accelerate your business with flash-fueled mongodb
DESCRIPTION
TRANSCRIPT
© 2014 Pure Storage, Inc. | 1
MongoDB on Pure Storage
© 2014 Pure Storage, Inc. | 2
Introducing Pure Storage
John Colgrove
Founder, CTO
Matt KixmoellerVP, Products
Mission: Enabling the all-flash enterprise
Team: A team of 500+ storage industry veterans globally
Backing: $245M from committed investors
Differentiated Technology: 50+ patents in process
Scott DietzenCEO
John HayesFounder
Chief Arch.
Michael Cornwell
VP, APJ
DavidHatfieldPresident
ChristopherZang
VP, Support
Matt BurrVP, Americas
Bob WoodVP, Engineering
ToddEngleVP, Ops
Joao“John”Silva
EMEA Sales
© 2014 Pure Storage, Inc. | 3
The Pure Storage FlashArray 400 Series SpecsFA-420FA-405
Scale• Up to 40+TBs usable**• 2.75 – 11 TBs raw flash
Performance*• 100,000 32K IOPS @ <1ms
average latency• 3 GB/s bandwidth
Host Connectivity• Up to 4 x 8Gb/s FC • Up to 4 x 10Gb/s Ethernet
iSCSI• Built-in Replication ports
Additional Specifications• 4 x Intel 8-core CPUs• 256GB DRAM• 2 x 1U HA Controllers, 660W
Scale• Up to 125+TBs usable**• 11 – 35 TBs raw flash
Performance*• 150,000 32K IOPS @ <1ms
average latency• 5 GB/s bandwidth
Host Connectivity• Up to 8 x 8Gb/s FC or 10Gb/s
Ethernet iSCSI or mixed• Built-in Replication ports
Additional Specifications• 4 x Intel 8-core CPUs• 512GB DRAM• 2 x 2U HA Controllers, 800W
* Up to. Performance depends on workload, I/O mix and hardware configuration. ** Usable depends on actual data reduction rate, average of 6-to-1. Usable includes RAID and metadata overhead, FlashCare reserve and GB-to-GiB conversion.
FA-450
Scale• Up to 250+TBs usable**• 34 – 70 TBs raw flash
Performance*• 200,000 32K IOPS @ <1ms
average latency• 7 GB/s bandwidth
Host Connectivity• Up to 12 x 16Gb/s FC or
10Gb/s Ethernet iSCSI or mixed
• Built-in Replication ports
Additional Specifications• 4 x Intel 12-core CPUs• 1024GB DRAM• 2 x 2U HA Controllers, 1000W
Dual Controller (Front View)
(Rear View)
© 2014 Pure Storage, Inc. | 4
The Purity Operating Environment:
Storage Software Born for Flash
© 2014 Pure Storage, Inc. | 5
How Purity Works
ADAPTIVE METADATA FABRIC
Flash Memory
Acknowledge• Checksum• Copy to NV-RAM• Acknowledge to Host
DEDUPE
Reduce
• Pattern removal• Deduplication• Compression
P P
Protect
• Create RAID-3D segments
• Flush writes via flash scheduler
• Encrypt
Validate &
Serve• Validate
checksums
FlashCare
• Global wear leveling & refresh
• Global deletion management• Integrity checking• Continuous optimization
Multi-PathRead
• Read from fastest path via scheduler
• Decompress
Host
© 2014 Pure Storage, Inc. | 6
Data
Reduct
ion C
om
ponent
Resu
lts
(x-t
o-1
)
Typical Deduplication vs. CompressionAll Pure Storage Arrays in Customer Deployments w/ Call-Home Enabled • Compression delivers
reliably 2-4x data reduction
• In arrays with total reduction <10-to-1 (90%), the total data reduction is dominated by compression
• Deduplication becomes dominant in arrays at >10-to-1, mostly VDI environments
• Total Data Reduction (xT) = Compression (xC) x Deduplication (xD)
• These results are from the ENTIRE Pure Storage customer base, with a wide range of DB, VM, and VDI workloads represented (average reduction = 6-to-1)Source: Pure Storage CloudAssist database, results accurate as of Nov 8 th, 2013
Average Total Data Reduction (x-to-1)Total = Dedupe * Compression
Dedupe AND Compression Are Required for Maximum Data Reduction
© 2014 Pure Storage, Inc. | 7
The Flash Price Barrier has Been Broken!
$/GBHA
USABLEW/ SW
2010 2011 2012 2013 2014 2015…
$6$5
$4 10/15KHDD
$2 $1.5 $1 SATA
HDD
>$20
$15
$10
$6
$3
FlashAppliances
$8
$5
$4
$3
$2
3-10xData Reduction!
© 2014 Pure Storage, Inc. | 8
Why Pure Storage for MongoDB?
1 Simplicity
2 Improved High Availability
3 Agile Development
© 2014 Pure Storage, Inc. | 9
Common Vision (Founders could be twins )
MongoDB is created to make it easy to code, scale and manage databases
Pure Storage is created to make it easy to provision, scale and manage storage
200% higher TCO vs. Big Storage
© 2014 Pure Storage, Inc. | 10
Simplicity Will Set You Free!
© 2014 Pure Storage, Inc. | 11
Improve TCO without compromising availability
© 2014 Pure Storage, Inc. | 12
Agile Development with Snaps & Clones
Instantly snapshot any volume– zero reservations or planning required– zero performance overhead– flexible consistency groups
Snapshots are just new volumes, with full capabilities – mount, read, write, snapshot again, etc.– zero dependencies on one another– all have full performance
=
Always full yet space saving– zero data duplication– always thin, deduped, compressed
Recover anything to anywhere– instantly roll forward or backward– recovery ANY volume from ANY snapshot
FlashArray Virtual Volumes100% Metadata
…1000s
© 2014 Pure Storage, Inc. | 13
A real-world DATA reduction with Pure Storage
RITA collects all this flight information, archives it and distributes it
© 2014 Pure Storage, Inc. | 14
Data Reduction on RITA Database14 years of flight data, ~500GB MongoDB database size, Only 47GB on Pure
© 2014 Pure Storage, Inc. | 15
The Love Your Storage™ Guarantee
If you don’t love it, we’ll take it back.
Dedupe? Guaranteed.
Performance? Guaranteed.
Resilience? Guaranteed.
Simplicity? Guaranteed.
Support? Guaranteed.
Within 30 days of delivery. Full terms at purestorage.com/guarantee
© 2014 Pure Storage, Inc. | 16
Introducing Forever Flash
Free Every Three• Receive controller upgrades every
3 years when renewing maintenance! • 5-Year maintenance plan now includes
controller upgrade after 3rd year• 4+5-Year maintenance renewal comes
with controller upgrade
Fresh Every Upgrade• Reset maintenance terms on your
entire array, at then-current pricing!• Upgrade controllers or add shelves to
make your array new again• Simplify maintenance terms and true-
ups
free
© 2014 Pure Storage, Inc. | 17
Next Steps – Try it for yourself, no risk
© 2014 Pure Storage, Inc. | 18
Motivation for Network Storage -Ride the CPU Curve
Separating CPU and Disk gives much better flexibility• Compute / Storage innovations follow different cycles• Better investment protection as compute and storage can be refreshed
separately• Improves Operational Efficiency• Networks are becoming faster and cheaper
Servers are becoming
DenserCheaper
CPU are becoming
FasterDenser
Cheaper