Practical experiences and best practices for SSD and IBM i

IBM

What’s the value and how to optimize
Performance with SSD and IBM i

Gottfried Schimunek
3605 Highway 52 North
Senior Architect Rochester, MN 55901
Application Design
IBM STG Software Tel 507-253-2367
Development Fax 845-491-2347
Lab Services
Gottfried@us.ibm.com
IBM ISV Enablement

Power is performance redefined ©2012 IBM Corporation

Acknowledgements

Thanks to Mark Olson, IBM WW Product Manager for providing
this comprehensive and detailed presentation

2 Power is performance redefined ©2012 IBM Corporation

Agenda

SSD introduction/overview

Implementing hot/cold

Performance insights

Roadmap for Power Systems SSD

SSD reference material


Solid State Drive (Flash Technology)

Processors Memory Disk
SSD

Very, very, Very, very, Very, very slow
very, very, very fast comparatively
very fast Fast

1,000,000 -
< 10’s ns ~100 ns ~200,000 ns
8,000,000 ns
Access Speed

There is a huge gap between memory and disk speed


New Solid State Drive (Flash Technology)

SSD

very fast Fast

1,000,000 -
< 10’s ns ~100 ns ~200,000 ns
8,000,000 ns
Access Speed

~33 ~ 12.5
~1 second
minutes hours

Human Time Context


Basic Problem --- Disk “Slowing” Down (Relatively)

Seagate 15k RPM/3.5" Drive Specifications

+35% 450
Capacity (GB)

Max Sustained
171 DR (MB/s)
+15% Read Seek (ms)
73
75

3.6
-1% 3.4

2002 2008
Capacity growing ok (35% per year), but Read/Seek -1% & Data Rate only
15% per year
While processors & memory speed up and add threads and cache
Net … a growing imbalanced between disk and processor/memory


177GB eMLC SSD Performance

Latency -
IO OPERATIONS PER SECOND Throughput
Response
(IOPS) (MB/s)
Time (ms)
Random
Random Random Mixed Single
Power SSD Read Write
Read Write 70% Read / Read
30% Write

177GB 2.5” SSD
15 k 4k 11 k 170 MB 64 MB .25 ms
in SAS SFF bays

For grins … 0.12 - 0.4 0.12 - 0.4 0.12 - 0.4 ~175 ~200 8.3 – 2.5
15k rpm HDD k k k MB MB ms

Note these are drive specific measurements and projections which can vary from what you might experience. The values assume
528 byte sectors running RAID-0 with no protection. Hypothetically if measured with unsupported 512 byte sectors, values would be
higher. The values are highly workload dependent. Factors such as read/write mix, random/non-random data, drive cache
hits/misses, data compressibility in the drive controller, large/small block, type of RAID or mirroring protection, etc will change these
values. These values produced by a server with plenty of processor, memory and controller resources to push this much I/O into the
SSD. Most client system applications don’t push SSD nearly this hard.


SSD Price Performance

SSD

very fast Fast

$/GB $/IOP
Power 740 (Max 512GB memory)
One 32GB Memory feature = $6,390 One 15k HDD = ~200 IOPS*
= $199.69 per GB = $2.49 / IOPS
One 177GB SSD = $3,588 One 177GB SSD = ~11,000 IOPS*
= $20.27 per GB = $0.33 / IOPS

SSD only 10% per GB SSD only 13% per
cost of memory IOPS cost of HDD
IOPS values vary depending on many configuration and workload variables.
The above values are fairly conservative for SSD are represent a 70/30
(read/write) mix. 200 HDD IOPS is better than many HDDs experience.

Prices are suggested list prices for a Power 740 and are subject to change. Reseller prices may vary.


World Class eMLC SSD Performance

Latency -
IO OPERATIONS PER SECOND Throughput
Response
(IOPS) (MB/s)
Time (ms)
Random
Random Random Mixed Single
Power SSD Read Write
Read Write 70% Read / Read
30% Write

177GB 2.5” SSD
15 k 4k 11 k 170 MB 64 MB .25 ms
in SAS SFF bays
387GB 2.5” SSD
39 k 22 k 24 k 340 MB 375 MB .20 ms
in SAS SFF bays

For grins … 0.12 - 0.4 0.12 - 0.4 0.12 - 0.4 ~175 ~200 8.3 – 2.5
15k rpm HDD k k k MB MB ms

Note these are drive specific measurements and projections which can vary from what you might experience. The values assume
528 byte sectors running RAID-0 with no protection. Hypothetically if measured with unsupported 512 byte sectors, values would be
higher. The values are highly workload dependent. Factors such as read/write mix, random/non-random data, drive cache
hits/misses, data compressibility in the drive controller, large/small block, type of RAID or mirroring protection, etc will change these
values. These values produced by a server with plenty of processor, memory and controller resources to push this much I/O into the
SSD. Most client system applications don’t push SSD nearly this hard.


