1 Building Peta-Byte Building Peta-Byte Servers Servers Jim Gray Jim Gray Microsoft Research Microsoft Research [email protected][email protected]http://www.Research.Microsoft.com/~Gray/talks http://www.Research.Microsoft.com/~Gray/talks Kilo Kilo 10 10 3 Mega Mega 10 10 6 Giga Giga 10 10 9 Tera Tera 10 10 12 12 today, we are here today, we are here Peta Peta 10 10 15 15 Exa Exa 10 10 18 18
50
Embed
1 Building Peta-Byte Servers Jim Gray Microsoft Research [email protected]/talks Kilo10 3 Mega10 6 Giga10 9 Tera10.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Building Peta-Byte ServersBuilding Peta-Byte Servers
TeraTera 10101212 today, we are here today, we are here PetaPeta 10101515
ExaExa 10101818
2
OutlineOutline
• The challenge: Building GIANT data storesThe challenge: Building GIANT data stores
– for example, the EOS/DIS 15 PB systemfor example, the EOS/DIS 15 PB system
• Conclusion 1Conclusion 1
– Think about Maps and SCANSThink about Maps and SCANS
• Conclusion 2:Conclusion 2:
– Think about ClustersThink about Clusters
3
The Challenge -- EOS/DISThe Challenge -- EOS/DIS
• Antarctica is melting -- Antarctica is melting -- 77% of fresh water liberated77% of fresh water liberated
– sea level rises 70 meters sea level rises 70 meters – Chico & Memphis are beach-front propertyChico & Memphis are beach-front property– New York, Washington, SF, SB, LA, London, Paris New York, Washington, SF, SB, LA, London, Paris
• Let’s study it! Let’s study it! Mission to Planet EarthMission to Planet Earth
• EOS: Earth Observing System EOS: Earth Observing System (17B$ => 10B$)(17B$ => 10B$)
– 50 instruments on 10 satellites 1997-200150 instruments on 10 satellites 1997-2001– Landsat (added later)Landsat (added later)
• EOS DIS: Data Information System:EOS DIS: Data Information System:
– 3-5 MB/s raw, 30-50 MB/s cooked.3-5 MB/s raw, 30-50 MB/s cooked.– 4 TB/day, 4 TB/day, – 15 PB by year 200715 PB by year 2007
4
The Process FlowThe Process Flow• Data arrives and is pre-processed.Data arrives and is pre-processed.
– instrument data is instrument data is calibrated, calibrated, griddedgriddedaveragedaveraged
– Geophysical data is derived Geophysical data is derived • Users ask Users ask for stored data for stored data
OROR to analyze and combine data.to analyze and combine data.• Can make the pull-push split dynamicallyCan make the pull-push split dynamically
Pull Processing Push ProcessingOther Data
5
Designing EOS/DIS (for success) Designing EOS/DIS (for success) • Expect that millions will use the system Expect that millions will use the system (online)(online)
Three user categories:Three user categories:– NASA 500 -- NASA 500 -- funded by NASA to do sciencefunded by NASA to do science– Global Change 10 k - Global Change 10 k - other dirt bagsother dirt bags– Internet 20 m - Internet 20 m - everyone elseeveryone else
=> discovery & access must be automatic => discovery & access must be automatic
• Allow anyone to set up a peer- nodeAllow anyone to set up a peer- node (DAAC & SCF)(DAAC & SCF)
• Design for Ad Hoc queries, Design for Ad Hoc queries, Not Just Standard Data ProductsNot Just Standard Data Products If push is 90%, then 10% of data is read (on average). If push is 90%, then 10% of data is read (on average).
=> A failure: no one uses the data, in DSS, push is 1% or less.=> A failure: no one uses the data, in DSS, push is 1% or less.
