Red Hat Gluster Storage, Container Storage & CephFS Plans
Presenter: Sayan SahaDate: 19-Jan-2015
Presentation Outline
▪RHGS recap, intro and state of the union▪Looking forward (Core RHGS, 6 – 8 months)▪Persistent Storage for Containers (RHxS)▪Hyper-Converged solution with RHEV▪Public Cloud (AWS, Azure)▪Red Hat Storage Life Cycle Manager▪CephFS Plans▪Q&A
2
3
The Past: Quick Recap
• VM image store, performance & stability• 6 updates released • End of Life June 2014
Red Hat Gluster
Storage 2.0
• Quota, Geo-Replication, UI, SMB 2.0• 6 updates released• End of Life Oct 2015
Red Hat Gluster
Storage 2.1
• Snapshots, Nagios, SNMP, Rolling Upgrade, RDMA, 3-way replication, small file performance enhancements, SELinux, SSL encryption
Red Hat Gluster
Storage 3+
4
The Present: RHGS 3.1 – Shipping Now
Red Hat Confidential5
The big 3 (fully supported)…
● Erasure Coding○ Data protection without using RAID & replication
● Bit Rot Detection○ Protection against “Silent Data Corruption”
● Active-Active NFSv4○ Secure, scalable, performant, table stakes
Robust Enterprise Class SDS Capabilities
RHGS - Erasure Coding
▪ Data protection without using RAID & replication
▪ Break data into smaller fragments, store and recover from a smaller number of fragments
▪ New type of volumes: Dispersed, dist-Dispersed
▪ Algorithm used is REED-Solomon ▪ Initial tested configurations: 8+3,
8+4 & 4+2
= Overall capacity reduction= Faster rebuild times
Red Hat Confidential6
RHGS - Bit Rot Detection
▪ Protection against “Silent Data Corruption”
▪ Two fundamental procedures –Signing using SHA256–Scanning/scrubbing for rot
▪ Lazy checksum maintenance (not inline to data path)
▪ Checksum calculation undertaken when a file is considered “stable”.
▪ Bit-rot scanning mode is admin selectable to control performance impact
Red Hat Confidential7
RHGS - Multi-headed NFSv4 (Active-Active)
▪ NFSv4 ACLs▪ Security
– Kerberos authentication using RPCSEC_GSS, krb5/i/p, spkm3
▪ Active/Active cluster-on-cluster– Up to 16 A/A NFS heads
▪ RHGS pool scales-out as usual▪ Ability to add & delete RHGS
volume exports to nfs-ganesha at run-time
Red Hat Confidential8
RHGS – Tiering (tech preview)
▪ Automated data movement between hot & cold tiers
▪ Movement based on access frequency
–Hot tiers could be SSDs, cold tiers are normal disks
▪ Attach & detach a tier to and from an existing Gluster volume
–Initial I/Os forwarded to hot tier
–I/O misses promotes data to hot tier
Red Hat Confidential9
Looking Forward...
1010
STORAGE TRENDS:MODERN IT INFRASTRUCTURES
11
Traditional Storage Next Generation Storage
Manual provisioning of LUNs and volumes with some degree of automation
Self-service provisioning by lines ofbusinesses and application developersStatic selection
Static selection of storage platforms basedon application needs
Catalog based storage service offerings withmetering & charge-back
Scale-up with some scale-out. Costly migrations.
Expand, Shrink and scale on demand. Easier upgrade.