SSD Price Performance

SSD

very fast Fast

$/GB $/IOP
Power 740 (Max 512GB memory)
One 32GB Memory feature = $6,390 One 15k HDD = ~200 IOPS*
= $199.69 per GB = $2.49 / IOPS
One 387GB SSD = $6,200 One 387GB SSD = ~24,000 IOPS*
= $16.02 per GB = $0.26 / IOPS

SSD only 8% per GB SSD only 10% per
cost of memory IOPS cost of HDD
IOPS values vary depending on many configuration and workload variables.
The above values are fairly conservative for SSD are represent a 70/30
(read/write) mix. 200 HDD IOPS is better than many HDDs experience.

Prices are suggested list prices for a Power 740 and are subject to change. Reseller prices may vary.


Power eMLC SSD Performance versus 15k rpm HDD
Drive to Drive comparisons
SSD offers up to New eMLC SSD HDD is 12X – 41X
66x – 250x more I/O nearly 2X WRITE slower (latency)
Operations Per throughput (MB/s) than SSD
Second (IOPS) than HDD (milliseconds)
1.8X 41X
250X

1X
125X

12X
66X
33X
1X 1.25X 1X
HDD SSD SSD HDD SSD SSD HDD SSD SSD
177GB 387GB 177GB 387GB 177GB 387GB
Drive-to-drive compares ignore any caching by SAS controller

It All Stacks Up

Depending on the workload, the HDD
stack may be much higher on the left.

This shows a fairly conservative HDD
vs SSD performance comparison

PLUS … the PCI slots for the SAS
controllers make the HDD stack higher

1/10th Qty HDD

1/20th Qty HDD 1/20th Qty HDD

15k rpm HDD 177 GB SSD 387 GB SSD 387 GB SSD
(SFF) (SFF) (SFF) (1.8”)


Power Solid State Drive

SSD

very fast Fast

1,000,000 -
< 10’s ns ~100 ns ~200,000 ns
8,000,000 ns
Access Speed

Sweet spots
1. Batch window reduction for disk bound applications
You can cut up to 40-50% off window
2. Response time - transaction/data base for disk bound applications
Internal drives or perhaps even SAN drives
3. Analytics
Key points
-- A modest quantity of SSD can often make a big difference
-- Both write-heavy and read-heavy work can be fine for today’s SSD –
biggest performance boost for random read workload


SSD Client - Batch Window Reduction Example 1

Associated Bank needed to reduce month end
batch run time from 4+ hours to under 3
hours

SSDs cut 1.5 hours from batch run time 40% Reduction
Plus a 16% reduction in # of disk drives
Batch Performance Runs
# of SAS # of Batch 5
Disk Drives SSDs Run Time
4

Base run 72 0 4:22
Hours

3

2
SSD run 1 72 8 2:43
1

SSD run 2 60 4 2:48 0

72 HDD 72 HDD + 8 SSD 60 HDD + 4 SSD

Placed eight DB2 Objects (table, index, view)
on SSD

Source: IBM Power Systems Performance and Benchmark Center 5-23-09


SSD Client - Batch Window Reduction Example 2

50% Reduction
with SSD
Customer in health care industry Batch Windows
needed to reduce batch windows
significantly
Daily batch running 10+ hours

H o u rs
Monthly batch running 30+ hours
Added 12 SSDs to 168 HDDs
Cut 50% from daily run time
Cut 50% from monthly run time

. Monthly Daily


SSD Client – Data Warehouse Example

SAP Customer using Business 60-87% reductions with
Intelligence Warehouse (BW 7.0) SSD
installed an all-SSD configuration of
about 800GB.
4 days
Huge improvement in aggregate
builds / compression
Significant reductions in queries