=> computation demand is enormous=> computation demand is enormous (pull:push is 100: 1)(pull:push is 100: 1)
6
The (UC alternative) ArchitectureThe (UC alternative) Architecture
• 2+N data center design2+N data center design
• Scaleable DBMS to manage the dataScaleable DBMS to manage the data
• Emphasize Pull vs Push processingEmphasize Pull vs Push processing
• Storage hierarchyStorage hierarchy
• Data PumpData Pump
• Just in time acquisitionJust in time acquisition
7
2+N Data Center Design2+N Data Center Design• Duplex the archive (for fault tolerance)Duplex the archive (for fault tolerance)
• Let anyone build an extract (the +N)Let anyone build an extract (the +N)
• Partition data by time and by space Partition data by time and by space (store 2 or 4 ways).(store 2 or 4 ways).
• Each partition is a free-standing DBMSEach partition is a free-standing DBMS(similar to Tandem, Teradata designs).(similar to Tandem, Teradata designs).
• Clients and Partitions interact via standard protocolsClients and Partitions interact via standard protocols
and ~ 100 Terabytes of DRAMand ~ 100 Terabytes of DRAM
• 1997 requirements are 1000x smaller1997 requirements are 1000x smaller
– smaller data ratesmaller data rate
– almost no re-processing workalmost no re-processing work
9
Hardware ArchitectureHardware Architecture
• 2 Huge Data Centers2 Huge Data Centers
• Each has 50 to 1,000 nodes in a clusterEach has 50 to 1,000 nodes in a cluster
– Each node has about 25…250 TB of storage (FY00 prices)Each node has about 25…250 TB of storage (FY00 prices)– SMP SMP .5Bips to 50 Bips .5Bips to 50 Bips 20K$ 20K$
– DRAMDRAM 50GB to 1 TB50GB to 1 TB 50K$ 50K$– 100 disks 100 disks 2.3 TB to 230 TB2.3 TB to 230 TB 200K$ 200K$– 10 tape robots10 tape robots 50 TB to 500 TB 50 TB to 500 TB 100K$ 100K$– 2 Interconnects2 Interconnects 1GBps to 100 GBps1GBps to 100 GBps 20K$ 20K$
• Node costs 500K$ Node costs 500K$
• Data Center costs 25M$ (capital cost)Data Center costs 25M$ (capital cost)
• System must scale to many processors, disks, linksSystem must scale to many processors, disks, links
• Organize data as a Database, not a collection of filesOrganize data as a Database, not a collection of files
– SQL rather than FTP as the metaphorSQL rather than FTP as the metaphor
– add object types unique to EOS/DIS (Object Relational DB)add object types unique to EOS/DIS (Object Relational DB)
• DBMS based on standard object modelDBMS based on standard object model
– CORBA or DCOM (not vendor specific) CORBA or DCOM (not vendor specific)
• Grow by adding componentsGrow by adding components
• System must be self-managingSystem must be self-managing
11
Storage HierarchyStorage Hierarchy• Cache hot 10% (1.5 PB) on disk.Cache hot 10% (1.5 PB) on disk.
• Keep cold 90% on near-line tape.Keep cold 90% on near-line tape.
• Remember recent results on speculation|Remember recent results on speculation| research challenge: how trade push +store vs. pull.research challenge: how trade push +store vs. pull.
• (more on this later Maps & SCANS) (more on this later Maps & SCANS)
15 PB of Tape Robot
1 PB of Disk
10-TB RAM 500 nodes
10,000 drives
4x1,000 robots
12
Data PumpData Pump
• Some queries require reading ALL the data Some queries require reading ALL the data (for reprocessing)(for reprocessing)
• Each Data Center scans the data every 2 days.Each Data Center scans the data every 2 days.– Data rate 10 PB/day = 10 TB/node/day = 120 MB/sData rate 10 PB/day = 10 TB/node/day = 120 MB/s
• Compute on demand small jobsCompute on demand small jobs• less than 1,000 tape mountsless than 1,000 tape mounts• less than 100 M disk accessesless than 100 M disk accesses• less than 100 TeraOps.less than 100 TeraOps.• (less than 30 minute response time)(less than 30 minute response time)
• For BIG JOBS scan entire 15PB database For BIG JOBS scan entire 15PB database • Queries (and extracts) “snoop” this data pump.Queries (and extracts) “snoop” this data pump.
