1
Rich Niemiec, Rolta (Thanks: Zhong Yang, Steven Lu, Jim Viscusi, Milton Wan, Damon Grube, Mike Messina,
Sri Avantsa, & Shyam Varan Nath (+ Oracle Learning Library for Examples)
Exadata 101 The Rise of the Machines!
2013 – Rochester, New York
2
Rich’s Overview
• Advisor to Rolta International Board
• Former President of TUSC – Inc. 500 Company (Fastest Growing 500 Private Companies)
– 10 Offices in the United States (U.S.); Based in Chicago
– Oracle Advantage Partner in Tech & Applications
• Former President Rolta TUSC & President Rolta EICT International
• Author (3 Oracle Best Sellers – #1 Oracle Tuning Book for a Decade): – Oracle Performing Tips & Techniques (Covers Oracle7 & 8i)
– Oracle9i Performance Tips & Techniques
– Oracle Database 10g Performance Tips & Techniques
• Former President of the International Oracle Users Group
• Current President of the Midwest Oracle Users Group
• Chicago Entrepreneur Hall of Fame - 1998
• E&Y Entrepreneur of the Year & National Hall of Fame - 2001
• IOUG Top Speaker in 1991, 1994, 1997, 2001, 2006, 2007
• MOUG Top Speaker Twelve Times
• National Trio Achiever award - 2006
• Oracle Certified Master & Oracle Ace Director
• Purdue Outstanding Electrical & Computer and Engineer - 2007
3
Audience Knowledge
• Manager or Tech?
• Exadata V1 Experience?
• Exadata V2-2 or V2-8 Experience?
• Exadata V3-2 Experience?
• Oracle Cloud Control 12c Experience?
• Goals – Overview & Tour of Oracle EM Cloud Control 12c
– Focus on a few nice tuning features of Oracle EM Cloud Control 12c
• Non-Goals – Learn ALL aspects of Tuning with Oracle EM Cloud Control 12c
– Learn how to install Oracle EM Cloud Control 12c 11g Grid
There is a fifth dimension beyond that which is know to the CEO,
CFO, or CIO. It is a dimension that as vast as Exadata disk space,
and as untimely as an infinite loop for those that fail to embrace it. It
is the middle ground between light and shadow, between science and
superstition, between developer and DBA, and it lies between the pit
of the DBA’s fears of whether a recovery will work, and the summit of
his knowledge required for a successful upgrade. This future
dimension was once the dimension of imagination only, where we
could dream of a single database filled with millions of terabytes of
data and media. A wonderland where developers could write any
query without penalty of being timed out, or being the victim of an
anonymous kill -9; it’s a place where speed and disk space no longer
matter (as much). This has now become a reality, and I call it the
“Exabyte Zone!” - Oracle Database 11g Performance Tuning Tips & Techniques
5
Overview
• Terminology & the Basics about Exadata (X2-2, X2-8)
• Flash Cache
• Storage Index
• Smart Scans
• Hybrid Columnar Compression (HCC)
• Enterprise Manager & Grid Control
• Enterprise Manager Exadata Simulation
• I/O Resource Manager
• Security, Best Practices
• Exalogic Elastic Cloud
• Exadata Storage Expansion Rack (July 2011)
• Oracle Exalytics Business Intelligence Machine (October 2, 2011)
• SPARC SuperCluster & Oracle Database Appliance (Sept. 2011)
• Other Hardware, Exadata X3-2 (9/2012), 12c Database / PDBs
1968: E. F. “Ted Codd”
Invents Relational Theory in his mind
“The SEQUEL/DML paper got accepted to 1974 SIGMOD. Several years later I got a call from a guy named Larry Ellison who’d read that paper;
he basically used some of the ideas from that paper to good advantage.”
– Don Chamberlin, then IBM (SQL Reunion, 1995)
“In fact, when I started Oracle, the goal was never to have a large company. At best, I hoped we would have fifty people in the company and make a good living. About five years into the company, it became pretty clear that the horizons were unlimited. The only limitations were us.” – Larry Ellison (Nicole Ricci Interview, 1998)
1977: Oracle Begins as SDL
Software Development Laboratories
8
“I remember seeing the Oracle system running for the first time. Larry knew about System R and about our work and he gave me a little demo. I was impressed, because it was obviously simple. It seemed fast. He loaded the database, queried it, and updated it, all in a few seconds. It was - I don't know how many - maybe five-hundred records. And it loaded instantly. The thing that impressed me the most was that it ran on a little DEC PDP-11. The machine looked to be the size of a carton of cigarettes. It must have been an LSI-11 version of the machine, if my recollection of the size is correct. And System R at the time in most of our joint studies and at IBM was running on 168s. Now a 168 is only maybe the power of a 486DX2 or something, but the fact of the matter is it was a huge machine which would probably not fit in this room (water cooled).”
- Mike Blasgen, IBM System-R Team
1979/1980: SIGMOD Conference
9
“I thought, "Simple, fast, cheap; that's neat. People will buy it."
- Mike Blasgen, IBM System-R Team
1979/1980: SIGMOD Conference
10
Fast Forward
3 Decades –
With Exadata …=>
(Addressable Memory):
1983 – First 32-bit
(V3: 4G possible)
1995 – First 64-bit
V7.3: (16E possible)
2009: Oracle Buys Sun
(Hardware Accelerates)
11
All Data in the world 2010 = 1000E or 1Z
2K – A typewritten page 5M – The complete works of Shakespeare 10M – One minute of high fidelity sound 2T – Information generated on YouTube in one day 10T – 530,000,000 miles of bookshelves at the Library of Congress 20P – All hard-disk drives in 1995 (or your database in 2010) 700P –Data of 700,000 companies with Revenues less than $200M 1E – Combined Fortune 1000 company databases (average 1P each) 1E –Next 9000 world company databases (average 100T each) 8E – Capacity of ONE Oracle11g Database (CURRENT) 12E to 16E – Info generated before 1999 (memory resident in 64-bit) 16E – Addressable memory with 64-bit (CURRENT) 161E – New information in 2006 (mostly images not stored in DB) 1Z – 1000E (Zettabyte - Grains of sand on beaches -125 Oracle DBs) 100TY - 100T-Yottabytes – Addressable memory 128-bit (FUTURE)
The Future: 8 Exabytes
Look what fits in one 11g Database!
Bigger Data – Get Ready for it…
Worldwide, data is growing rapidly*:
2000: 800 Terabytes (1012)
2006: 160 Exabytes (1018)
2009: 500 Exabytes (just Internet)
2012: 2.7 Zettabytes (1021)
2020: 35 Zettabytes …?
Data generated in ONE day*….?
Twitter: 7 TB
Facebook: > 10 TB
Big data: The next frontier for innovation, competition, and productivity
McKinsey Global Institute 2011
Brain: 2.8 x 1020 bits of Memory
Space – John von Neumann, Harvard * Data collated from various online sources
How Much Data …
• 2004 monthly internet traffic >1E; 2010 it was 21E/month.
• In 2012, 2.5E data created every day (about 1Z=1000E per year)
• June 2012 – Facebook has 100P Hadoop cluster
• Facebook: 500T processed daily – (210T/hr DWHSE scanned)
• A Single Jet Engine – 20T/hour –same rate as Facebook!
• Gmail has 450 million users
• Wal-Mart – 1 million customer transactions/hour (2.5P DB)
• Large Hadron Colider produced 13P in one year
• Business data doubles every 1.2 years
• 19% of $1B companies have >1P of data (31% in 2013)
• 2011 – First Exabyte tape library from Oracle
• Decoding Human Genome took 10 yrs; Now takes a week!