87%
20 min
reduction
65%
reduction
60%
7 min reduction 1/2 day
5 min
2 min

HDD SSD HDD SSD HDD SSD

Sample query 1 Sample query 2 Sample build/compression


SSD Client – Deep Analytics

~85% reductions
with SSD
Client doing complex financial
48 hour
analysis of financial transactions Run
combined with WW economical
factors. Millions of records.
Insights guide client in making
tactical and strategic decisions. ~85%
reduction

Transformed application from
“interesting and helpful” to 6-7
hours
“critical and essential”

HDD SSD

Sample build/compression


SSD Faster Analytics Data
Warehouse

Faster answers
Faster/better decision making
--------------------------
SSD example: Leveraging DB2 temp space with
Iterative /
modest amount of SSD capacity
drill-down
First query draws data from HDD “relatively slow”, question
but data then cached on SSD. Second, third,
fourth, etc query MUCH faster !!!! 1
2
When query work of this data ceases, SSD space 3
reused for new queries. 4

Questions/analysis are usually iterative. The first
question generates additional drill-down or expanded
questions. The raw data for the first question gets
reused a lot until the topic exploration is completed.
Example – “what am i selling?” “what am i selling in
the evening?” “what am i selling in the evening by
store?” etc, etc

SSD Client Example – IPL Reduction

IBM Development has projected Less downtime
modest improvement for IBM i and
for AIX, but …..
IBM i POWER6 520 client with 16
drives (4 SSD and 12 HDD)
Client put load source on SSD
Now reporting 3 minute IPLs
POWER7 IBM i client also reporting 3
minute IPLs
Much faster PTF applications

What is your cost per minute or cost per hour for down
time? How long is your typical IPL?


Many Systems Buy “Extra” Disk Arms for Performance

HDD % utilization of capacity is held low to help ensure higher I/O performance
and more consistent response time - arm movement, spinning platter an issue
Write cache and read cache attempt to buffer the impact

Often less than 30-50% full for performance sensitive workloads

SSD % utilization of capacity not restricted

Can run much closer to 100% full


Mixed SSD + HDD Can be Great Solution
It is typical for data bases to have a large percentage of data which is infrequently
used (“cold”) and a small percentage of data which is frequently used (“hot”)

Hot data may be only 10-20% capacity, but represent 80-90% activity

SSD offers best price performance when focused on “hot” data
HDD offers best storage cost, so focus it on “cold” data …. a hierarchical approach

Cold May be able to use larger HDD and/or a larger % capacity used
Hot Can run SSD closer
to 100% capacity


Save Space/Energy/Cooling
A few SSD can often replace many HDD
Fewer total drives
Maybe fewer controllers
Fewer I/O drawers
Fewer cables

I/O drawer maintenance reduction
can offset SSD maintenance


A Power 740 Scenario (your situation will vary)
Acquisition
Power 740
Single LPAR
IBM i V6R1
9%
All HDD, 28 drives, mirrored
versus
Mixed HDD & SSD, mirrored
12 15k RPM HDDs
4 SAS-bay-based 177GB SSDs
All HDD Mix HDD & SSD
Achieves similar performance results
Maintenance Energy Rack "U's"
22
% 39 40
% %

All HDD Mix HDD & SSD All HDD Mix HDD & SSD All HDD Mix HDD & SSD


Should You Use SAN/SVC SSD or Internal SSD?

Pro’s Con’s

Use
Internal yes yes
SSD

Use
SAN/SVC yes yes
SSD

“Internal” = drives located in a system unit, in an I/O drawer with PCI slots or in a disk-only I/O drawer.


DAS & SAN - Two Good Options
Both options are strategic
Both options have their strengths
Can use both options on the same
server

DAS SAN
Direct Attached Storage Storage Area Network
(“internal”) (“external”)

• Fastest (lower • Fast
latency) • Multi-server sharing
• Typically lower cost • Advanced functions/values
Flash Copy, Metro/Global
hardware/software
Mirror, Live Partition Mobility,
• Often simpler config Easy Tier


