Académique Documents
Professionnel Documents
Culture Documents
Day 1
John Banchy
Systems Architect
Version 4
100%
Gain by Innovation
80% Gain by Technology Scaling
60%
40%
20%
0%
180 nm 130 nm 90 nm 65 nm 45 nm 32 nm 22 nm
P4 P4+ P5+ P6 P7 P7+ P8
Carbon Nanotubes
IBM demonstrated a 10,000 transistor chip
IBM demonstrated a 9nm transistor
5x-10x better performance
Graphene
IBM demonstrated a 100 GHz transistor
IBM demonstrated a functional circuit
10,000x better performance
0
IBM Microsoft Google Intel HP Oracle
System 1 32 40 32 48
Control Node cores cores cores SOD
Unit 2 64 80 64 96
Nodes cores cores cores SOD
Node 2
3 96 144
Nodes SOD SOD
* IBM Statement of Direction for 2015. For more details see the IBM Power E880 Announcement Letter released on October 6, 2014. IBMs
statements regarding its plans, directions, and intent are subject to change or withdrawal without notice at IBMs sole discretion.
14 IBM Corporation, 2015
Server Details
512 GB (1S)
Max Memory
1 TB (2S)
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice.
16 IBM Corporation, 2015
POWER8 S814
System (8286-41A) 4U
One 4-core 3.02 GHz (rack/tower)
POWER8
One 6-core 3.02 GHz (rack/tower)
Processors
One 8-core 3.72 GHz (rack)
Sockets 1
64 GB (4-core)
Max Memory
512 GB (6 or 8-core)
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice.
17 IBM Corporation, 2015
POWER8 S822
System (8284-22A) 2U
Sockets 1 or 2
512 GB (1S)
Max Memory
1 TB (2S)
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice.
18 IBM Corporation, 2015
POWER8 S824L
System (8247-42L) 4U
Sockets 2
Max Memory 1 TB
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice.
19 IBM Corporation, 2015
POWER8 S812L
System (8247-21L) 2U
Sockets 1
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice.
20 IBM Corporation, 2015
POWER8 S822L
System (8247-22L) 2U
Sockets 1 or 2
Max Memory 1 TB
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice.
21 IBM Corporation, 2015
POWER8 E880
9119-MHE Building Block System Maximum
POWER8 32 x 4.35 GHz 128 x 4.35 GHz
Processors 48 POWER8 * 192 POWER8 *
Sockets 4 8 going to 16 *
DVD Bays 1 1
DVD Bays 1 1
Max Disk
Drawers 64 64
(24 SFF bays)
Integrated PCIe 8 PCIe Gen3 x16 16 PCIe Gen3 x16
Max I/O Drawers* 2 going to 4* 4 going to 8 *
Max PCIe Slots* 24 going to 48* 48 going to 96*
Right-half
Left-half
Left Right
System
FSP Clock Clock FSP
Card Card Card Card Control Unit
Dual
Power
Cords
System
Node
IBMs statements regarding its plans, directions, and intent are subject to change or withdrawal without notice at IBMs sole discretion.
28 IBM Corporation, 2015
EXP24S SFF Gen2-bay Drawer
Redundant power
Rear
* Applies to orders for AIX, Linux, and VIOS, IBM i is ordered as 1 set
29 IBM Corporation, 2015
Optimized for Data
Core to business
Update transactional
High volume
Secure
Competitive advantage
Ubiquitous
Unpredictable
Competitive advantage
Anytime, anywhere
Untrusted devices
Exposes internal IT
Batch
Memory is slow
1
clock cycle
Core relative to
cache
1-100
Cache
clock cycles
400-800
Memory
clock cycles
Ejected New
Cache
Cache
C C C C C C
Mem L4 L4 Mem
L1 L1 L1 L1 L1 L1
L2 L2 L2 L2 L2 L2
Mem L4 L4 Mem
Shared L3 Cache
Mem L4 L4 Mem
L2 L2 L2 L2 L2 L2
L1 L1 L1 L1 L1 L1
Mem L4 L4 Mem
C C C C C C
Other
Large working sets
Memory Write burst traffic
Mixed reads and writes
On-Module Off-Module
1 Hop 1 to 3 Hops
Maximum of 2 hops
82
64
56
+14 % +28%
A A
B B
C C
D D
E E
Restart
Time
Systems of Insight
Systems of Engagement
1.00
POWER7+ POWER8
x x x x
GX++ PCIe Gen3 x16 8 1 8 1 PCIe Gen3 x8
(2) 20 GB/s (2) 32 GB/s 6 6 (2) 16 GB/s
40 GB/s 96 GB/s
Peak Bandwidth Peak Bandwidth
x x x x x x
PCIe Gen3 x16 1 1 PCIe Gen3 x16 8 1 8 1 PCIe Gen3 x8
(2) 32 GB/s 6 6 (2) 32 GB/s 6 6 (2) 16 GB/s
64 GB/s 96 GB/s
Peak Bandwidth Peak Bandwidth
POWER8
192 GB/s
S824
POWER7+
60 GB/s
740
POWER7
50 GB/s
740
POWER8
POWER8
230 GB/s
Enterprise
POWER7 68 GB/s
32
POWER6
GB/s
CPU
Read and write
data to and from
network or storage
1 Gb Ethernet
Large
Memory Memory Memory number of
nodes
Scale server
and disk
together
Replicated
local disk
Memory Memory
GZIP
Compression 10 Gb Ethernet
Accelerator
1/3 1/2
Spectrum
Of the Storage Number of Servers
Scale (TeraSort Benchmark)
Later
Pre-Defined Real-Time
Limited user mix Broad user mix
Historical data Actionable data
Batch like queries Speed matters
Human tuned Self optimizing
Actionable compression
c1 c2 c3 c4 c5 c6 Columnar store
Compression
Data skipping
No indexes
Data In Raw
Memory Data
POWER8 Single
POWER8 Threading
Instruction Multiple Data
A B C D
Compare
B
Result
Based on IBM internal tests as of April 7, 2014 comparing IBM DB2 with BLU Acceleration on Power with a comparably tuned competitor row store
database server on x86 executing a materially identical 2.6TB BI workload in a controlled laboratory environment. Test measured 60 concurrent user
report throughput executing identical Cognos report workloads. Competitor configuration: HP DL380p, 24 cores, 256GB RAM, Competitor row-store
database, SuSE Linux 11SP3 (Database) and HP DL380p, 16 cores, 384GB RAM, Cognos 10.2.1.1, SuSE Linux 11SP3 (Cognos). IBM configuration:
IBM S824, 24 cores, 256GB RAM, DB2 10.5, AIX 7.1 TL2 (Database) and IBM S824, 16 of 20 cores activated, 384GB RAM, Cognos 10.2.1.1, SuSE
Linux 11SP3 (Cognos). Results may not be typical and will vary based on actual workload, configuration, applications, queries and other variables in a
production environment.
82x calculation based on geometric mean calculation giving equal weighting to the report per hour (RPH) improvements in the three categories of
simple, intermediate, and complex reports. GEOMEAN(RPH_simple,RPH_intermediate,RPH_complex) = GEOMEAN(18.85,40.07,747.63)=82.66