Flash’s Role in Big Data, Past Present, and Future Jim Handy Tutorial: Fast Storage for Big Data Hot Chips Conference August 25, 2013 Memorial Auditorium Stanford University O BJECTIVE A NALYSIS
O BJECTIVE A NALYSIS Semiconductor Market Research • Market consulting/research firm – Market analysis, strategies, white papers • Highly-respected lead analysts – Jim Handy: Memories & SSDs – Lane Mason: Memory chips – Tom Starnes: Processors • Industry experience & 25+ years in field • Reports, Competitive Analysis, Consulting O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Haven’t We Seen You Before? • HotChips 2010: The Inevitable Rise of NVM in Computing O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Agenda • The problem • SSDs as a solution • The role of future memories • Tomorrow’s computing architecture O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Big Data • Data is exploding. By 2020: – Annual data production will be 35 ZB (CSC) – 50 billion “things” on the Internet (Cisco) • Systems aren’t keeping pace – CPUs moving along nicely – HDD/DRAM/Flash capacities still growing – Interfaces are not moving fast enough • Result: Break the system into smaller chunks O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
The DRAM/HDD Speed Gap 1.E+07 10 7 Semiconductor 1.E+06 10 6 memories becoming faster and cheaper Bandwidth (MB/s) 1.E+05 10 5 Disks becoming 10 4 1.E+04 cheaper, but no faster 10 3 1.E+03 10 2 1.E+02 Result: 10 1 1.E+01 Speed gap is widening 10 0 1.E+00 10 -1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 Price per Gigabyte From: Solid State Drives in the Enterprise O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
“Sharding” • Big Data? Subdivide the problem! Sharded System Multiple Servers O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Agenda • The problem • SSDs as a solution • The role of future memories • Tomorrow’s computing architecture O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Add Flash, Cut Server Count Sharded System Single Server with SSD SSD SSD Multiple Servers O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Other Flash Benefits • Less power & cooling • Improved reliability – The fewer things there are, the fewer will fail • Floor space reduction • Lower licensing fees • Faster error recovery – RAID rebuilds – Backup restore – Snapshots O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Agenda • The problem • SSDs as a solution • The role of future memories • Tomorrow’s computing architecture O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
How to Maintain this Inertia? $10,000 Average Price per Gigabyte $1,000 DRAM $100 $10 $1 0 1 2 3 4 5 6 7 8 9 0 1 2 0 0 0 0 0 0 0 0 0 0 1 1 1 Q Q Q Q Q Q Q Q Q Q Q Q Q 1 1 1 1 1 1 1 1 1 1 1 1 1 From: Hybrid Drives: How, Why, & When? O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
NAND’s Scaling Limit • NAND will reach a limit – Too few electrons per gate – Needs constant shrinks for cost reductions – 4-bit/cell hard to make • This may be the maximum possible • Other technologies will scale past NAND – PCM, MRAM, RRAM, FRAM…. • Not yet clear which will win O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
An NVM Timeline • 2013: 1Xnm planar cell – Requires Hi-k gate oxide • 2015: 1Ynm planar cell – May be the last planar cell • 2017: 3D NAND in volume • 2019: 3D – “The Next Generation” • 2021: Final (?) 3D generation • 2023: NAND yields to new technology O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
How Alternatives Will Emerge 100 Relative Cost per Bit Flash New Tech 10 1 0 180nm 65nm 23nm 8nm Process Node O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
New Memories are Better NAND New Memory • Serial read • Random read • Erase before write • Overwrite • Block erase/page write • Byte write • Slow write • Fast write • Inherent bit errors • Lower error rates • Wear • Low/no wear Opens pathway to “Storage Class Memory” O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Impact of New Memories O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Agenda • The problem • SSDs as a solution • The role of future memories • Tomorrow’s computing architecture O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
NVM Won’t Cross HDD $/GB $1,000.00 Price per Gigabyte $100.00 $10.00 $1.00 20X the Price! $0.10 $0.01 2004 2005 2006 2007 2008 2009 2010 2011 2012 HDD Pricing Courtesy of PriceG2 O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
An Evolving Computing Framework Yesterday Today Tomorrow CPU CPU CPU DRAM DRAM DRAM NVM SSD HDD HDD HDD O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
NAND Fits in Computers Today 1.E+07 10 7 1.E+06 10 6 Bandwidth (MB/s) 1.E+05 10 5 1.E+04 10 4 1.E+03 10 3 1.E+02 10 2 1.E+01 10 1 10 0 1.E+00 10 -1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 Price per Gigabyte From: Solid State Drives in the Enterprise O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Flash as Memory Hard Way Easy Way • It’s nonvolatile • It’s cheaper than DRAM – Write new code – It’s faster than HDD – Create new topologies • Who cares about • Worry about volatility? coherency • Handle coherency like • Everything is new you do with DRAM – Use existing code O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Flash More Economical than DRAM From: How PC NAND O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com Will Undermine DRAM
Enterprise SSD Forecast 4 3 Millions of Units $3.5 Billion in 2016 59% Unit CAGR 2 43% Revenue CAGR 1 0 2009 2010 2011 2012 2013 2014 2015 2016 From: Solid State Drives in the Enterprise O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Summary • Flash belongs in all computers today – So does HDD • Flash vs. DRAM, not flash vs. HDD • New NVMs will require new computing architectures – “Flash as memory” can be used today O BJECTIVE A NALYSIS – www.O BJECTIVE -A NALYSIS .com
Thank You! Jim Handy O BJECTIVE A NALYSIS
Recommend
More recommend