Evaluate Pure Storage

Compare the FlashArray with other flash alternatives.

Pure Storage Flash Array

Flash Array

Disk arrays with flash tiering / caching

Disk arrays with
flash tiering / caching

Flash Appliance

Flash Appliance

PCIe Card

PCIe Card


FlashArray vs. Disk array
with flash tier / cache

  • < 1ms latency
  • No outlier IOs
  • < 1ms latency when IOs hit flash
  • 5-10+ ms latency when IOs hit disk
  • Average performance depends on cache hit ratio
  • 100% flash, one large pool
  • Independent disk and flash tiers
  • Caching policies to size/manage
  • Each application must be analyzed for cache suitability
  • $5-10/GB usable
    (HA, RAID, dedupe included)
  • $15-30/GB usable
  • 400,000 IOPS
  • <1ms average latency
  • 10,000s IOPS
  • 5-10ms average latency
Suitable Workloads
  • All major enterprise applications
  • Applications which cache well (small databases, non-random IO)
  • Tier 2 applications without consistent performance needs
Space & Power
  • 1,000 watts for 50TB usable
  • 8U for 50TB usable
  • ~5,000 watts for 50TB usable
  • ~40U for 50TB usable

FlashArray vs. Flash Appliances

  • 400,000 IOPS
  • <1ms latency
  • Up to 500,000 IOPS
  • <1ms latency, potentially 100s ms
  • $5-10/GB usable
    (HA, RAID, dedupe included)
  • $20-40/GB usable
    (HA, RAID, hot spares, spare capacity for flash management included)
  • 100s TBs usable
  • Expandable by adding storage shelves
  • 5-20 TB usable
  • Not expandable
Resiliency & Serviceability
  • Active/active independent clustered controllers
  • Hot-swappable controllers
  • Hot-swappable SSDs
  • Two controller cards in the same chassis
  • Internal non-hot-swap controllers
  • Internal non-hot-swap flash modules
Software Stack
  • Dedupe, compression, thin provisioning, snapshots, VMware integration & plug-in
  • No integrated software

FlashArray vs. PCIe flash cards

  • 400,000 IOPS
  • <1ms latency
  • Flash management executed on dedicated CPUs in array
  • Up to 500,000 IOPS
  • 100s ms latency
  • Flash management executed on host CPUs
  • $5-10/GB usable
    (HA + RAID + dedupe)
  • $10/GB raw, per server
  • $20/GB HA (either mirrored in a server or between servers)
High Availability & Serviceability
  • Active/active clustered HA controllers
  • Hot-swap flash modules and controllers
  • HA via mirroring two cards in a server
  • No cross-server HA model
  • Dedupe, compression, thin provisioning, VMware integration
  • RAID, HA, data integrity, snapshots
  • Flash management
  • Mirroring
  • Flash management
Suitable Workloads
  • All major enterprise applications
  • Small database applications which can fit in the card, and where high availability is not required, or is handled by the application
  • Non-persistent database files