SSD - SAN/SVC or Internal …. Or Mixed?
Pro’s Con’s
1. Lower latency - internal is about 1. If already have SAN/SVC, adds
0.3 milliseconds faster (per access complexity of having internal
Use .. it can add up) (can be larger) PLUS SAN/SVC
2. If don’t already have SAN/SVC,
Internal lower cost & less learning
2. Reduces value of SAN/SVC high
function capabilities – Flash
SSD 3. Multiple Config options Copy, MetroMirror, HA, partition
4. Handle just like other internal mobility, etc
storage
1. Leverage SAN/SVC well-known 1. If don’t have SAN/SVC, need
value proposition of initial SAN/SVC investment in
Use manageability, control, hardware, software, skills
functionality, performance, etc 2. Many medium/entry SAN/SVC
SAN/SVC 2. If already have SAN/SVC, less don’t offer SSD
SSD learning, reduced cost to add
3. DS8000 & V7000 EasyTier

MIXED: If using a SAN/SVC which doesn’t have SSD, OR if not using the higher
function capability of a SAN/SVC (for example you do backup/restore functions just like
it was an internal drive); THEN combining internal Power Systems SSD with SAN/SVC
storage can be a good thing.
“Internal” = drives located in a system unit, in an I/O drawer with PCI slots or in a disk-only I/O drawer.

SSD Technology “Myth”

“Beware SSD as they wear out as they have a limited number of
writes”
Key facts to correct myth …. this is not a concern for IBM Power
Systems SSD
IBM Power Systems SSD are industrial/enterprise drives designed to
handle this. Design points are many years of 24x7x365 heavy write
workloads. These are not “PC-grade” or “consumer-grade” flash drives.
Even if you do somehow “use up all the writes”, the SSD reports status prior
to it being a real problem to the server and server sends a message to the
operator or to IBM for a future scheduled repair action. This is just like a
disk drive reporting a weakening status. Replacement SSD is covered
under IBM Maintenance.
Just like disk drives, you want to protect contents using system mirroring or
RAID protection


Agenda







Implementing “Hot” and “Cold” Data

AIX IBM i
Ability to granularly/flexibly select and
locate hot files on SSD
#1 Best integrated, automated
Key AIX performance tools are filemon capability available in the industry
& iostat. today
Database vendors provide hot data “Trace and Balance” function part of
analysis tools, example: DB2 Snapshot. IBM i – automated in i 7.1
Migrating hot data Monitors by partition or ASP (Aux
Migratepv is often useful Storage Pool) to determine hot/cold
Softek Migration Tool provides a nice Upon request, automatically moves hot
suite of functions and can be combined data to SSD, cold to HDD
with IBM Services Can re-monitor & rebalance any time
See Total Storage Productivity Center A few key OS files can automatically
SSD white paper for additional be placed on SSD
insights Can specify specific data base
New/enhanced tools being developed objects to be placed on SSD
See white paper for additional insights
Additional enhancements being
developed


IBM i Load Balancer

Industry leading automated capability
Monitors partition/ASP using “trace” IBM i intelligent hot/cold placement
makes a big difference vs normal IBM
User turns trace on during a peak time striping / scattering of data across all
User turns trace off after reasonable drives.
sample time
Negligible performance impact expected This example 72 HDD + 16 SSD
Tool monitors “reads” to identify hot data
Looks at 1MB stripes of data
Upon command, automatically moves hot
Application Response time

data to SSD, cold data to HDD 72 HDD + 16 SSD No Balance
72 HDD + 16 SSD Data Balanced
Minimal performance impact, done in
background
Can remonitor and rebalance any time
Probably a weekly or monthly activity Trans/min

Perhaps less often if data not volatile

Predicting/analyzing what % of data is hot for presale analysis to
help size the number of SSD required:
• Use PEX tool/output
• Output from monitor (will need technical person to interpret)


IBM i SSD Balancing One of 5 ASP Balancing Types
Balance data between busy units and idle units (STRASPBAL
TYPE(*USAGE))
Make all of the units in the ASP have the same percent full
(STRASPBAL TYPE(*CAPACITY))
Drain the data from a disk, to prepare unit it to be removed from
the configuration (STRASPBAL TYPE(*MOVDTA))
(Almost obsolete) move hot data off of a compressed disk, and
move cold data to the compressed disk (STRASPBAL
TYPE(*HSM)) Requires specific disk controllers with compression
capability – feats #2741/2748/2778. Compression only allowed in user ASPs

Move cold data to HDDs and move
hot data to SSDs
(STRASPBAL TYPE(*HSM))


Using TRCASPBAL to place hot data on SSDs – IBM i
HDD1 HDD2 HDD3 HDD4 SSD
100 500 100 2000 6000
10000
100 1200 6000
900 3000 10000
100
300 800 500 900 6000
900
0 4000
300 300 400 4000
300
200 600 700 1000 100
100