14
Goals …
• Goals
– Overview of Exadata – Easy
– Exadata Fit in Oracle Strategy
– Some Tech & Grid Examples
– Exalogic, Exadata Expansion Rack
• Non-Goals
– Making you the Expert
15
Computing has Shifted from
Monolithic to Decentralized
Source: Forrester Research, Inc. - 2002
16
Acceleration! Next Generation Hardware
• Exadata/Exalogic from Oracle
• EMC V-Max
• IBM pureScale (for DB2)
• Flash Arrays (Violin Memory)
• Fusion-IO
Where to find Gold … X marks the spot!
Future Goal is to do this for Others:
Applications Acquisitions
Not to be confused with… Fusion Middleware/BI Acquisitions:
18
How BIG Oracle is Getting - OW
19
How about the Oracle JAVA World?
20
14 Storage Servers
- 14x12=168 Disks
- 100T SAS or
- 504T SAS
- 22.4TB+ flash storage!
8 Compute Servers
• 8 x 2 sockets x 8 cores = 128 cores
• 2T DRAM
InfiniBand Network
• 40 Gb/sec each direction
• Fault Tolerant
Exadata (X3-2)
(Oracle’s picture)
21
Exadata is MORE than Hardware*
1 TB
with compression
10 TB of user data
Requires 10 TB of IO
100 GB
with partition pruning
20 GB
with Storage Indexes
5 GB
with Smart Scans
Sub second
On Database
Machine
Data is 10x Smaller, Scans are 2000x faster
*Oracle Slide – Thanks!
22
Audience Experience?
Exadata V1? Exadata X2-2? X3-2?
10x faster than any Oracle DW 5x faster than V1
23
Big Difference…
Much Improved!
Exadata V1?
Exadata X2?
Exadata X3?
24
Terminology & The Basics
25
Some Terms
• High Capacity 3T SAS Disk (504T)–
Big/Slower – Like a 33 1/3 <7200 RPM>
• Used to be 2T SATA=Serial Advanced
Technology Attachment
• High Perf. 600G SAS Disk (100T) –
Small & Fast – Like a 45 <15K RPM>
• SAS=Serial Attached SCSI (Small
Computer System Interface)
26
More SPEED Coming… Get Ready…
This guy does not ever slow down!!
27
WHAT is it (X3-2)?
• Hardware ready for an internal 8-Node RAC cluster
• All the CPU power you need (128 cores)
• Mega DRAM Server Memory (2T)
• Super-Mega Flash Memory (22.4T)
• Super fast interconnect (40Gb/s)
• 100T of SAS disk (45T useable)
• Database could be MUCH larger with compression!
• If you need it & can afford it – You want it!!
Introduction to RAC - Shared Data Model
Exadata puts it back into One Machine
Shared Disk Database
Shared Memory/Global Area
shared
SQL
log
buffer
. . . . . . Shared Memory/Global Area
shared
SQL
log
buffer
Shared Memory/Global Area
shared
SQL
log
buffer
Shared Memory/Global Area
shared
SQL
log
buffer
Instance 1 Instance 2 Instance N-1 Instance N
29
How BIG is it?
• 128 Cores (16 eight core CPUs) on compute servers +168 cores on storage servers = 296 cores total-full rack
• 2T server DRAM
• 22.4T of flash cache (100G/sec)
• 100T SAS disk (45T useable) –15K RPM (25G/s; 50K IO/s)
OR
• 504T disk space (224T useable) – SAS 7.2K RPM
• SAS High Performance (15K RPM) or SAS High Capacity (7.2K RPM)
• SAS=Serial Attached SCSI (Small Computer System Interface)
• DRAM – Dynamic Random Access Memory
30
How FAST can it be?
• ALL Disks Combined:
– SAS – 25G/s (50,000 IOPS = 300 IOPS x 12 disk x 14)
– SAS High Capacity – 18 G/s (28,000 IOPS)
– SATA (Original 2T drives) – 12 G/s (20,000 IOPS)
• ALL Flash Cache Combined (5.4G/s per cell):
– 75G/s (1,500,000 IOPS); <20x more random I/O; 2x seq)
• Max Data Bandwidth with Disk + Cache + Compress:
– 500G/s (10x compression)
• Data Load Rate:
– 16T/hour (Full Rack)
31
How FAST is it?
Compared to the competition & previous Oracle:
• 5 – 100x for Data Warehousing
• 20x faster for OLTP
• Also - Miscellaneous:
– Hot Swappable Redundant Power
– Each Database Server - Dual Port InfiniBand 40Gb/s card
– Database Servers have Disk Controller HBA (Host Bus Adapter) has 512M battery backed up cache
– Each DB Server has 4 x 1GbE interfaces & ILOM (Integrated Lights Out Management – Remote power on)
– Two 10G Ethernet ports (optical)
32
How they got these NUMBERS?
(FYI Only)
• 8 compute servers – 8 servers x 2 CPU sockets x 8 cores = 128 cores (Xeon E5-2690)
• 8 servers x 256G DRAM = 2T DRAM
• 14 Storage Servers total 336G DRAM = 2.3T+ Total DRAM
• 3 InfiniBand Switches x 36 ports = 108 ports
• 14 Storage Servers (100-504T) with Flash Cache (22.4T) – 400G x 4 banks = 1.6T flash cache per storage server
– 14 storage servers x 1.6T = 22.4T Flash Cache
– 12 disks per storage server x 14 servers = 168 disks
– 168 disks x 600G SAS = 101T High Performance SAS
– 168 disks x 3T SAS = 504T High Capacity SAS
– Additional total storage of 9.6T on Database Servers (300G drives)
• 14 storage servers x 2 six core L5640 = 168 additional cores
33
Compute Servers – Like 8 Node RAC!
• 8 compute servers (Intel Xeon E5-2690’s at 2.9 GHz) – 8 servers x 2 CPU sockets x 8 cores = 128 cores
• 8 compute servers x 256G DRAM = 2T DRAM
• 4 x 300G drives x 8 = 9.6T (in addition to storage servers)
x8
DRAM
34
Storage Servers – Full Rack
• 14 Storage Servers with Flash Cache
– 24Gx14= 336G of DRAM (in addition to database servers)
– 400G x 4 cards = 1.6T per storage server of flash cache
– 14 storage servers x 1.6T = 22.4T Flash Cache
– 12 disks per storage server x 14 servers = 168 disks
– Additional 168 CPU cores on the storage!
– 168 disks x 600G SAS = 101T SAS
– 168 disks x 3T SATA = 504T SAS
12 Disks
Hot Swappable
Flash
Cache
(400G each)
35
InfiniBand – 40Gb/s Each way
• 3 InfiniBand Switches x 36 ports = 108 ports
• Leaf and spine switches wired at factory depending on
needs and how many Racks you’ll have – careful!
36 Ports
36
Put it all together – Oracle’s picture
of the X3-2
14 Storage Servers
- 14x12=168 Disks
- 100T SAS or
- 504T SAS
- 22.4TB+ flash storage!
8 Compute Servers
• 8 x 2 sockets x 8 cores = 128 cores
• 2T DRAM
InfiniBand Network
• 40 Gb/sec each direction
• Fault Tolerant
37
NEW X3-2 - One more time…
How they got these NUMBERS?
• 8 compute servers – 8 servers x 2 CPU sockets x 8 cores = 128 cores (Xeon E5-2690)
• 8 servers x 256G DRAM = 2T DRAM
• 14 Storage Servers total 336G DRAM = 2.3T+ Total DRAM
• 3 InfiniBand Switches x 36 ports = 108 ports
• 14 Storage Servers (100-504T) with Flash Cache (22.4T) – 400G x 4 banks = 1.6T flash cache per storage server
– 14 storage servers x 1.6T = 22.4T Flash Cache
– 12 disks per storage server x 14 servers = 168 disks
– 168 disks x 600G SAS = 101T High Performance SAS
– 168 disks x 3T SAS = 504T High Capacity SAS
– Additional total storage of 9.6T on Database Servers (300G drives)
• 14 storage servers x 2 six core L5640 = 168 additional cores
38
OLD X2-2 - One more time…
How they got these NUMBERS?
