Call a Specialist Today! 844-356-5142

Pavilion Hyperparallel Flash Array

The Power of Parallelism

By disaggregating storage from compute, Pavilion lets you compose volumes of NVMe SSDs on demand, without the performance bottlenecks of traditional dual-controller arrays.

Pavilion Hyperparallel Flash Array

 

Performance and Density @ the Lowest Latency
Built like a network switch, Pavilion separates and aggregates network and CPU resources for unprecedented composability.

 

 

Pavilion Hyperparallel Flash Array Blazing Performance

Blazing Performance

90GB/sec read and write throughput, 20M IOPS, 40μsec latency

Pavilion Hyperparallel Flash Array Modular 4RU Design

Modular 4RU Design

Up to 20 active-active controllers with 40 @ 100Gbps Ethernet or InfiniBand ports, 72 standard 2.5” NVMe SSDs

Pavilion Hyperparallel Flash Array 100% Standards Compliant

High Availability

Redundant and hot-swappable components assure no single point of failure

High Availability

100% Standards Compliant

x86 processors, no FPGAs or custom ASICs. NVMe-oF/TCP & NVMe-oF/ RDMA, NFS, iSCSI, Ethernet and InfiniBand

 

Pavilion Hyperparallel Flash Array Hardware Top Down Standard x86 components

Our Hardware

Standard x86 components, including off the shelf processors and SSDs. No custom ASICs or FPGAs. Rapid innovation leveraging the latest technologies leads to unparalleled density and performance.

 

Pavilion Hyperparallel Flash Array Standard NVMe SSDs

Standard NVMe SSDs

Up to 72 2.5” NVMe SSDs configurable in RAID 6 volumes. OpenChoice Storage™ Storage™ Storage™ with support for Intel® Optane™ TLC and SCM NAND.

 

Pavilion Hyperparallel Flash Array Line Cards

Active-Active Line Cards

Up to 10 Line Cards. Each has 4 @ 100Gbps network ports, 2 x86 Active-Active Storage Processors with 4,096 NVMe queues/controller. That’s 81,920 NVMe queues per array.

 

Pavilion Hyperparallel Flash Array Dual Management Controllers

Dual Management Controllers

Sophisticated GUI, CLI, and APIs for integration with V-Center, Kubernetes CSI, RedFish™ and SwordFish™.

 

Pavilion Hyperparallel Flash Array Redundant-Fabric

Redundant Fabric Modules

Dual, hot-swappable 1.6 Terabit fabrics form the core of storage array designed like a network switch.

 

Pavilion Hyperparallel Flash Array Modular Cooling

Modular Power & Cooling

Redundant, hot-swappable power and cooling modules. Power up and keep cool as you scale.

 

 

 

Over 16 Million IOPS, less than 1-microsecond network latency, 0 packet drops. Cisco Nexus 9000 DC switch with in-house ‘cloud scale’ ASIC and Pavilion NVMe-oF/RoCEv2 storage. With Pavilion now supporting 15+ TB SSDs, and Intel® Optane™, this array is a game-changer.

– Kamal Bakshi

DC Technology Evangelist, Cisco Systems

 

 

Pavilion Hyperparallel Flash Array

Accelerate legacy and modern data applications

The Pavilion Hyperparallel Flash Array (HFA) is the world’s first NVMe-oF Storage Array. It is an end-to-end NVMe solution, from the host to the media. The array is 100% standards compliant with zero host-side presence, contains no ASICs or FPGAs. It leverages off the shelf processors and SSDs, and is designed to accelerate legacy and modern rack-scale web and analytics applications.

  • Industry’s First Hyperparallel Flash Array
  • Up to 20 Storage Controllers and Parallel Architecture Boost Performance
  • 40 Ethernet or InfiniBand Ports for Flexible Connectivity
  • 90 GB/s R/W Throughput, 40 µs Latency, and 20M 4K Random Read IOPS*
  • 4RU Chasis
  • Active/Active Data Access
  • No Single Point of Failure
  • End-to-End NVMe with simultaneous NVMe/Ethernet, NVMe/InfiniBand, and NVMe/TCP protocols
  • Support of iSCSI and NFS for legacy applications

End-to_end NVMe

  • Up to 72 U.2 2.5" NVMe NAND and SCM SSDs
  • OpenChoice StorageTM
  • NVMe-oF/RDMA (Ethernet and InfiniBand)
  • NVMe-oF/TCP
  • iSCIS and NFS for legacy applications
  • Standard drivers

OpenChoice StorageTM

Pavilion OpenChoice Storage disrupts conventual selling practices and avoids vendor lock-in by providing customers the freedom of using their own SSD media in the Pavilion HFA.