Trace ASP balance counts the read operations based
on 1MB stripes
TRCASPBAL SET(*ON) ASP(1) TIMLMT(*NOMAX)
Start ASP balance moves the data
STRASPBAL TYPE(*HSM) ASP(1) TIMLMT(*NOMAX)
Target is 50% of read operations to be on SSD
Cold data is moved (multiple threads) to HDDs, hot data is
moved (single thread) to SSD


How to Find Hot Tables and Indexes - IBM i

Performance Explorer
BY FAR the best solution
Perform analysis based on read complete and write
complete events
DB2 maintains statistics about the number of
operations on a table or index
Statistics are zeroed on each IPL
Statistics only identify candidates (logical operations include
both random and sequential operations)
Available via:
Display file description (DSPFD)
Application programming interface (API) QUSRMBRD
System i Navigator Health Center (V6R1 only)
SQL catalog queries


Agenda







Summary Performance Comparisons of 3 SSD Options

SSD PCIe-based SAS-Bay-Based

GB capacity 177 GB 69 GB 177 GB

Flash Technology eMLC SLC eMLC

SSD drive Base compare Similar, may be
(reference a little slower in
(no controller Similar
point) write-heavy
considerations) usage

35 Power is performance redefined * Assuming same number of drives per controller/adapter ©2012 IBM Corporation

Performance Comparisons of 69/177GB SSD Options
SSD PCIe-based SAS-Bay-Based
GB capacity 177 GB 69 GB 177 GB
Flash Technology eMLC SLC eMLC
SSD only performance (no
Base compare
controller) read heavy Similar, but can vary a little
(reference point)
Similar, but can vary a little
workload
SSD-only performance (no Typically somewhat slower
Base compare
controller) write heavy Similar
(reference point) (0-15%)
workload (due to add’l function in SSD)

Controlled by
Tied directly to ONE controller Controlled by #5805/5903
#5805/5903 or by #5906
with no write cache with max 4 or by #5906 adapter up to
Controllers / adapters drives. RAID-5 slowed due to
adapter up to 9/8 drives
9/8 drives per
per controller/controller
no write cache. controller/controller pair
pair

SSD PLUS Controller
Base compare
performance Typically similar *
(reference point)
Typically similar
Read heavy workload
Similar throughput, but
SSD PLUS Controller potentially slower response Typically somewhat slower
Base compare (0-5% mixed HDD/SSD)
performance time due to no write cache * ---
(reference point)
-- (0-15% SSD only)
Write heavy workload (due to add’l function in SSD)
Bigger impact if RAID-5/6

* Assuming same number of drives per controller/adapter


Other Performance Insights – background reading
• Time to Write on a HDD is the same as time to read on an HDD (1:1 ratio)
• Time to Write on an SSD is much longer than a read on an SSD around (3:1 ratio)
• Thus the biggest SSD performance boost is for reads compared to HDD.
SSDs do reads at < 1 ms while from HDD are typically around 3-15 ms
• Writes to SSD are faster than writes to HDD …. BUT …. Write cache on a SAS controller is even faster than
a write to SSD. SSD can empty out the write cache faster than HDD, so SSD+write cache is generally faster
than HDD+write cache, but comparisons have more “it depends” caveats
• The size of information being handled makes a difference. Handling bigger blocks of data takes longer,
especially if doing writes.
• Randomness of information also plays a part. If there are a lot of smaller blocks of information which happen
to be close together where they will be eventually be stored, then they are more likely to be in the same page
of space in the SSD device’s internal controller. Thus the access and/or writes can happen “above” the real
flash memory and is faster.
• HDD have had ongoing cost/GB improvements over the last 5-10years, but have had little access
performance enhancements. Most performance boosts have been through enhancements in controllers of
HDD, not the spinning platter or arm. No significant HDD performance improvements are currently foreseen.
Thus the gap between server memory and HDD will continue to widen.
• If operating system can predict the data needed, it can issue a read prior to the application making the request
reducing HDD delay to application. But random reads by definition are not feasible to pre-fetch. Applications
with lots of random reads are thus THE ideal area for SSD benefits.
• Alternatives to SSD – 1) Can “pin” information in memory. This is even faster than SSD. But SSD can
provide a lot more GB of capacity and is usually more cost effective per GB. 2) Also could go buy a lot more
adapters with write cache and/or spread data out, but this is less cost effective than SSD.
• If an HDD fails, it can take hours to format and bring the new HDD into sync with either its mirrored pair or with
the rest of the RAID-5 array. The larger the drive’s capacity, the longer the resync time can be. Until re-
synced there is risk a second failure could crash the partition or application. If an SSD is replaced, the time to
resync so that the RAID array or mirroring protection is restored is MUCH, MUCH faster (minutes vs hours)
with SSD than HDD.
• For absolute best performance use internal SSD vs SAN SSD. There in about 0.3ms (3/10ths milliseconds)
additional latency in accessing SAN. For I/O intensive applications this adds up. But SANs can offer a lot of
other operational and managerial advantages for many clients. The performance trade off may be a good
thing to do in the overall decision.
• For VERY rough rules of thumb for mixing HDD & SSD – use about 5-15% of the GB of the data – or use
about 5-10% of the drives.