• 8 compute servers (Intel Xeon X5675’s)
– 8 servers x 2 CPU sockets x 6 cores = 96 cores
• 8 compute servers x 96G DRAM = 768G DRAM
– (Expandable to 1.1T Total DRAM)
• 3 InfiniBand Switches x 36 ports = 108 ports
• 14 Storage Servers with 168 CPU cores & Flash Cache
– 96G x 4 banks = 394G DRAM per storage server
– 14 storage servers x 394G = 5.376T Flash Cache
– 12 disks per storage server x 14 servers = 168 disks
– 168 disks x 600G SAS = 101T SAS (High Performance)
– 168 disks x 3T SATA = 504T SAS (High Capacity)
The X3-2 is much more than X2-2 …
• 4x more Flash Memory
• 20x increase in Write Performance (Smart Flash Cache Write-Back – could age out in months/years)
• 33% more Data Throughput
• 10 – 30% more Power Savings (-3KW per rack)
• 33% faster CPUs & 75% more Memory
• Same price except 6-core to 8-core software increase
• Can expand V2 or X-2 (½ or ¼ rack & add X3-2)
39
(Oracle’s picture)
*** If on 11.2.3.2+ … Does not require
Database/ASM / Cluster upgrade
40
X3-2: Full Rack or ½ or ¼ or 1/8
Full Half Quarter Eighth
Memory/DRAM 2T 1T 512G 512G
Servers/cores 8/128 4/64 2/32 2/16
Servers/disks** 14/168 7/84 3/36 3/18
Flash Memory 22.4T 11.2T 4.8T 2.4T
SAS (High Performance) 100T 50T 21.6T 10.8T
SAS (High Capacity) 504T 252T 108T 54T
Flash IOPs 1,500,000 750,000 375,000 187K
InfiniBand Switches 3 3 2 2
Data Load Rates 16T/hr* 8T/hr* 4T/hr* 2T/hr * Estimate ** 600G High Performance SAS or 3T High Capacity SAS; X3-2 has 1/8 Rack, the X2-2 does not!
41
Exadata X-3-8: In-Memory Database
4 T DRAM / 22 T Flash Cache
42
How will NEW 3-8 change these …
How they got these NUMBERS?
• 2 compute servers (E7-8870 CPU at 2.4 GHz & 5T SAS)
– 2 servers x 8 CPU sockets x 10 cores = 160 cores
• 2 compute servers x 2T DRAM = 4T DRAM
• 3 InfiniBand Switches x 36 ports = 108 ports
• 14 Storage Servers with 168 CPU cores & Flash Cache
– 400G x 4 banks = 1.6T DRAM per storage server
– 14 storage servers x 1.6T = 22.4T Flash Cache
– 12 disks per storage server x 14 servers = 168 disks
– 168 disks x 600G SAS = 101T SAS
– 168 disks x 2T SATA = 504T SAS
43
Where did all my disk space go?
• Lost Space:
– 100T SAS = 45T usable
– 504T SAS = 224T usable
• Apply some compression & get it back:
– 45T usable x 10 = 450T SAS
(High Performance SAS)
– 224T usable x 10 = 2.24P SAS
(High Capacity SAS)
44
-The Past is History
- The Future is a Mystery
- Today is a Gift
- That’s why they call it the present!
Where are YOU in History…
45
Some Case Studies
• Australian Finance Group – Process broker commission from 37 hours to 9 hours; Improved Siebel CRM 8X
• Turkey Mobile operator (97% of country) – Compression 40T to 10T
• Banca Transilvania– 30x faster queries; 30% energy savings
• Bokwang Family mart – Process 900,000 orders from 50 minutes to 7 minutes; Order filtering from 15 minutes to 30 seconds
• Enkitec – Consolidate 20 servers (two racks) into one Exadata saving an 100k power/cooling retrofits
• Finansbank – Used advanced indexing/compression to reduce 18T to 9.5T; DWHSE load from 341 to 250 minutes. Daily backups instead of a limited weekend only backup.
46
Some Case Studies
• Hiscom – Cut data processing speed (data loaded from 6,000+ Coca-Cola machines) by 3,000% by consolidating Oracle and Teradata into Exadata. Also eliminated Teradata license costs.
• LinkShare – Reduced data center floor space and power by 400%; 8x increase in query speed while reducing servers and storage by eight-fold; Use Enterprise Manager for tuning/monitoring
• Polk – 10x faster queries; Reduced storage with 10-15x compression; Partitioning for performance improvements.
• SoftBank Mobile – 8x faster queries; The system is processing one trillion items in a few seconds; Customer logs analysis went from 25 hours to 7 hours.
47
Some Case Studies
• Sogeti USA – Apps migrated to Exadata; Reduced weekly tape backup from 4 hours to a disk backup in 5 minutes with replication to the cloud. Reduced the number of servers and consolidated databases.
• Turkcell – 1.5 billion call records a day; Used HCC to go from 250T to 25T; Reduced report time from 27 minutes to 3 minutes; Reduced 10 storage cabinets to 1 Exadata rack.
• Yamazaki Baking – 3.5 M transactions per day. Moved from 9i to 11g Exadata (32 bit to 64 bit) and improved performance 30x. Some reports used to never finish at month end – they now finish – smart scans were a huge boost.
• Thomson Reuters – 2 large UNIX servers to 1 Exadata – 11x faster
• BNP Paribas - 4 large UNIX servers to 1 Exadata – 17x faster
48
How Oracle saved $1B:
CONSOLIDATION! & Process
Process
$750M
Technology
$200M
People
$50M
49
Proof of Concept – Wait Events
Our own ¼ Rack Exadata!
• Rolta POC –Consistent Improvement; Reduced waits by over 50x in some cases.
• Testing included disparate workloads with consistent improvement across all tests. Time spent waiting on I/O improved drastically. Better I/O in turn lowered or eliminated other wait events
50
Proof of Concept – Features
Our own ¼ Rack Exadata!
• Rolta POC –Advanced Compression and query parallelism boosted performance anywhere from 13x to 1700x faster (different query sets on x-axis).
51
Proof of Concept – Run Time
Our own ¼ Rack Exadata!
• Rolta POC – Exadata specific SQL Tuning, enabling compression and Parallel Query features further improved run times on both long and short/quick running queries.
• The improvements were 400% – 700% range
52
What’s Making it FAST?
• Fast Hardware!
• Many CPUs
• Flash Cache
• Lot’s of DRAM (Parallel Query in DRAM in 11.2)
• Smart Scan (save 4x-10x)
• Storage Indexes (save 5x-10x)
• Compression (save 10x-70x)
• Partition Pruning (save 10x-100x)
• Turn a 1T search into a 500M search or even 50M
53
Smart Scans
54
Smart Scans – 10x savings common
• HARDWARE Scans with NO Code Change:
– Filters based on WHERE clause (predicates)
– Filters on row / column / join condition
– Incremental Backup Filtering
• Works with:
– Uncommitted data
– Locked rows
– Chained rows
– Compressed Data
– Encrypted Data (11.2)
• You can SEE the benefit with Grid Control (OEM)
55
Oracle performance test…
• Without Smart Scan (Push whole table via network)
– 5T Table Scan
– Network bandwidth (40Gb/s) slows things
• 40Gb/s = 5GB/s; with 14 storage cells = 0.357GB/s each
– 16 minutes, 40 seconds (5T/5GB/s)
• With Smart Scan (Limit first at hardware level)
– 5T Table Scan
– Limit result BEFORE it hits the network
• Effectively scan 21GB/s (1.5G/storage cell * 14 cells)
– 3 minutes, 58 seconds (5T/21G/s)
56
The SMART Flash Cache
ALL Flash Cache Combined (5.4G/s per cell): 75G/s (1,500,000 IOPS)
20x more random I/O; 2x more sequential I/O (vs. disk)
57
Flash Cache – 20x-50x faster than disk
• Caches HOT Data – Does as LAST step!