• Buy best product that day: commodityBuy best product that day: commodity
• Depreciate over 3 years so that facility is fresh. Depreciate over 3 years so that facility is fresh. • (after 3 years, cost is 23% of original). 60% decline peaks at 10M$(after 3 years, cost is 23% of original). 60% decline peaks at 10M$
• The challenge: Building GIANT data storesThe challenge: Building GIANT data stores
– for example, the EOS/DIS 15 PB systemfor example, the EOS/DIS 15 PB system
• Conclusion 1Conclusion 1
– Think about Maps and SCANSThink about Maps and SCANS
• Conclusion 2:Conclusion 2:
– Think about ClustersThink about Clusters
19
Meta-Message:Meta-Message: Technology Ratios Are Important Technology Ratios Are ImportantMeta-Message:Meta-Message: Technology Ratios Are Important Technology Ratios Are Important
• If everything gets faster & cheaper If everything gets faster & cheaper
at the same rate at the same rate THEN nothing really changes.THEN nothing really changes.
•
Things getting MUCH BETTER:Things getting MUCH BETTER:
• Things staying about the sameThings staying about the same– speed of light (more or less constant)speed of light (more or less constant)– people (10x more expensive)people (10x more expensive)– storage speed (only 10x better)storage speed (only 10x better)
20
Today’s Storage Hierarchy : Today’s Storage Hierarchy : Speed & Capacity vs Cost TradeoffsSpeed & Capacity vs Cost TradeoffsToday’s Storage Hierarchy : Today’s Storage Hierarchy : Speed & Capacity vs Cost TradeoffsSpeed & Capacity vs Cost Tradeoffs
1015
1012
109
106
103
Typ
ical
Sys
tem
(by
tes)
Size vs Speed
Access Time (seconds)10-9 10-6 10-3 10 0 10 3
Cache
Main
Secondary
Disc
Nearline Tape Offline
Tape
Online Tape
104
102
100
10-2
10-4
$/M
B
Price vs Speed
Access Time (seconds)10-9 10-6 10-3 10 0 10 3
Cache
MainSecondary
DiscNearline
TapeOffline Tape
Online Tape
21
Storage Ratios ChangedStorage Ratios Changed
• 10x better access time10x better access time
• 10x more bandwidth10x more bandwidth
• 4,000x lower media price4,000x lower media price
• DRAM/DISK 100:1 to 10:10 to 50:1DRAM/DISK 100:1 to 10:10 to 50:1
Disk Performance vs Time
1
10
100
1980 1990 2000
Year
acce
ss t
ime
(ms)
1
10
100
ban
dw
idth
(M
B/s
)
Disk Performance vs Time(accesses/ second & Capacity)
1
10
100
1980 1990 2000
Year
Acc
esse
s p
er
Sec
on
d
0.1
1
10
Dis
k C
apac
kty
(GB
)
Storage Price vs Time
0.01
0.1
1
10
100
1000
10000
1980 1990 2000
Year
$/M
B
22
What's a TerabyteWhat's a Terabyte1 Terabyte 1,000,000,000 business letters 100,000,000 book pages 50,000,000 FAX images 10,000,000 TV pictures (mpeg) 4,000 LandSat images
Library of Congress (in ASCI) is 25 TB 1980: 200 M$ of disc 10,000 discs 5 M$ of tape silo 10,000 tapes
1997: 200 K$ of magnetic disc 120 discs 250 K$ of optical disc robot 200 platters 25 K$ of tape silo 25 tapes
Terror Byte !!.1% of a PetaByte!!!!!!!!!!!!!!!!!!
150 miles of bookshelf 15 miles of bookshelf 7 miles of bookshelf 10 days of video
23
The Cost of Storage & AccessThe Cost of Storage & AccessThe Cost of Storage & AccessThe Cost of Storage & Access
All information comes electronically entertainment publishing business
Information Network, Knowledge Navigator, Information at Your Fingertips
Multimedia: Text, voice, image, video, ...