SSD Config Options --- April 2012

Power Systems (internal / DAS)
PCIe-based
SAN-based
SSD SAS-bay-based SSD
DS8000 #5805 #5913 #ESA1/A2 #5888 Ultra
#2053/54/55
SVC In CEC w/ int Gen1 PCIe Gen2 PCIe Gen2 PCIe Drawer
RAID & SSD
V7000 SAS contrlr 380MB 1800MB 0 MB 3100MB
SAS Adapter
XIV cache cache cache cache

Many DAS SSD config options* for Power Clients
Options vary
Performance
Price
Physical size
Where tested/supported
Function

* PCI-X 1.5GB SAS adapter not shown to
focus on most current technologies


Power DAS SSD Options: April 2012
PCIe-based
SSD SAS-bay-based SSD
In CEC w/ int #5805 & #5913 & #ESA1/A2 #5888 Ultra
#2053/54/55
SAS contrlr #5887*** #5887*** & #5887*** Drawer
Number PCIe slots used 2 (4 mirror) 0 2 2 2 0
Number GX slots used 0 0 0 0 0 1-2 (710/730=2)

Max SSD attach 4 3-8 mdl dependent 9 24 24 30
Max 177GB busy SSD W1 4 W1 2-3 W1 4-6 W1 ~24 W1 ~24 N/A
reasonably supported @ W2 3-4 W2 1-2 W2 3-4 W2 ~24 W2 ~24
Max 387GB busy SSD N/A W1 1-2 W1 2-3 W1 ~20 W1 ~18 W1 ~22
reasonably supported @ W2 1 W2 1-2 W2 ~14 W2 ~14 W2 ~14
Write cache (MB) 0 175 380 1800 0 3100
GB / SSD 177 177 or 387 177 or 387 177 or 387 177 or 387 387
Servers supported
- Newest 710-740 (C models) Y Y Y 710/730 limit** Y 710/730 limit** Y 720/740 limit * 710 740
- Rest POWER7 710-795 Y, except 795 Y, except 795 Y not 710/730 Y not 710/730 N N (SOD)
- POWER6 Y, except 595 N Y 177GB SSD Y 177GB SSD N N
AIX/IBM i/Linux support Y Y Y Y Y AIX / Linux
Mix HDD & SSD N Y N Y N N (SOD)
Rack space needed depends N/A 2U+ 2U+ 2U+ 1U
Easy Tier N N N N N N (SOD)
Approximate USA list price $3k + 2PCIe 0 $4.4k/pair + $15k/pair + $6.1k + 6.1k $24.5k + $2k
with zero SSD for Mdl 740 slots $5.4k drawer $5.4k drawer + $5.4k drwr PCIe adpter

PowerHA: share w/ 2 servers N N Y Y Y Y
@ This is a simple rule of thumb. Actual reasonable maximum depends on many factors.
Prices subject to * ESA1 not in 720/740 system unit. Can place in #5802/5877 I/O drawer attached to 720/740
change. Reseller ** Not in 710/730 system unit, but 730 (8231-E2C) can have in #5802/5877 I/O drawer
prices can vary. *** Possible to use #5802 or 5803 I/O drawer instead of #5887 EXP24S Drawer. Max 177GB SSD attach would differ for #5913 and ESA1/ESA2
W1 = transaction/command/CPW type workload, smaller block, IOPS sensitive RAID5
W2 = save/restore/large-file type workload, Throughput sensitive. RAID5 Assumes #5887 is in mode2 using two SAS ports for higher bandwidth.