• PCIe based Flash cards (PCI = Peripheral Component Interconnect express)
• Knows which objects NOT to cache (FTS)
• Can specify WHAT you want to cache
– STORAGE (CELL_FLASH_CACHE KEEP)
– Table/Partition level with CREATE or ALTER
• Write through caches is used to accelerate reads – Data written to disk also written to cache for future reads.
58
Flash Cache
• Caches
– Hot Data/Index Blocks
– Control File reads/writes
– File header reads/writes
• Does NOT cache
– Mirror copies / Backups / Data Pump
– Tablespace Formatting
– Table Scans (rare)
24G x 4 doms = 96G (dom = disk on module – “solid state”)
400G x 4 flash cards = 1.6T per storage server of flash cache
14 storage servers x 1.6T = 22.4T Flash Cache
59
Flash Cache LRU
• CELL_FLASH_CACHE storage clause
– DEFAULT (normal – large I/O’s not cached)
– KEEP (use flash cache more aggressive / May not occupy > 80%)
– NONE (flash cache not used)
• CACHE (NOCACHE) Hint
– I/O cached/not-cached in the flash cache
– SELECT /*+ CACHE */ …
• EVICT Hint – Data removed from the flash cache
• ASM rebalance data is evicted from cache when done
• Large I/O (Full Table Scans) on objects with CELL_FLASH_CACHE set to DEFAULT are not cached
60
Using the KEEP cache
ALTER TABLE CUSTOMER
STORAGE (CELL_FLASH_CACHE KEEP);
Table Altered.
SELECT TABLE_NAME, TABLESPACE_NAME,
CELL_FLASH_CACHE
FROM USER_TABLES
WHERE TABLE_NAME = ‘CUSTOMER’;
TABLE_NAME TABLESPACE_NAME CELL_FL
------------ ------------------- -------
CUSTOMER R_TEST KEEP
61
How it works…
• DB Request comes to CELLSRV (Cell storage server)
• CELLSRV (first time) gets data from disk – Data cached based on settings, hints … etc.
– Data to WRITE may also be cached after written if it is deemed that it may be needed again.
• CELLSRV (next time) checks: – In Memory Hash Table that lists what is cached
– If cached – goes to flash cache
– In not cached …may cache based on settings…etc.
• CELLCLI> list flashcache detail (allows you to monitor)
• CELLCLI> list flachcachecontent where ObjectNumber=62340 detail
(Select DATAOBJ# =from obj$ where name = ‘CUSTOMER’;)
62
Is it working for me…
SELECT NAME, VALUE
FROM V$SYSSTAT
WHERE NAME IN (
‘physical read total IO requests’,
‘physical read requests optimized’);
Name Value
---------------------------------------------- --------
physical read total IO requests 36240
physical read requests optimized 23954
(this second line (*8192) is flash cache used)
63
It IS working … 4G query
SELECT NAME, VALUE, VALUE*8192 VALUE2
FROM V$SYSSTAT
WHERE NAME IN (
‘physical read total IO requests’,
‘physical read requests optimized’);
NAME VALUE VALUE2
--------------------------------- -------- --------
physical read total IO requests 10,862,844 88,988,418,048
physical read requests optimized 2,805,003 22,978,584,576
run2.....
physical read total IO requests 11,320,185 92,734,955,520
physical read requests optimized 3,203,224 26,240,811,008
run4 .....
physical read total IO requests 11,993,845 98,253,578,240
physical read requests optimized 3,793,000 31,072,256,000
64
It IS working … V$SQL
Select sql_text, optimized_phy_read_requests, physical_read_requests, io_cell_offload_eligible_bytes
from v$sql
where sql_text like '%FIND YOUR SQL%'
SQL_TEXT OPTIMIZED_PHY_READ_REQUESTS PHYSICAL_READ_REQUESTS
------------ --------------------------- ----------------------
IO_CELL_OFFLOAD_ELIGIBLE_BYTES
------------------------------
SELECT.... 567790 688309
4.2501E+10
Run 2.....
SELECT… 762747 906729
4.9069E+10
run 4 ....
SELECT... 1352166 1566537
6.8772E+10
65
Storage Indexes (11.2)
** Thanks Oracle for this image
66
Storage Index - 10x is common (11.2)
• Storage Indexes maintain summary information about the data– (like Meta Data in a way)
• A CELL LEVEL (storage) Memory Structure
• Groups things into Min/Max for various columns
• Eliminates I/Os where there is no match
• Transparent to the user
• Done at the hardware level
• Typically one index for every 1M of disk
• NOT like a B-Tree Index…more like partition elimination to skip data NOT meeting conditions
• 100% done by Oracle – NO COMMANDS NEEDED!!
67
Is it working for me…
SELECT NAME, VALUE
FROM V$SYSSTAT
WHERE NAME LIKE (‘%storage%’);
NAME VALUE
--------------------------------------------- -------
cell physical IO bytes saved by storage index 25604736
(actual savings from Exadata built storage index)
68
Check BOTH servers…
SELECT NAME, VALUE
FROM GV$SYSSTAT
WHERE NAME LIKE (‘%storage%’);
NAME VALUE
--------------------------------------------- -----------
cell physical IO bytes saved by storage index 19693854720
cell physical IO bytes saved by storage index 0
(actual savings from Exadata built storage index)
69
Case two – More advanced comparing
ORDERED data to NON-ORDERED
SQL> desc table_no_order
Name Null? Type
----------------------------------------- -------- ----------------------------
OBJECT_ID NUMBER
OBJECT_NAME VARCHAR2(128)
SQL> desc table_ordered
Name Null? Type
----------------------------------------- -------- ----------------------------
OBJECT_ID NUMBER
OBJECT_NAME VARCHAR2(128)
70
Case two – More advanced comparing
ORDERED data to NON-ORDERED
select count(*)
from table_no_order;
COUNT(*)
--------------
482246656
select count(*)
from table_ordered;
COUNT(*)
--------------
482246656
alter system flush buffer_cache;
System altered
71
Non-Ordered does not use Storage Index
select count(*)
from table_no_order
where object_id=20;
COUNT(*)
--------------
32768
Elapsed: 00:00:01.64
select name ,value from v$mystat m,v$statname n
where m.statistic#=n.statistic#
and n.name like '%storage%‘
and m.value>0;
no rows selected
alter system flush buffer_cache;
System altered
72
Over 10x faster… uses Storage Index
select count(*)
from table_ordered
where object_id=20;
COUNT(*)
--------------
32768
Elapsed: 00:00:00.11
select name ,value from v$mystat m,v$statname n
where m.statistic#=n.statistic#
and n.name like '%storage%‘
and m.value>0;
NAME VALUE
---------------------------------------------------------------- --------------
cell physical IO bytes saved by storage index 1.5211E+10
73
Hybrid Columnar Compression (11.2)
74
Hybrid Columnar Compression
2. Stored in Compression Units
(Better compression when column data stored together)
1. Column Data Compressed
(Archive)
(Warehouse)
** Thanks Oracle for these images
75
Exadata Hybrid Columnar Compression
(EHCC) – 4-10x & 30x is common
• What is it (a HYBRID of column & row storage)?
– Data organized by column and compressed vs. row
– Tables organized in Compression Units (CU)-1000 rows?
– CU’s span many blocks (32K)
– Good for data bulk loaded (not for OLTP – single block)
• What’s it for?
