Page 1
G. Tarnopolsky /November 2003/ 1© 2003 INSIC SC 2003
Hard Disk Drive Evolution &Data Storage Systems
ProgramGiora J. Tarnopolsky
TarnoTek - www.tarnotek.com
Storage on the Lunatic FringeSC2003 - Phoenix - AZ
19 November 2003
Information Storage IndustryConsortium
Page 2
G. Tarnopolsky /November 2003/ 2© 2003 INSIC SC 2003
Agenda
Who/What/When INSICExhilarating growth of volumetric density
Drive economicsIs it likely to continue?
1 Terabyte drive likelyINSIC’s research programs
EHDRHAMR
INSIC & Data Storage Systems Research
Page 3
G. Tarnopolsky /November 2003/ 3© 2003 INSIC SC 2003
Information Storage IndustryConsortium
* Limited Member
MITISICNIST
IDEMAIDAHOMITRE
PURDUESTANFORD
MINNESOTAVANDERBILT
SANTA CLARAGEORGIA TECH
ARIZONA STATENORTHWESTERN
CARNEGIE MELLONARGONNE NAT’L LAB
WASHINGTON UNIVERSITYLAWRENCE BERKELEY NAT’L LAB
LAWRENCE LIVERMORE NAT’L LABNATIONAL UNIVERSITY OF SINGAPORE
FLORIDA INTERNATIONAL UNIVERSITYDATA STORAGE INSTITUTE (DSI)
LOS ALAMOS NATIONAL LABCENTRAL LANCASHIRE
COLORADO STATEJOHNS HOPKINSNORTHEASTERN
UC SAN DIEGOMANCHESTERUC BERKELEYOHIO STATECOLORADOPLYMOUTHMISSOURINEBRASKAVIRGINIAALABAMAHARVARDALBERTA
ARIZONAILLINOIS
IBMECD*IDC*SONYMAXELLIMATIONAPRILIS*QUANTUMSAMSUNGCERTANCEMAGNECOMP*STORAGETEKDOWA MINING*MEMS OPTICALAGERE SYSTEMSWESTERN DIGITALTORAY INDUSTRIESHEWLETT PACKARDVEECO INSTRUMENTSADVANCED RESEARCHSEAGATE TECHNOLOGYEUXINE TECHNOLOGIESHUTCHINSON TECHNOLOGYHITACHI GLOBAL STORAGE TECHNOLOGIES
collaborative research consortium for the worldwide information
storage industryestablished 1991 Conduct Joint Research on High Risk Pre-competitive
Storage Technologies Develop Technology Roadmaps
Maximize Value of University Research Obtain Government Funding
Speak for the Industry
Page 4
G. Tarnopolsky /November 2003/ 4© 2003 INSIC SC 2003
The Future of Hard Disk Drive TechnologyLab Demos: Possible HDD Areal Density Progression
Laboratory Demonstrations
1.0
10.0
100.0
1000.0
10000.0
100000.0
Jan-90 Jan-93 Jan-96 Jan-99 Jan-02 Jan-05 Jan-08 Jan-11 Jan-14 Jan-17
Date
Are
al D
ensi
ty (G
b/ in
2 )
perpendicular recording
heat-assisted mag recording
self-organized arrays ?
patterned media ?
50% CAG ra
te line
70now
highest in products
1 Terabit per inch2 goal
Page 5
G. Tarnopolsky /November 2003/ 5© 2003 INSIC SC 2003
Slower Progression of HDD Technology?
Laboratory Demonstrations
1.0
10.0
100.0
1000.0
10000.0
100000.0
Jan-90 Jan-93 Jan-96 Jan-99 Jan-02 Jan-05 Jan-08 Jan-11 Jan-14 Jan-17
Date
Are
al D
ensi
ty (G
b/ in
2 )
?
30% CAG line
Lab Demos: Possible HDD Areal Density Progression
Page 6
G. Tarnopolsky /November 2003/ 6© 2003 INSIC SC 2003
Precipitous decline in $/GB
Cost per Gigabyte, 95 mm Desktop ≤ 7,200 rpm
0.01
0.10
1.00
10.00
100.00Ja
n-98
Jan-
99
Jan-
00
Jan-
01
Jan-
02
Jan-
03
Jan-
04
Jan-
05
Jan-
06
Jan-
07
Jan-
08
Jan-
09
Time
$/G
B
HGSTMaxtorSeagateWDC
(C) 2003 TarnoTek
- 44%/ year
Page 7
G. Tarnopolsky /November 2003/ 7© 2003 INSIC SC 2003
Price of HD Drive ≥ Components
Price of HDD, 95 mm Desktop ≤ 7,200 rpm
10
100
1,00
0
Jan-
98
Jul-9
8
Jan-
99
Jul-9
9
Jan-
00
Jul-0
0
Jan-
01
Jul-0
1
Jan-
02
Jul-0
2
Jan-
03
Jul-0
3
Date
$/D
rive
HGST MaxtorSeagate WDC
(C) 2003 TarnoTek
Page 8
G. Tarnopolsky /November 2003/ 8© 2003 INSIC SC 2003
Capacity Growth
HDD Capacity vs. Time, 95 mm Desktop ≤ 7,200 rpm
1
10
100
1000Ja
n-98
Jan-
99
Jan-
00
Jan-
01
Jan-
02
Jan-
03
Jan-
04
Jan-
05
Jan-
06
Jan-
07
Jan-
08
Jan-
09
Time
HD
D C
apac
ity (G
B)
All HDDEarly HDD
(C) 2003 TarnoTek
Page 9
G. Tarnopolsky /November 2003/ 9© 2003 INSIC SC 2003
Capacity Growth: Sustainable?
