Mellanox provides high performance networking solutions for Ceph storage clusters. They discussed how Ceph relies on high performance networks for scalability and availability. Mellanox offers end-to-end 40/56GbE and InfiniBand solutions with full CPU offloading. They presented examples of how customers deploy Ceph with Mellanox's 40GbE interconnects across cluster, client, and public networks. Mellanox also discussed ongoing work to integrate RDMA support into Ceph to further improve performance.

1. Mellanox High Performance Networks for Ceph
Building World Class Data Centers
Ceph Day, October 22nd, 2014

2. Agenda
 Who are we and what does it have to do with CEPH?
 Benefits of High Performance Interconnect for CEPH
 Typical deployment scenarios
 Advanced performance work in progress
 Examples of Commercial Solutions
© 2014 Mellanox Technologies - Mellanox Confidential - 2

3. Leading Supplier of End-to-End Interconnect Solutions
Server / Compute Switch / Gateway
Virtual Protocol Interconnect
Storage
Front / Back-End
Virtual Protocol Interconnect
56G IB & FCoIB 56G InfiniBand
10/40/56GbE & FCoE 10/40/56GbE
Comprehensive End-to-End InfiniBand and Ethernet Portfolio
ICs Adapter Cards Switches/Gateways Host/Fabric Software Metro / WAN
Cables/Modules
© 2014 Mellanox Technologies - Mellanox Confidential - 3

4. The Future Depends on Fastest Interconnects
1Gb/s 10Gb/s 40/56Gb/s
© 2014 Mellanox Technologies - Mellanox Confidential - 4

5. From Scale-Up to Scale-Out Architecture
 Only way to support storage capacity growth in a cost-effective manner
 We have seen this transition on the compute side in HPC in the early 2000s
 Scaling performance linearly requires “seamless connectivity” (ie lossless, high bw, low latency,
cpu offloads)
Interconnect Capabilities Determine Scale Out Performance
© 2014 Mellanox Technologies - Mellanox Confidential - 5

6. CEPH and Networks
 High performance networks enable maximum cluster availability
• Clients, OSD, Monitors and Metadata servers communicate over multiple network layers
• Real-time requirements for heartbeat, replication, recovery and re-balancing
 Cluster (“backend”) network performance dictates cluster’s performance and scalability
• “Network load between Ceph OSD Daemons easily dwarfs the network load between Ceph Clients
and the Ceph Storage Cluster” (Ceph Documentation)
© 2014 Mellanox Technologies - Mellanox Confidential - 6

7. How Customers Deploy CEPH with Mellanox Interconnect
 Building Scalable, Performing Storage Solutions
• Cluster network @ 40Gb Ethernet
• Clients @ 10G/40Gb Ethernet
 Directly connect over 500 Client Nodes
• Target Retail Cost: US$350/1TB
 Scale Out Customers Use SSDs
• For OSDs and Journals
8.5PB System Currently Being Deployed
© 2014 Mellanox Technologies - Mellanox Confidential - 7

8. CEPH Deployment Using 10GbE and 40GbE
 Cluster (Private) Network @ 40GbE
• Smooth HA, unblocked heartbeats, efficient data balancing
 Throughput Clients @ 40GbE
• Guaranties line rate for high ingress/egress clients
 IOPs Clients @ 10GbE / 40GbE
• 100K+ IOPs/Client @4K blocks
Admin Node
Throughput Testing results based on fio benchmark, 8m block, 20GB file,128 parallel jobs, RBD Kernel Driver with Linux Kernel 3.13.3 RHEL 6.3, Ceph 0.72.2
IOPs Testing results based on fio benchmark, 4k block, 20GB file,128 parallel jobs, RBD Kernel Driver with Linux Kernel 3.13.3 RHEL 6.3, Ceph 0.72.2
Client Nodes
10GbE/40GbE
Public Network
10GbE/40GBE
Cluster Network
40GbE
20x Higher Throughput , 4x Higher IOPs with 40Gb Ethernet Clients!
(http://www.mellanox.com/related-docs/whitepapers/WP_Deploying_Ceph_over_High_Performance_Networks.pdf)
Ceph Nodes
(Monitors, OSDs, MDS)
© 2014 Mellanox Technologies - Mellanox Confidential - 8

9. I/O Offload Frees Up CPU for Application Processing
Without RDMA With RDMA and Offload
~88% CPU
Efficiency
System Space User Space
~53% CPU
Efficiency
~47% CPU
Overhead/Idle
~12% CPU
Overhead/Idle
System Space User Space
© 2014 Mellanox Technologies - Mellanox Confidential - 9

10. CEPH+RDMA: Coming Soon
 Phase 1: RDMA (multi-transport) integration of XioMessenger based
on Accelio
• Basic integration complete, in up-stream contribution process
• http://wiki.ceph.com/Planning/Blueprints/Hammer/Accelio_RDMA_
Messenger
 Eliminating XioMessenger bottleneck uncovered several others…
 Increased focus on performance, and weekly community calls
involving RedHat, Cisco, Fujitsu and others:
• http://pad.ceph.com/p/performance_weekly
Accelio Native simple
messenger
Xio messenger Ceph
TCP
Ceph
RDMA
Expected
Ramdisk local
perf
BW Read (MB/s) 3200* 980** 3200* 950 1300 2534
BW Write (MB/s) 3200* 980** 3200* 435 500 2361
* Measured on gen2 PCIE system, expected 5000 in gen3 systems
** Increased pipe depth to 10000 and 1000000
© 2014 Mellanox Technologies - Mellanox Confidential - 10

11. Commercial Solutions in the Market Today
Healthcare
 ISS: Storage Super Core
• 82,000 IOPS on 512B reads
• 74,000 IOPS on 4KB reads
• 1.1 GB/sec on 256KB reads.
 http://www.iss-integration.com/supercore.html
Cloud
 OnyxCCS: ElectraStack
 Turnkey IaaS multi-tenant computing system
 5X Faster Node/Data Restoration
 https://www.onyxccs.com/products/8-series
High Performance Storage
 Scalable Informatics: Unison
 Archive like storage density with tier 1
performance
• 60 drives in a 4U chassis all driven at full bandwidth
 https://www.scalableinformatics.com/unison.html
Media / Entertainment
 Daystrom/StorageFoundry: Nautilus
 Massive-scale storage platform
• Single client reaching up to 2.2GB/sec on read throughput.
 Easy integration with existing FC storage
 http://storagefoundry.net/
Generally Available, Pre-configured, Pre-tuned Solutions leveraging 40/56Gb/s Cluster Networks
© 2014 Mellanox Technologies - Mellanox Confidential - 11

12. Summary
 CEPH cluster scalability and availability rely on high performance networks
 End to end 40/56 Gb/s transport with full CPU offloads available and being deployed
• 100Gb/s around the corner
 RDMA and more performance improvements coming soon
© 2014 Mellanox Technologies - Mellanox Confidential - 12

13. Thank You