– Query Data / DWHS (NOT frequently Updated)
• How much does it compress (old OLTP was 2-3x)?
– 10x in a typical data warehouse compression; (we got 4-11)
– 15x to 70x in archive compression (cold data); (we got 32)
76
Hybrid Columnar Compression
• Faster Operations: Query runs without decompression
– Compressed/Processed in FLASH CACHE; lower I/O!
– Compressed when sent over InfiniBand!
– Cloned compressed!
– Backed Up compressed!
– Scans MUCH less (compressed) data
• Worth Noting:
– Use standard table compression for OLTP
– Single block lookup FASTER than other columnar storage
– Updated rows migrate to normal / lower level compression
77
Hybrid Columnar Compression
• Fully supported:
– B-Tree Indexes
– Bitmap Indexes
– Text Index
– Materialized Views
– Partitioning
– Parallel Query
– Data Guard Physical Standby
– Logical Standby and Streams (FUTURE release)
– Smart Scans of HCC tables!
78
Other Oracle Compression
• Data Pump Compression
– Compression = {ALL | DATA_ONLY | NONE}
• RMAN Backup Compression
– Compression Level LOW/HIGH (New in 11.2)
• Secure File Compression
– LOW/MEDIUM/HIGH (2-3x compression)
– Deduplication & Encryption
• Normal OLTP Table Compression (since 9.2)
– 11g now supports INSERT/UPDATE
– FASTER Algorithm
• Data Guard Redo Transport Compression
79
Also remember … bad queries still bad!
Insert row by row…
declare v_t_object_id number; v_t2_object_id number; v_t_object_name varchar2(100); cursor mycur is select object_id from t2; begin open mycur; loop fetch mycur into v_t2_object_id; exit when mycur%notfound; begin select distinct t.object_id, t.object_name into v_t_object_id, v_t_object_name from t where t.object_id=v_t2_object_id; if SQL%ROWCOUNT =1 then insert into t3 values ( v_t_object_id,v_t_object_name ); end if; exception when no_data_found then null; end; end loop; close mycur; commit; end; /
80
Performance Improved by 60x
In Exadata… write smart queries!
alter session enable parallel dml;
Session altered.
SQL> @row-by-row
PL/SQL procedure successfully completed.
Elapsed: 00:03:44.94
Insert all at once:
insert into t2
select distinct t.object_id, t.object_name
from t,t2 where t.object_id=t2.object_id;
98 rows created.
Elapsed: 00:00:04.95
81
Benefits Multiply*: Access 1/2000th the data; It’s
like getting 8P memory resident in 4T of an X3-8
1 TB
with compression
10 TB of user data
Requires 10 TB of IO
100 GB
with partition pruning
20 GB
with Storage Indexes
5 GB
with Smart Scans
Sub second
On Database
Machine
Data is 10x Smaller, Scans are 2000x faster
*Oracle Slide – Thanks!
82
Enterprise Manager &
Grid Control for Exadata
83
Cloud Control 12c – Monitor Exadata
84
Cloud Control 12c – Monitor Exadata
85
SQL Performance Analyzer
12c – Exadata Simulation
Upgrade
Options
86
Exadata Simulation
87
Resource Management (IORM)
(FYI Only)
88
IORM - I/O Resource Management
• Set I/O resources for different instance
– Instance A = 50%
– Instance B = 30%
– Instance C = 20%
• Further set I/O based on users and tasks
– Instance A Interactive = 50%
– Instance A Reporting = 25%
– Instance A Batch = 15%
– Instance A ETL = 15%
• Best Solution for MIXED workloads & many instances
89
DBRM – Database Resource Manager
• Enhanced for Exadata
• Allows management of inter and intra DB I/O
• Inter-DB – Managed via IORM & Exadata storage software
• Intra-DB - Managed via Consumer Group
• CPU
• Undo
• DOP (Degree of Parallelism)
• Active Sessions
90
Grid Control - Resource Manager
91
Security – FYI Only
92
Oracle Audit Vault
Oracle Database Vault
DB Security Evaluation #19
Transparent Data Encryption
EM Configuration Scanning
Fine Grained Auditing (9i)
Secure application roles
Client Identifier / Identity propagation
Oracle Label Security (2000)
Proxy authentication
Enterprise User Security
Global roles
Virtual Private Database (8i)
Database Encryption API
Strong authentication (PKI, Kerberos, RADIUS)
Native Network Encryption (Oracle7)
Database Auditing
Government customer
Oracle Database Security Built over MANY years...
2007+
1977 *Oracle Slide – Thanks!
93
Security
• Audit Vault
• Total Recall / Flashback
• Database Vault
• Label Security
• Advanced Security
• Secure encrypted backup (also available: incremental
backup with Change Tracking File – much faster)
• Data Masking
• Data Guard
• Failure Groups – (automatic-for storage cell failure)
Enhanced security of Audit Data with new
AUDIT_ADMIN role
• Also SYSBACKUP privilege (don’t need SYSDBA for RMAN)
• Update strong user authentication using kerberos
• Simplified Vault administration
Security Enhancements in 12c
95
Best Practices
96
MUST haves & DON’T do!
• Must have Latest Bundle Patch (See note: 888828.1 for latest)
• Must have the correct data center COOLING! – 3 tiles with holes for full rack (400 CFM/tile) – don’t melt it!
• Must have the correct power needs
• Must use Oracle Linux 5.8 (x86_64) or Solaris 11 (Selectable at install time) & Oracle11.2
• Must have ASM & use RMAN for backups
• Consider StorageTek SL500 Tape backup
• Use an ASM allocation unit (AU) size of 4M
• Don’t add any foreign hardware or – No Support!
• Don’t change BIOS/Firmware or – No Support!
97
Best Practices
• Create ALL celldisk and griddisks
• Use DCLI to run on ALL Storage Servers at once
• Use IORM
• Decide Fast Recovery Area (FRA) & MAA Needs
• Database 11.2.0.1+ (11.2.1.3.1) and ASM 11.2.0.1+
• COMPATIBLE 11.2.0.1+
• Logfile size at 32G (Whoa!)
• LMT (Locally Managed Tablespaces) with at 4M uniform extents
• Move Data with Data Pump (or use INSERT /*+ APPEND */)
98
It’s the Real Deal!
Oracle: Aggressive!
• Fast Hardware!
• Many CPUs!
• Fast Flash Cache!
• Lot’s of DRAM on Database Servers and Storage
• Compression (save 10x-70x)
• Partition Pruning (save 10-100x)
• Storage Indexes (save 5-10x)
• Smart Scan (save 4-10x)
• Turn a 1T search into a 500M search or even 50M
99
Exadata = Paradigm Shift!
100
March 4, 1986 – Sun
(Stanford University Network)
101
March 12, 1986 – Oracle
ORCL IPO:
Open:15
Close:20.75
Up 38%
102
March 13, 1986 – Microsoft Next?
103
Oracle Taking over Hardware…
** Thanks Oracle for this image
1/8 Rack
104
What’s Next – Exalogic Elastic Cloud!
Built for Applications Tier (Note: There is a ½, ¼ & 1/8 Rack)
• Some points here – Oracle is leveraging those acquisitions! – Coherence is a great product / NEW Linux – Unbreakable
Enterprise Kernel! Tuxedo is part of X3-2.
– X2-2:- 360 CPU cores, 2.9T DRAM, 4T FlashFire SSD Read Cache, 60T SAS – Will help Fusion Apps Smoke!
– X3-2:- 480 CPU cores, 7.68T DRAM, 4T FlashFire SSD Read Cache, 60T SAS – Will help Fusion Apps Smoke!
– 1M HTTP/sec (X2-2) – When released, it could fit Facebook on 2 of these even thought there were 500M people on Facebook
105
What’s Next –
Exadata Storage Expansion Rack X3-2! (Note: Also ½ & ¼ Rack, and 600G drives)
• Oracle is leveraging those acquisitions!
