3% - 10% load on compute processing

Applications and storage stacks can co-exist on same compute substrate. This achieves higher server utilization and lower operational costs across the cloud

Page 12: Containerized Storage

Architecture

Containerized Docker containers running on commoditized converged hardware

Page 13: Containerized Storage

State of the Union - Simplified Deployment via containers - Synchronous replication for high availability - Geo-Replicated for disaster recovery - Seamless upgrade and rollback - Dynamic scaling (up) of capacity as needed - Performance and health metrics collection via

container monitoring agents and sending alerts.

Page 14: Containerized Storage

Examples of Usage and Scale.Volume Media Bricks Size Geo -

ReplicatedFile Store HDD 28 112 TB

and growing

Yes

Cache SSD 20 8 TB No

Geo-Rep Vehicle

HDD 12 48 TB Yes

Page 15: Containerized Storage

Implementation Challenges- DNS to resolve Gluster container hostnames

- Persistence of Gluster metadata

- Gluster FUSE client optimization

- Geo-replication performance

Page 16: Containerized Storage

Solved Challenges• New Technology - Trailblazing issues• Provisioning resources based on demand, not

projections• Unproven Software, production issues• Storage in docker environment is relatively

unsolved• Deploying and orchestration • Automated Seamless upgrade.

Page 17: Containerized Storage

BUSINESS BENEFITS

- Leveraging stable proven technology in a new environment

- Higher Resource utilization, reducing costs and capex requirements

- Increase speed and reliability of product delivery, reducing time to market and increase speed of innovation

- Reduce cost and reliance on specialized hardware by running on commoditized hardware with open source software

- For disaster recovery, we have reduced the time to access backups from a couple of days to a couple of hours

- Reduced dependence on external clouds.

Page 18: Containerized Storage

Proposed Futures- Data migration - Better Management of resources and gluster

volumes – Heketi ???

Page 19: Containerized Storage

Engagement with Red Hat- Weekly Calls- Responsive and meaningful interactions- Developer Support - TAM engagement- Training and certification

Page 20: Containerized Storage

Q&A