26
Thesis: Performance =Storage AccessesThesis: Performance =Storage Accesses not Instructions Executed not Instructions ExecutedThesis: Performance =Storage AccessesThesis: Performance =Storage Accesses not Instructions Executed not Instructions Executed• In the “old days” we counted instructions and IO’sIn the “old days” we counted instructions and IO’s
• Now we count memory referencesNow we count memory references
• Processors wait most of the timeProcessors wait most of the time
SortDisc Wait
Where the time goes: clock ticks used by AlphaSort Components
SortDisc WaitOS
Memory Wait
D-Cache Miss
I-Cache MissB-Cache
Data Miss
27
The Pico ProcessorThe Pico ProcessorThe Pico ProcessorThe Pico Processor
1 M SPECmarks
106 clocks/ fault to bulk ram
Event-horizon on chip.
VM reincarnated
Multi-program cache
Terror Bytes!
10 microsecond ram
10 millisecond disc
10 second tape archive 100 petabyte
100 terabyte
1 terabyte
Pico Processor
10 pico-second ram1 MM
3
megabyte
10 nano-second ram 10 gigabyte
28
Storage Latency: How Far Storage Latency: How Far Away is the Data?Away is the Data?Storage Latency: How Far Storage Latency: How Far Away is the Data?Away is the Data?
RegistersOn Chip CacheOn Board Cache
Memory
Disk
12
10
100
Tape /Optical Robot
10 9
10 6
Sacramento
This CampusThis Room
My Head
10 min
1.5 hr
2 Years
1 min
Pluto
2,000 YearsAndromeda
29
The Five Minute RuleThe Five Minute Rule
• Trade DRAM for Disk AccessesTrade DRAM for Disk Accesses
• Cost of an access (DriveCost / Access_per_second)Cost of an access (DriveCost / Access_per_second)
• Cost of a DRAM page ( $/MB / pages_per_MB)Cost of a DRAM page ( $/MB / pages_per_MB)
• Break even has two terms:Break even has two terms:
• Technology term and an Economic termTechnology term and an Economic term
• Grew page size to compensate for changing ratios.Grew page size to compensate for changing ratios.
• Still at 5 minute for random, 1 minute sequentialStill at 5 minute for random, 1 minute sequential 1
ofDRAMPricePerMB
skDrivePricePerDi
skecondPerDiAccessPerS
ofDRAMPagesPerMBtervaleferenceInBreakEvenR
1ofDRAMPricePerMB
skDrivePricePerDi
skecondPerDiAccessPerS
ofDRAMPagesPerMBtervaleferenceInBreakEvenR
30
Shows Best Page Index Page Size ~16KBShows Best Page Index Page Size ~16KB
Index Page Utility vs Page Size and Index Elemet Size
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Page Size (KB)
Uti
lity
16 B 0.64 0.72 0.78 0.82 0.79 0.69 0.54
32 B 0.54 0.62 0.69 0.73 0.71 0.63 0.50
64 B 0.44 0.53 0.60 0.64 0.64 0.57 0.45
128 B 0.34 0.43 0.51 0.56 0.56 0.51 0.41
2 4 8 16 32 64 128
16 byte entries
32 byte
64 byte
128 byte
Index Page Utility vs Page Size and Disk Performance
– RAM: RAM: MB and $/MB: today at 10MB & 100$/MBMB and $/MB: today at 10MB & 100$/MB– Disk:Disk: GB and $/GB: today at 10 GB and 200$/GBGB and $/GB: today at 10 GB and 200$/GB– Tape: Tape: TB and $/TB: today at .1TB and 25k$/TB TB and $/TB: today at .1TB and 25k$/TB
(nearline)(nearline)
• Access time (latency)Access time (latency)– RAM:RAM: 100 ns100 ns– Disk: Disk: 10 ms 10 ms– Tape: 30 second pick, 30 second position Tape: 30 second pick, 30 second position
• Transfer rateTransfer rate– RAM:RAM: 1 GB/s 1 GB/s– Disk:Disk: 5 MB/s - - - Arrays can go to 1GB/s 5 MB/s - - - Arrays can go to 1GB/s– Tape: 5 MB/s - - - striping is problematicTape: 5 MB/s - - - striping is problematic
32
New Storage Metrics: New Storage Metrics: Kaps, Maps, SCAN?Kaps, Maps, SCAN?New Storage Metrics: New Storage Metrics: Kaps, Maps, SCAN?Kaps, Maps, SCAN?