– 18 storage servers (216 3T drives): 648T raw disk (288T useable) 6.75T Flash Cache, 216 CPU cores (22.8T Flash Cache in the X3-2)
– InfiniBand connected.
– 1.9M Flash read IOPS, 32K disk IOPS
– On-disk backup – 27T/hour (Great for older data / images)
8 Full Racks = 5.2 Petabytes of raw disk!
(with compression store even more!)
106
What’s Next –
Exadata Storage Expansion Rack X3-2!
• Exadata Storage Server Software:
– Smart Scan Technology (even Data Mining Scores)
– Smart Flash Cache
– Storage Indexes
– Hybrid Columnar Compression
– IORM / DBRM
– Smart Scans of Data Mining Model Scoring
– ASM (Automatic Storage Management)
– Backup with RMAN
– Restores using Flashback Technologies
– Redundant power & InfiniBand Switches
107 * Thanks Oracle for images (Original ODA above and New X3-2 ODA below)
108
PE
RF
OR
MA
NC
E
CAPACITY
HIG
HE
R
HIGHER
Oracle Systems Family*
(Updated for the ODA 3-2 & X3-2)
Oracle Exadata Quarter Rack (X3-2)
32 Database Cores
3 Exadata Storage Servers
72 TB Storage
1.1 TB Smart Flash Cache
Smart Scan / Storage Indexes
Hybrid Columnar Compression
Large Flash I/O (4.8T)
Fully Expandable
Oracle Database Appliance 2 cores
Oracle Database Appliance 32 cores
Oracle Database Appliance (3-2)
2 to 32 Cores
Cores can be disabled
18 TB Storage** (36T Possible)
800 GB Flash*** (for Redo Logs)
One Button Deployment, Patching, and Support
* This Slide (edited by me) from Oracle Corporation – Thanks!
** Note that the disk speed was 15K RPMs on the previous ODA and only 10K RPMs on the ODA 3-2
*** Note that the Flash is put in the expansion area (could limit disk space)
108
4.8
109
Oracle Database Appliance Small Business (SMB) & Departmental Focus
• 2 cores to 32 cores (Xeon) – 2 core minimum/4 core for RAC
• Cluster in a box (Oracle Clusterware comes installed)
• 256G RAM x 2 servers = 512 G DRAM (+800G redo Flash)
• 36 T* of raw storage (up to a 12T data warehouse / 3x mirror)
– *Standard is 18T or 36T with optional storage expansion shelf!
• Software on this is 11gR2 / ASM / RAC / Oracle Linux
• Oracle Appliance Manager (Patching / Managing… - NEW)
• High Availability Fault Tolerant! 2 nodes / dual server
• Ready to go in a couple of hours (software & RAC preloaded)
• Auto Memory Management, Auto Tuning, Auto Disk Backup
• Phone home … calls to get service; One button patching
110 This Slide from Oracle Corporation – Thanks!
Oracle Appliance Manager – Configurator
111
“Pay as You Grow” Database Pricing (Subject to Change – Check with Oracle)
• Old single Hardware System SKU: $50,000**
• (New Version is 60K + 40K (expansion shelf))
• Standard Oracle software licensing applies
– Starting software list price still: $47,500
• “Pay as You Grow”
– CPU cores can be powered down, and don’t have to be
licensed. You can use the cores you need and align
software license costs with business requirements
– You do this through a special area of My Oracle Support
– NEW in 3-2 –Use other CPUs for other tasks** This Slide from Oracle Corporation edited – Thanks!
** Check with your Oracle Sales Rep to Confirm this previous price! You do this using virtualization.
112
ODA or ¼ Rack Exadata? (Comparison on previous ODA & X2-2)
• Depends on your needs…
• If you need less than 7000 IO’s/sec then ODA (previous version of ODA was around 4000 IO’s/sec)
• Need less than 5 G/s (was 3 G/s) throughput then ODA
• ODA is around ½ the price fully loaded and well less than ½ the price if you don’t need 24 CPU’s
• If you need > 7000 IO’s/sec or > 5G/s throughput or need to grow greater than 32 CPU’s then you need Exadata!
• ¼ Rack Exadata was about 4 – 20x faster than a fully loaded ODA in the last versions depending on workload & other factors. The 1/8th rack is half the size of ¼ rack.
113
What Early Customers Are Saying …
• "At Beijing Cable TV, we need to be able to deploy highly available database solutions quickly with minimum resources. The Oracle Unbreakable Database Appliance was extremely simple to deploy and manage. We were able to set up a highly available database solution in less than an hour. We also like the fact that the appliance is a complete Oracle engineered system, which includes two server nodes, software, networking, and storage, in a single box, so we don’t need to work with multiple hardware and software vendors.“
– Jianlian Wu, Vice Chief Engineer, Beijing Cable TV
This Slide from Oracle Corporation – Thanks!
114
What’s up Next …
Exalytics Business Intelligence Machine!
** Thanks Oracle for this image
115
Oracle Exalytics BI Machine!
Leveraging the Times Ten Acquisition
• 1T DRAM
• 3.6T storage uncompressed
• 40 Gb/s InfiniBand
• 4 x 10 core = 40 cores
• 200G/sec scan x 5x compression = 1T/sec scan
• Runs Times Ten, OBIEE, Memory Optimized Essbase…
• In memory data access & columnar compression
• Smart Storage Manager to keep the right things in memory
• Software seems pricey – but, huge impact possible!
“It’s so fast the response time is … well there is no response time, it’s just done.”
116
What’s Next … SPARC SuperCluster!
** Thanks Oracle for this image
117
SPARC SuperCluster T4-4 The Future is coming faster…
• New SPARC T4 microprocessor
• Runs 5x faster than T3 microprocessor
• Exadata Flash Disk Storage System
• Solaris 11 & 10 8/11 Operating System
• Oracle VM for SPARC
• Targeting SPARC Solaris install base to migrate to SPARC SuperCluster
• Run Oracle Database Applications faster and less expensive than anything available from IBM. (from Oracle Earnings announcement call Sept. 2011)
118
The new T4 – Why it’s SOOOOO fast!
• 5x faster than T3
• L1/L2 cache specific to a core – 16K each cache
• L3 shared by all 8 cores – 4M (new on T4)
• Prefetching of instructions & data (new on T4)
• Out of order execution (new on T4)
• Dual instruction use (new on T4)
• Memory Management Unit page size = 2G
• Dynamically threaded (see next slide)
• 14 on chip crypto functions & 10GbE networking
• Up to 8 Racks w/o adding any InfiniBand changes
119
8 core T4 processor has 64 threads
Up to 2 threads run simultaneously
** Thanks Oracle for this image 8 cores x 8 threads each = 64 threads x 4 T4’s x 4 servers
120
SuperCluster … 1.2M IO’s/sec
How they got these NUMBERS? (Note: There is also ½ Rack now)
• 4 compute servers (SPARC T4 CPU at 3 GHz & 4.8T SAS) – 4 servers x 4 CPU sockets (T4) x 8 cores = 128 cores
– 8 threads per core 128 cores = 1024 threads (1200 threads per box)
• 4 compute servers x 1T DRAM= 4T DRAM (19.2T disk)
• 3 InfiniBand Switches x 36 ports = 108 ports
• 8 Storage Servers with additional cores & Flash Cache – ZFS 7320 controllers with 2 servers (two 4-core CPU & 2x10 disks)
– 9.6T Flash Cache
– 10x2 = 20 disks x3T (in ZFS) + 6x4 = 24 disks x 600G (in Servers) = 14.4T SAS
– 12 disks per storage server x 8 servers = 96 disks (4 disks are 18G)
– 72 disks x 600G SAS + 20 disks x 3T = 103.2T Mix (117T total including the 14.4G)
– Or: 92 disks x 3T SAS High Capacity = 276T SAS (290T total including the 14.4G)
• 42 GB/s storage bandwidth
SuperCluster about to get Faster?