HDD Capacity vs. Time, 95 mm Desktop ≤ 7,200 rpm
y = 6E-28e0.0018x
R2 = 0.9626
1
10
100
1000
Jan-
98
Jan-
99
Jan-
00
Jan-
01
Jan-
02
Jan-
03
Jan-
04
Jan-
05
Jan-
06
Jan-
07
Jan-
08
Jan-
09
Time
HD
D C
apac
ity (G
B)
All HDDEarly HDD
(C) 2003 TarnoTek
? INSIC 1 Tb/inch2
demo goal
Page 10
G. Tarnopolsky /November 2003/ 10© 2003 INSIC SC 2003
Magnetic Areal Density Tradeoffs
INSIC’s Extreme HighDensity Recording programstrives for concurrent highSNR, permanency of therecorded bit, and ability torecord (EHDR)
INSIC’s Heat AssistedMagnetic Recordingprogram uses heat toachieve recordability(HAMR)
Patterned mediaTilted perpendicular mediaSelf-organized media
recordability
sign
al-to
-no
ise
ratio
thermal stability
goal
lowSNR
unstable
can’twrite
Page 11
G. Tarnopolsky /November 2003/ 11© 2003 INSIC SC 2003
Speculation on 1 Tb/in2 HDD
Adapted from R. Wood,"The feasibility of magneticrecording at 1 Terabit persquare inch," IEEE Trans.Magn. 36, No. 1, 36-42(2000)
ParametersAreal density 1000 Gb/in2
Linear density 2500 kbpiTrack density 400 ktpiRotation 15,000 rpmDisk 33 mmCapacity 125 GB/platterInternal transfer rate 320 Mbyte/s
DriveCapacity 1 TerabytePlatters 8Heads 16Width 2 (50) in (mm)Height 1.4 (36) in (mm)Length 4 (100) in (mm)
SystemCapacity 8 TBDrives 8Physical sizeWidth 16 inHeight 2 inDepth 6 in
Page 12
G. Tarnopolsky /November 2003/ 12© 2003 INSIC SC 2003
Parting shots in “SAN Essentials”
Storage Area Networks Essentials, Richard Barkerand Paul Massiglia, John Wiley & Sons, pgs. 379-380.
“Since the evolution of SCSI in the mid-1980s, thefunctional definition of a disk (or a tape) has beenessentially constant … the basic model of a singlevector of numbered blocks has remained the same.”
“Researchers today are questioning whether thetried-and-true disk (or tape) functional model is the mosteffective way to provide network storage services.”
(text in italics added by GT)
“Since the evolution of SCSI in the mid-1980s, thefunctional definition of a disk (or a tape) has beenessentially constant … the basic model of a singlevector of numbered blocks has remained the same.”
“Researchers today are questioning whether thetried-and-true disk (or tape) functional model is the mosteffective way to provide network storage services.”
(text in italics added by GT)
Implementation & management ofTerabyte drives & systems
Page 13
G. Tarnopolsky /November 2003/ 13© 2003 INSIC SC 2003
Pre-competitive DSSResearch
Page 14
G. Tarnopolsky /November 2003/ 14© 2003 INSIC SC 2003
What is pre-competitive research?
Work that: is not now subject of industrial competition
towards commercialization of a product addresses goals that are beyond the
immediate industrial roadmaps may tackle high-risk/high-reward speculative
ideasenhances the research return-on-investment
to the benefit of all partners and of the globaldata storage systems endeavor
Page 15
G. Tarnopolsky /November 2003/ 15© 2003 INSIC SC 2003
What is pre-competitive research?
Advances whose implementation wouldrequire industrial consensus among parties thatotherwise may compete in the marketplace
e.g. storage-device OSPre-competitive research is, essentially,
research “co-opetition.”Are there pre-competitive research
opportunities in data storagesystems work?