• Kaps: How many kilobyte objects served per secondKaps: How many kilobyte objects served per second– The file server, transaction processing metricThe file server, transaction processing metric– This is the OLD metric.This is the OLD metric.
• Maps: How many megabyte objects served per secondMaps: How many megabyte objects served per second– The Multi-Media metricThe Multi-Media metric
• SCAN: How long to scan all the dataSCAN: How long to scan all the data– the data mining and utility metricthe data mining and utility metric
How To Get Lots of Maps, SCANsHow To Get Lots of Maps, SCANsHow To Get Lots of Maps, SCANsHow To Get Lots of Maps, SCANs• parallelism: use many little devices in parallelparallelism: use many little devices in parallel
• Beware of the media mythBeware of the media myth
• Beware of the access time mythBeware of the access time myth
1 Terabyte
10 MB/s
At 10 MB/s: 1.2 days to scan
1 Terabyte
1,000 x parallel: 100 seconds SCAN.
Parallelism: divide a big problem into many smaller ones to be solved in parallel.
35
The Disk Farm On a CardThe Disk Farm On a CardThe Disk Farm On a CardThe Disk Farm On a CardThe 100GB disc cardThe 100GB disc cardAn array of discsAn array of discsCan be used asCan be used as 100 discs100 discs 1 striped disc1 striped disc 10 Fault Tolerant discs10 Fault Tolerant discs ....etc....etcLOTS of accesses/secondLOTS of accesses/second bandwidthbandwidth
14"
Life is cheap, its the accessories that cost ya.
Processors are cheap, it’s the peripherals that cost ya (a 10k$ disc card).
36
Tape Farms for Tertiary StorageTape Farms for Tertiary StorageNot Mainframe SilosNot Mainframe SilosTape Farms for Tertiary StorageTape Farms for Tertiary StorageNot Mainframe SilosNot Mainframe Silos
Scan in 27 hours.many independent tape robots(like a disc farm)
The Metrics: The Metrics: Disk and Tape Farms Win Disk and Tape Farms Win The Metrics: The Metrics: Disk and Tape Farms Win Disk and Tape Farms Win
Data Motel:Data checks in, but it never checks out
38
Tape & Optical: Tape & Optical: Beware of the Beware of the Media MythMedia MythTape & Optical: Tape & Optical: Beware of the Beware of the Media MythMedia Myth
Optical is cheap: 200 $/platter 2 GB/platter => 100$/GB (2x cheaper than disc)
Tape is cheap: 30 $/tape 20 GB/tape => 1.5 $/GB (100x cheaper than disc).
39
Tape & Optical Tape & Optical RealityReality: : Media is 10% of System CostMedia is 10% of System CostTape & Optical Tape & Optical RealityReality: : Media is 10% of System CostMedia is 10% of System Cost
( more expensive than mag disc ) Robots have poor access times Not good for Library of Congress (25TB) Data motel: data checks in but it never checks out!