The new T5 and new T5-4 (& T5-8)
121
• New SPARC T5
• Runs 2.3x faster than T4
• New T5-5
• Similar to T4-5
• Has the T5 processor
• If SuperCluster replaces T4-4’s with T5-4’s it will be: • 4 servers x 4 CPU sockets (T5) x 16 cores = 256 cores • 8 threads per core 256 cores = 2048 threads (About 2200 threads per box = DOUBLE THE CURRENT!)
• 4 compute servers x 2T DRAM= 8T DRAM (19.2T disk)
Double the CPU & Double the DRAM if SuperCluster gets it!
122
2K – A typewritten page 5M – The complete works of Shakespeare 10M – One minute of high fidelity sound 2T – Information generated on YouTube in one day 10T – 530,000,000 miles of bookshelves at the Library of Congress 730T – Information generated in YouTube in a year 20P – All hard-disk drives in 1995 (or your database in 2010) 700P –Data of 700,000 companies with Revenues less than $200M 1E – Combined Fortune 1000 company databases (average 1P each) 1E –Next 9000 world company databases (average 100T each) 8E – Capacity of ONE Oracle11g Database (CURRENT) 12E to 16E – Info generated before 1999 (memory resident in 64-bit) 16E – Addressable memory with 64-bit (CURRENT) 161E – New information in 2006 (mostly images not stored in DB) 1Z – 1000E (Zettabyte - Grains of sand on beaches -125 Oracle DBs) 100TY - 100T-Yottabytes (1000Z=1Y) – Addressable 128-bit (FUTURE)
The Future: 8 Exabytes
Look what fits in one 11g Database!
All Data in the world 2010 = 1000E or 1Z
123
What’s Next – World On-Line in Oracle Cloud
Super Oracle System!
• Oracle’s email system - 9 Exadata Racks
• What’s possible: – 32 x X3-8’s (64 node RAC cluster)
• 10,240 CPUs on the compute servers
• 5,376 CPUs on the storage servers
– 2048 x X3-2 Storage Expansion Racks • Roughly 1327 PB of raw storage
• That’s 13.27 Exabytes compressed (at 10x)
• To Exceed the 8E Oracle Maximum (mirrored) – 32 X3-8’s (10.7 PB)
– 4,096 Storage Expansion Racks (13+Exabytes at 10x)
– 5,000+ years of YouTube storage (2.55P/year)
– 70x Compression will give you 90+ Exabytes
– 64 bit will allow 16E to be in memory
124
Exadata V2 / V2-8 & Exalogic
125
Exadata V2 / V2-8 & Exalogic
126
Exadata V2 / V2-8 & Exalogic
The X3-2 is much more than X2-2 …
• 4x more Flash Memory
• 20x increase in Write Performance (Smart Flash Cache Write-Back – could age out in months/years)
• 33% more Data Throughput
• 10 – 30% more Power Savings (-3KW per rack)
• 33% faster CPUs & 75% more Memory
• Same price except 6-core to 8-core software increase
• Can expand V2 or X-2 (½ or ¼ rack & add X3-2)
127
(Oracle’s picture)
*** If on 11.2.3.2+ … Does not require
Database/ASM / Cluster upgrade
128
SuperCluster
129
Headlines: Oracle is now a very serious
hardware company!
130
Big Revenue & Big Profits Coming…
Need more hardware … ?
131
Pillar (SAN), ZFS, Big Data, StorageTek
from www.oracle.com
HUGE iBridge Sale– RJN NOTE
132
Pillar, ZFS (NAS or SAN), Big Data,
StorageTek
133
Pillar, ZFS, Big Data, StorageTek
(TAPE) – Store 1E (string of 10 w/ 2:1)!
134
135
Pillar, ZFS, Big Data, StorageTek
648T Raw Storage – RJN NOTE
Oracle Big Data Solutions
In-Database MapReduce (12c)
137
Oracle Public Cloud
138
The Oracle Social Network
Trends: Gartner Hype Cycle 2012
140
More SPEED Coming… Get Ready…
This guy does not ever slow down!!
141
Get Ready for Pluggable Databases
This guy and his team working hard to make your life easier!
What is your System of the Future?
142
143
Rich Niemiec, Rolta (www.rolta.com) (Thanks: Andy Mendelsohn, Debbie Migliore, Maria Colgan, Penny Avril, Jacob Niemiec , & Lucas Niemiec)
Oracle Disclaimer: The following is intended to outline Oracle's general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle's products remains at the sole discretion of Oracle.
The Best Oracle 12c New Features
(see 12c presentation for more)
144
Overview – 12c
• Know the Oracle!
• Start Me Up – Using Memory Target, The Buffer Cache & The Result Cache
• Invisible Columns (12c) & virtual columns (11g)
• Multiple indexes on the same Column (12c) & Invisible Indexes (11g)
• Adaptive Execution Plans (12c) & Adaptive Cursor Sharing & Bind Peeking (11g)
• Runaway query Management (12c)
• Change Table Compression at import Time (12c) & (Partition Compression – 11g)
• Create Views as Tables (12c)
• Online Move Partition (12c) & Interval Partitioning (11g)
• Partial Indexes for Partitioned Table (12c)
• Pluggable Databases (12c)
• Enhanced DDL Online (12c)
• Exadata and Big Data (In-Database MapReduce in 12c)
• Consolidated Database Replays & Better Reporting (12c)
• Automatic Diagnostics Repository (12c)
• Security Enhancements (12c)
• Other 12c New Features
145
Exadata X-3: In-Memory Database
4 T DRAM / 22 T Flash Cache
Compelling Technology Statistics!
0
5
10
15
20
25
30
35
40
Radio TV Cable Internet Wireless
Years to Reach 50M
Users
147
Friedman’s 6 Dimensions
of Understanding Globalization*
• Politics (Merging)
• Culture (Still disparate)
• Technology (Merging/Merged)
• Finance (Merging/Merged)
• National security (Disparate)
• Ecology (Merging)
* Sited from Mark Hasson, PSU, Global Pricing and International Marketing.
148
Waves of Acceleration!
Country Time to Oust Ruling Communist Govt.
Poland 10 Years
Hungary 10 Months
E. Germany 10 Weeks
Czechoslovakia 10 Days
Romania 10 Hours
“Human history becomes more and more a race between education and disaster.” - HG Wells
149
NEW Waves of Acceleration!
Country Time to Oust Ruling Dictators or Monarchy
Iraq ?? Years (with help)
Afghanistan ?? Years (with help)
Egypt Months
Tunisia Months
Libya Months
Yemen ?? Months
Syria ?? Months or Years
Saudi Arabia ?? Years or months
Iran ?? Decades or weeks
150
Where are YOU in History…
-The Past is History
- The Future is a Mystery
- Today is a Gift
- That’s why they call it the present!
-Melissa
151
2K – A typewritten page 5M – The complete works of Shakespeare 10M – One minute of high fidelity sound 2T – Information generated on YouTube in one day 10T – 530,000,000 miles of bookshelves at the Library of Congress 20P – All hard-disk drives in 1995 (or your database in 2010) 700P –Data of 700,000 companies with Revenues less than $200M 1E – Combined Fortune 1000 company databases (average 1P each) 1E –Next 9000 world company databases (average 100T each) 8E – Capacity of ONE Oracle11g Database (CURRENT) 12E to 16E – Info generated before 1999 (memory resident in 64-bit) 16E – Addressable memory with 64-bit (CURRENT) 161E – New information in 2006 (mostly images not stored in DB) 1Z – 1000E (Zettabyte - Grains of sand on beaches -125 Oracle DBs) 100TY - 100T-Yottabytes (1000Z) – Addressable in 128-bit (FUTURE)
The Future: 8 Exabytes
Look what fits in one 11g Database!