Little to no flexibility in selecting optimum storage back-end for workloads
Policy based storage back-end selection
Key Trends’ Impact on Gluster
▪Consumption Model– API based dynamic provisioning, healing, tuning & balancing
– Secure multi-tenancy– Cloud scale & stability at scale
▪Performance: performant storage back-end for a wide variety of workloads▪Advanced data services: tiering, compression, de-
duplication
12
RHGS - Core Roadmap
Red Hat Confidential13
Feb’ 16 April/May’ 16 Q4 CY 2016
3.1.2
Core- Tiering full support- Writable Snapshots- SMB Perf
Enhancements- Bit Rot StatusMgmt- Dynamic Volume
allocation (Preview)- RHGS-C offline install- Docker image support
3.1.3
Core - Sharded volume support
for VM image store including hyper-convergence
- Arbiter Quorum- Refresh Swift APIs to
OpenStack KiloProtocols- SMB 3.0 multi-channel
supportPerf• Multi-threaded self-heal
3.2
Core- Subdirectory level
exports with FUSE (multi-tenancy)
Protocols- NFSv4 delegations
Mgmt- REST APIs- New Storage Life
Cycle Manager
Roadmap Feature Details
14
3.1.2 Tiering HSM like, tiered volumes, promote-demote based on access frequency
Writable Snapshots Create a share from a snapshot
Dynamic Volume Provisioning Allocate Gluster volumes programmatically and dynamically
SMB performance enhancements Using Async I/O and other enhancements
Bit Rot Status Interactive CLI listing impacted files
RHGS-C offline install Provide OVA image
Docker Image Support Official docker image in Red Hat’s Container Registry
3.1.3 Sharding Sharded-replicated volumes for hyper-converged VM storage
Arbiter Quorum Reliable quorum without 3-way replication penalty
SMB 3.0 multi-channel Network fault tolerance & performance
RHS Portfolio Integration with OpenShift Enterprise
1515
RHGS + RHCS & Containers - Phase 1
▪ Persistent data store for containers in independent compute and storage clusters
▪ Fully support RHGS & RHCS as a storage backend for OSE & AEP
▪ Available Now!
16
RHGS & Containers - Phase 2
▪ Run containerized storage + containerized compute hyper-converged from same set of hosts
▪ Use Kubernetes to provide unified control plane for compute and storage
▪ Availability with OSE 3.2, 3.3
▪ Rinse & Repeat with RHCS
17
Hyper-Converged with Containers
18
Hyper-Converged RHGS Solution with RHEV
1919
RHGS - Hyper-Converged ROBO Solution
▪ A planned low footprint storage/compute offering, currently under development, integrating RHEL, RHGS, & RHEV-M
▪ Simplified acquisition, deployment, and management
▪ Support planned for wide range of workloads
▪ Currently in customer Pilot programs
▪ Post RHEV 3.6 GA
20
Remote Replication
Red Hat Gluster in Public Cloud
21
Public Cloud Availability & Plans
▪RHGS available via Cloud Access in AWS & Azure–BYOS model. Fully supported by RHT.–Allows customers to use RHGS in the Cloud (existing use-case or net new)
–Lift & Shift existing applications to Cloud without re-writing the app (POSIX compatible file store)
▪Exploring Cloud marketplaces and on-demand pricing for cloud consumers
22
Red Hat Storage Life Cycle Manager
23
Life Cycle Manager for Red Hat Storage
Deploy, Config, Manage, Monitor
– Consistent UX across storage portfolio
– Streamlined workflows– Integration / leverage
other RHT tools & projects– Integration with manager
of managers like RHT CloudForms
Architecture
CORE
Business logic, etc.
UI
REST API
AGENT
Runs on storage nodes
PROVIDER
Per-SDS driver / plugin
INSTALL
Mariner
Components
Product Profile...
26
▪ An independent product with it’s own maint stream and support lifecycle (3yr)
– distinctly versioned from storage products it manages
▪ Included with existing SKUs (entitlements)
– Existing SKUs of the storage providers will be modified to include
▪ Conceptually a “layered product” on top of the storage providers
▪ OS Support: RHEL7 mgmt station, Ubuntu Ceph nodes (agents)
CephFS Plans
27
Use Case: OpenStack Manila Backend (Phase 1)
▪CephFS will be tech preview in RHCS 2.0. FUSE client will ship then. ▪Kernel client in RHEL
7.3▪Ship native driver with
RHELOSP 9 for Manila▪Tenant VMs directly
access the storage network
28
Secure Plumbing (Phase 2)
▪VMware vsockets designed for VM to host communication ▪Mount CephFS on host
+ kNFS or run NFS-Ganesha ▪Export to VM’s VSOCK
address▪Tenant does not access
storage net
29
Thanks. Q&A.
Red Hat Confidential30