40
The Access Time MythThe Access Time MythThe Access Time MythThe Access Time MythThe Myth: seek or pick time dominatesThe Myth: seek or pick time dominatesThe reality: (1) Queuing dominatesThe reality: (1) Queuing dominates (2) Transfer dominates BLOBs(2) Transfer dominates BLOBs (3) Disk seeks often short(3) Disk seeks often shortImplication: many cheap servers Implication: many cheap servers
better than one fast expensive server better than one fast expensive server– shorter queuesshorter queues– parallel transferparallel transfer– lower cost/access and cost/bytelower cost/access and cost/byte
This is now obvious for disk arraysThis is now obvious for disk arraysThis will be obvious for tape arraysThis will be obvious for tape arrays
Seek
Rotate
Transfer
Seek
Rotate
Transfer
Wait
41
OutlineOutline
• The challenge: Building GIANT data storesThe challenge: Building GIANT data stores
– for example, the EOS/DIS 15 PB systemfor example, the EOS/DIS 15 PB system
• Conclusion 1Conclusion 1
– Think about Maps and SCAN & 5 minute ruleThink about Maps and SCAN & 5 minute rule
• Conclusion 2:Conclusion 2:
– Think about ClustersThink about Clusters
42
Scaleable ComputersScaleable ComputersBOTH SMP and Cluster BOTH SMP and Cluster
SMPSuper Server
DepartmentalServer
PersonalSystem
Grow Up with SMPGrow Up with SMP4xP6 is now standard4xP6 is now standard
Grow Out with ClusterGrow Out with Cluster
Cluster has inexpensive partsCluster has inexpensive parts
Clusterof PCs
43
What do TPC results say?What do TPC results say?• Mainframes do not compete on performance or priceMainframes do not compete on performance or price
They have great legacy code (MVS)They have great legacy code (MVS)
• PC nodes performance is 1/3 of high-end UNIX nodesPC nodes performance is 1/3 of high-end UNIX nodes
– 6xP6 vs 48xUltraSparc 6xP6 vs 48xUltraSparc
• PC Technology is 3x cheaper than high-end UNIXPC Technology is 3x cheaper than high-end UNIX
• Peak performance is a clusterPeak performance is a cluster
– Tandem 100 node clusterTandem 100 node cluster
– DEC Alpha 4x8 clusterDEC Alpha 4x8 cluster
• Commodity solutions WILL come to this marketCommodity solutions WILL come to this market
44
Cluster AdvantagesCluster Advantages
• Clients and Servers made from the same stuff.Clients and Servers made from the same stuff.
– Inexpensive: Built with commodity components Inexpensive: Built with commodity components
6k$/slice)6k$/slice)• Teradata, Tandem, DEC moving to NT+low slice priceTeradata, Tandem, DEC moving to NT+low slice price
• IBM: 512 nodes ASCI @ 100m$ (200k$/slice)IBM: 512 nodes ASCI @ 100m$ (200k$/slice)• PC clusters (bare handed) at dozens of nodes PC clusters (bare handed) at dozens of nodes
web servers (msn, PointCast,…), DB serversweb servers (msn, PointCast,…), DB servers
• KEY TECHNOLOGY HERE IS THE APPS.KEY TECHNOLOGY HERE IS THE APPS.– Apps distribute dataApps distribute data– Apps distribute executionApps distribute execution
46
Clusters are winning the high endClusters are winning the high end• Until recently a 4x8 cluster has best TPC-C performanceUntil recently a 4x8 cluster has best TPC-C performance• Clusters have best data mining story (TPC-D)Clusters have best data mining story (TPC-D)• This year, a 32xUltraSparc cluster won the MinuteSort This year, a 32xUltraSparc cluster won the MinuteSort
1.0E+02
1.0E+03
1.0E+04
1.0E+05
1.0E+06
1.0E+07
1985 1990 1995 2000
Sort Records/second vs Time
M68000
Cray YMP
IBM 3090
Tandem
Hardware Sorter
Sequent
Intel Hyper
SGIIBM RS6000
NOW
Alpha
Next NOW (100 nodes)
47
Clusters (Plumbing)Clusters (Plumbing)
• Single system imageSingle system image
– namingnaming
– protection/securityprotection/security
– management/load balancemanagement/load balance
• Fault ToleranceFault Tolerance
• Hot Pluggable hardware & SoftwareHot Pluggable hardware & Software
48
So, What’s New?So, What’s New?• When slices cost 50k$, you buy 10 or 20.When slices cost 50k$, you buy 10 or 20.• When slices cost 5k$ you buy 100 or 200.When slices cost 5k$ you buy 100 or 200.• Manageability, programmability, usability Manageability, programmability, usability
become key issues (total cost of ownership).become key issues (total cost of ownership).• PCs are MUCH easier to use and programPCs are MUCH easier to use and program