  • Up to 72 U.2 2.5" NVMe NAND and SCM SSDs
  • OpenChoice StorageTM
  • NVMe-oF/RDMA (Ethernet and InfiniBand)
  • NVMe-oF/TCP
  • iSCIS and NFS for legacy applications
  • Standard drivers

Features:

  • Simultaneous NVMe/Ethernet, NVMe/TCP, NVMe/InfiniBand, NFS v3/v4, and iSCSI With Standard Drivers
  • Up to 40 100Gb Ethernet and InfiniBand Ports

Flexible Management

  • Web GUI and CLI
  • vCenter, Kubernetes, and OpenStack
  • RESTful API, DTMF/Redfish, and Swordfish

Endurance & Reliability

  • Continuous Operations
  • Modular Architecture With Hot-Plug Support
  • 24/7 Proactive Support
  • Telemetry to Cloud-based Portal
  • Designed for High-Reliability with No Single Point of Failure

Benefits:

No-Compromise

  • World’s First Hyperparallel Flash Array
  • High-Performance, Low-Latency, Linear Scaling, and Compact Chassis Maximizes Data Center Efficiency
  • Concurrent Block & File Protocols

End-to-End NVMe

  • 72 U.2 2.5” NVMe drives in 1-4 Media Groups
  • • Supports Storage Class Memory (SCM)
  • Simultaneous NVMe-oF/TCP, NVMe-oF/Ethernet, NVMe-oF/InfiniBand

Storage Disaggregation

  • 25X Faster Than AFA for $/IOPS
  • 50% Faster Than DAS for $/BW
  • 2X Denser Compute
  • 67% Faster Searches
  • 2-3X Lower Storage TCO
  • 3X Better Management with SAN-Like Features

Specifications

Performance Capacity Physical
Elite Very High

Bandwidth: Up to 90 GB/s R/W

IOPS: Up to 20M

Latency: Down to 40µs

Capacity: 230 TB – 1.1 PB RAW – 202 TB-978 TB Usable

Elite Medium

Bandwidth: Up to 90 GB/s R/W

IOPS: Up to 20M

Latency: Down to 40µs

Capacity: 230 TB – 1.1 PB RAW – 202 TB-978 TB Usable

Premium High

Bandwidth: Up to 50 GB/s

IOPS: 6M

Latency: Down to 150µs

Capacity: 230 TB – 1.1 PB RAW – 202 TB-978 TB Usable

Premium Low

Bandwidth: Up to 50 GB/s

IOPS: Up to 6M

Latency: Down to 150µs

Capacity: 230 TB – 1.1 PB RAW – 202 TB-978 TB Usable

Basic Medium

Bandwidth: Up to 25 GB/s

IOPS: Up to 10M

Latency: Down to 200µs

Capacity: 57 TB – 550 TB RAW; 50 TB – 489 TB Usable

Basic Very Low

Bandwidth: Up to 25 GB/s

IOPS: Up to 10M

Latency: Down to 200µs

Capacity: 57 TB – 550 TB RAW; 50 TB – 489 TB Usable

 

Common Specs

Power Consumption 3200 Max Watts
Heat Dissipation 10922 Max BTU
System Height 4 Rack Units
Protocols & Transports Simultaneous NVMe/Ethernet using RoCE v2, NVMe/InfiniBand, NVMe/TCP, NFS v3/v4, iSCSI
Software Services Snapshots, Clones, Data at Rest Encryption, Thin Provisioning, RAID, Rapid “SWARM” RAID Rebuild, Data Assurance
Management Web GUI, CLI, vCenter, Kubernetes, RESTful API, OpenStack, DTMF/Redfish, Swordfish, and SNMP v2c
Support 24/7 Proactive Support with Cloud-based Telemetry
Physical Interfaces Up to 40 ports of 100 Gb Ethernet or InfiniBand for Storage Services
Up to 2 ports of 10 GigE for Management
Dimensions 17.5” W x 31.1” D x 6.9” H
Weight 147 lbs. (66.7) (loaded with SSDs)
Safety IEC/EN 60950, CB Certificate, UL60950-1 CAN/USA-C22.2, CE Mark.

Capacity-Based Model

The Pavilion HFA supports multiple applications with the capacity and performance ideal for one or many applications. Choose the tier needed and move between tiers or move up to a higher performance/capacity tier, increasing performance and capacity linearly.

 

Elite Performance Tier
Up to 90GB/s speed, supporting capacities of 230TB to 1.1PB

Elite Performance
at a Medium Capacity

Elite Performance
at a Very High Capacity

 

Premium Performance Tier
Up to 50GB/s speed, supporting capacities of 115TB to 1.1PB

Premium Performance
at a Low Capacity

Premium Performance
at a High Capacity

 

Basic Performance Tier
Up to 25GB/s speed, supporting capacities of 57TB to 550TB

Basic Performance
at a Very Low Capacity

Basic Performance
at a Medium Capacity

Documentation:

Download the Pavilion Hyperparallel Storage Array (.PDF)