152
• All databases of the largest 1,000,000 companies in the world (3E). or • All Information generated in the world in 1999 (2E) or • All Information generated in the world in 2003 (5E) or • All Email generated in the world in 2006 (6E)
or • 1 Mount Everest filled with Documents (approx.)
8 Exabytes:
Look what fits in one 11g Database!
153
With Java…
Now Oracle goes Commercial… ?
What’s Next …
154
155
What’s comes after the Exadata Zone?
YOU will soon be in for more…
Directly Addressable Indirect/Extended
4 Bit: 16 (640)
8 Bit: 256 (65,536)
16 Bit: 65,536 (1,048,576)
32 Bit: 4,294,967,296
64 Bit: 18,446,744,073,709,551,616 … quantum leaps are next!
• Qubits allow multiple states so that you can look at all of the possibilities/probabilities at one time.
• The “Quantum Zone” next (Quantum Physics is incomplete – Einstein)
– Just 512 qubits would store 512-bits of addressable memory or 2512 (which is well over a googol or 1 with 100 zero’s after it – a googol is about 2332).
– Brush up on your Eigenvectors, Eigenvalues, Pauli Matrices & Grover’s Algorithm
– Create Singularity … all atoms of a person by 2045 (I think earlier); 12-Monkeys
– Private universes – Is there one for each person? (Schroeder’s cat – I think not)
– Rearranging atoms to create new objects; Nanotech + Quantum Physics coming!
156
What’s comes after the Exadata Zone?
YOU will soon be in for more…
Consider 3-bit or 23 (Addressable memory is 3)
000 100
001 101
010 110
011 111
With just 3 cubits I get (ket notation) – looking at many states at once:
a|000> + b|001> + c|010> + d|011> + e|100> + f|101> + g|110> + h|111>
For the state (a,b,c,d,e,f,g,h): Note that: |010> = (0,0,1,0,0,0,0,0)
As we move from zeros and ones to cubits represented by the Bioch Sphere… many challenges occur including quantum decoherance and issues mapping entanglement & superposition to computing.
157
158
Not a tech person…
Know the Tech Mind!
159
Before Tech they were Engineers
160
This is How DBA View Themselves!
161
How everyone else views them!
Do users think of this when they think of their DBA?
Data
162
1970’s/1980’s High School –
Classic Under-Achievers
Future DBA Future Developer
Future Manager says…
Try Oracle…
163
Fast Forward to 2013…
Current Manager says…Try Exadata X3-2…
164
Teach your kids Oracle… 4:33 AM
165
Don’t want to Embrace the Future???
166
Summary – We Covered…
• Terminology & the Basics about Exadata (X3-2, X3-8)
• Flash Cache
• Storage Index
• Smart Scans
• Hybrid Columnar Compression (HCC)
• Enterprise Manager & Grid Control
• Enterprise Manager Exadata Simulation
• I/O Resource Manager
• Security, Best Practices
• Exalogic Elastic Cloud
• Exadata Storage Expansion Rack (July 2011)
• SPARC SuperCluster & Oracle Database Appliance X3-2 (March 2013)
• Oracle Exalytics Business Intelligence Machine (October 2, 2011)
• Summary
167
• Oracle 11g Release 2
Performance Tuning Tips &
Techniques; Richard J.
Niemiec; Oracle Press
(Available now)
“If you are going through hell, keep going” - Churchill
For More Information
168
#1 Selling Oracle Database Book on
Amazon since it came out in February!
• Also available at other
places like Barnes &
Noble…etc.
• Available on the Kindle
and other book readers
• Why is it #1?
169
References to wish for…
170
更多信息
• www.tusc.com
• Oracle9i Performance Tuning
Tips & Techniques; Richard
J. Niemiec; Oracle Press (May
2003)
• Oracle 10g Tuning (June 11,
2007)
“成功只访问那些没空追求它的人。”
- Henry David Thoreau
171
References
• Exadata V2 – Sun Oracle DB Machine, Oracle
• Oracle Exadata Implementation Workshop, Oracle Corporation, McLean, Virginia - Multiple Exadata sessions
• Oracle Learning Library – multiple sessions/topics
• Oracle 11g R1/R2 & Oracle 12c Best Features, Rich Niemiec
• Oracle Enterprise Manager Deployment and High Availability Best Practices, Jim Viscusi (Oracle Corporation), Jim Bulloch (Oracle Corporation), Steve Colebrook-Taylor (Barclays Global Investors)
• Oracle11g Performance Tuning Tips & Techniques, Rich Niemiec, Oracle Press McGraw-Hill
• Advanced Compression with Oracle Database 11g Release 2, Oracle Corporation, Steven Lu
• Tech Crunch
• Twilight Zone Series
• Rod Serling; Submitted for Your Approval, American Masters
• YouTube/oracle Oracle OpenWorld On Demand
172
173
V$ View Poster – Booth 1355
174
Rolta– Your Partner ….
Accomplished in Oracle!
2012 Oracle Excellence Award
(9 Partner of the Year / Titans / Excellence Awards)
Prior Years Winner 2002, 2004*, 2007*,2008,2010, 2011 *Won 2 Awards
175
How to Make a Difference in the World!
176
Rolta’s Oracle Services
• Oracle
– E-Business Suite implementation, R12 upgrades, migration & support
– Fusion Middleware and Open Systems development
– Business Intelligence (OBIEE) development
– Hyperion Financial Performance Management
– DBA and Database tactical services
– Strategic Global Sourcing
• IT Infrastructure
– IT Roadmap - Security & Compliance - Infrastructure Management
– Enterprise Integration / SOA - High Availability and Disaster Planning
• Profitability & Cost Management
– Financial Consolidation - Budgeting & Forecasting
– Profitability & Risk Analysis - Enterprise Performance Management
– Operational, Financial & Management Reporting
• Rolta Software Solutions
– iPerspective™ - rapid data & systems integration
– Geospatial Fusion™ - spatial integration & visualization
– OneView™ - business & operational intelligence
177
Copyright
Information
• Neither Rolta nor the author guarantee this document to be error-free. Please provide comments/questions to [email protected]. I am always looking to improve!
• Rich Niemiec/ Rolta © 2013. This document cannot be reproduced without expressed written consent from Rich Niemiec or an officer of Rolta TUSC, but may be reproduced or copied for presentation/conference use.
Contact Information
Rich Niemiec: [email protected]
www.rolta.com
178
Rich’s Overview…
• Advisor to Rolta International Board
• Former President of TUSC – Inc. 500 Company (Fastest Growing 500 Private Companies)
– 10 Offices in the United States (U.S.); Based in Chicago
– Oracle Advantage Partner in Tech & Applications
• Former President Rolta TUSC & President Rolta EICT International
• Author (3 Oracle Best Sellers – #1 Oracle Tuning Book for a Decade): – Oracle Performing Tips & Techniques (Covers Oracle7 & 8i)
– Oracle9i Performance Tips & Techniques
– Oracle Database 10g Performance Tips & Techniques
• Former President of the International Oracle Users Group
• Current President of the Midwest Oracle Users Group
• Chicago Entrepreneur Hall of Fame - 1998
• E&Y Entrepreneur of the Year & National Hall of Fame - 2001
• IOUG Top Speaker in 1991, 1994, 1997, 2001, 2006, 2007
• MOUG Top Speaker Twelve Times
• National Trio Achiever award - 2006
• Oracle Certified Master & Oracle Ace Director
• Purdue Outstanding Electrical & Computer and Engineer - 2007