Page 16
G. Tarnopolsky /November 2003/ 16© 2003 INSIC SC 2003
Workshop
Brainstorm the Future
Page 17
G. Tarnopolsky /November 2003/ 17© 2003 INSIC SC 2003
Workshop Target Participants
Data storage systems, software, and storagedevices’ companies
Academics & academic centersEntities vested in the storage enterprise:
government, sponsored research efforts withDSS aspects
Current sponsors: INSIC, CMRR UCSD,DTC/DISC U Minn, INSIC member companies
In progress: CAL IT2, others
Page 18
G. Tarnopolsky /November 2003/ 18© 2003 INSIC SC 2003
INSIC DSS Program Topics
Research topics will emerge from industrial &academic consensus
Gratefully acknowledge discussions with manyresearchers in creating a roster of candidatetopics.
David B. Anderson / SeagateWalter Burkhard / UC San DiegoRandal Burns / John Hopkins UniversityDavid Du / DTC, U MinnGreg Ganger / CMUGarth Gibson / CMU & PanasasGordon F. Hughes / UC San DiegoJoseph F. JaJa / U. MarylandDarrell Long / UC Santa CruzPatrick Mantey / UC Santa CruzPaul Massiglia / Veritas Software
Ethan Miller / UC Santa CruzReagan Moore / SDSC/UCSDRichard New / Hitachi GSTDavid Patterson / UC BerkeleyMichael O. Rabin / Harvard University and Hebrew
UniversityErik Riedel / Seagate ResearchTom Ruwart / DTC/DISC U MinnPaul Siegel / UC San DiegoEd Skalko / Seagate ResearchBarry Schechtman / INSICBoleslaw K. Szymanski / Rensselaer Polytechnic
Institute
Thanks!
Page 19
G. Tarnopolsky /November 2003/ 19© 2003 INSIC SC 2003
Spectrum of Research TopicsSpectrum of Research Topics
Non-volatile, lowlatency memory in
device
Non-volatile, lowlatency memory in
device
SemanticcontinuitySemantic
continuity
Application-aware storage
Application-aware storage
Digitalpermanency
Digitalpermanency
Digitalpermanency
Digitalpermanency
hardware-ish software-ish
networked,distributed
limitedlocale
OSD, object-oriented storagedevice
Performance& reliabilityPerformance& reliability
Performance& reliabilityPerformance& reliability
Multi-dimensionalspace and
projections
Page 20
G. Tarnopolsky /November 2003/ 20© 2003 INSIC SC 2003
Fred van den Bosch / Veritas
Page 21
G. Tarnopolsky /November 2003/ 21© 2003 INSIC SC 2003
There is no assured scheme for perpetual contentpreservation
Semantic continuity: make computerlanguages evolve like natural languages,assure comprehension
Storage management independent of themedium and of the content itself
Systems hold “eternal” data in devices bearing a ~three-year warranty§,*
Digital assets have undergone migrations todevices of higher performance & volumetric density.No more.
Long-term Storage: ≥ 10 years
Safe harbor: §) H/W MTBF ≥ 30 yr.*) Tape cartridges guaranteed “for life”
Page 22
G. Tarnopolsky /November 2003/ 22© 2003 INSIC SC 2003
Multi-GB, ≤ µs-latency, non-volatile memory Flash, M-RAM, or MEMSNon-volatile stratum would be “virtual disk”
such as disk is “virtual tape” to tape systemVarious application archetypes could have
assigned non-volatile storage streams - untilobjects of associated types are transferred tospecific areas on the media
Intelligent space allocation, self-defragmentation 500 GBMag 12 GBNV TT
Large, non-volatile storagestratum
Page 23
G. Tarnopolsky /November 2003/ 23© 2003 INSIC SC 2003
OSD - Object-based StorageDevice
OSDs take the storage-device-specificcomponent of the file system into the storagedevice itself
Ability of device to manage its own capacityAbility of device to export file-like objects to
their clientsWhere in the storage hierarchy is the OSD
concept to be applied? Disk remains byte-bucket but RAID is intelligent storage?
Page 24
G. Tarnopolsky /November 2003/ 24© 2003 INSIC SC 2003
Summary
Digital storage has irreversibly changed thetransaction of human activity
Change has occurred rapidly, and will continueChange will be driven by global industry &
academic effortsPlethora of unresolved issuesINSIC recognizes significant opportunities for
pre-competitive, industry/academia cooperativedata-storage-systems research, and willpromote it.
Page 25
G. Tarnopolsky /November 2003/ 25© 2003 INSIC SC 2003
Paul D. FrankPaul D. FrankExecutive DirectorExecutive Director
INSIC PROGRAMSINSIC PROGRAMS
Information Storage Industry ConsortiumInformation Storage Industry Consortium