Performance: PCIe-Based vs. 69GB SAS-Bay-Based SSD
Using 380MB Write Cache RAID Adapters running SAS-Bay-Based
(2) #5903/5278 adapters + (8) 69GB SSD (2) #2055 + (8) 177GB SSD
Running RAID-5 Running RAID-5
2 PCIe slots 4 PCIe slots

Similar performance – RAID-5 to RAID-5
10

SAS-Bay-
Appl Response time (ms)

1 Based PCIe-Based

0.1

0.01

SAS-bay-based
177GB SSD 0.001
0 20000 40000 60000 80000 100000

similar to 69GB Appl Trans/m in

SSD

Notes:
• Data points measured using an artificial transaction workload which IBM believes will reasonably represent many commercial client
workloads. Differences in application usage and data, for example the percentage of random reads vs percentage of writes can
impact this comparison. (Your results may differ.) The above measurement used 60% reads and 40% writes.
• Write cache was turned on for #5903/5278 adapters. #2055 has no write cache. .



(2) #5903/5278 adapters + (8) 69GB SSD (2) #2055 + (8) 177GB SSD
Running RAID-5 Running mirroring
2 PCIe slots 4 PCIe slots

Mirrored PCIe-based outperforms Medium cache RAID-5 SAS-Bay-Based
10
SAS-Bay- PCIe-Based
1
Based

0.1

SAS-bay-based 0.01

177GB SSD
0.001
similar to 69GB 0 50000 100000 150000 200000

SSD Application trans/min

Notes:
impact this comparison. (Your results may vary.) The above measurement used 60% reads and 40% writes.
• Write cache was turned on for #5903/5278 adapters. #2055 has no write cache. .


Using 1500MB Write Cache RAID Adapters running SAS-Bay-Based SSD

(1) #2055 + (4) 177GB SSD (1) #5904/6/8 adapters + (4) 69GB SSD
Running RAID-5 Running RAID-5
2 PCIe slots 2 PCI-X slots

Large cache PCI-X adapter & SSD outperforms PCIe-Based SSD
( RAID-5 to RAID-5 )
10
PCIe-Based

1

0.1

SAS-bay-based 0.01
SAS-Bay-
Based
177GB SSD
similar to 69GB 0.001

SSD 0 20000 40000 60000 80000

Application trans/min

Notes:
impact this comparison. (Your results may vary.) The above measurement used 60% reads and 40% writes.
• Write cache was turned on for #5904/6/8 adapters. #2055 has no write cache.


Performance: HDD vs. SSD+HDD Hybrid

(2) 5903/5278 + (18) HDD (2) #5903/5278 adapters ) + (18) HDD
Running mirrored + (2) #2055 + (8) 177GB SSD
Running mirrored
Hot data located on SSD

Adding SSD and placing hot data on SSD, grew workload capability by nearly 3X
Appl Response Time (ms) 10
HDD only

1
SSD+HDD hybrid

0.1

0.01

0.001
0 50000 100000 150000 200000 250000

Application Trans/min

Notes:
impact this comparison. The above measurement used 60% reads and 40% writes.
• Write cache was turned on for #5903/5278 adapters. #2055 has no write cache.


SSD Analyzer Package

Provides:
- SSD Analysis Overview
- Disk Read Time
- Job & Tasks by Read Time

Drilldown from here into CS &
GHI perspectives

Download from SSD Techdoc site:
http://www-03.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS3780


Disk Response Time Charts

Detailed charts are based upon the
increased detail QAPMDISKRB groups

Use tooltips to fly over & get
specific data


CSI – SSD Candidate Screening

Offered at 7.1 only, helps a user determine if SSDs could help
performance. It needs data from QAPMDISKRB.


CSI – SSD Candidate Screening

Offered at 7.1 only, helps a user determine if SSDs could help
performance.


CSI – SSD Candidate Screening job details report

This is similar to the green screen SSD Analyzer tool:
http://www-03.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS3780


Agenda







Power Systems SSD Statements of Direction
Provided August 2010

IBM plans to enhance its Power Systems Enterprise Class
SSD solutions with technology designed to continue to
provide significant improvements in performance and
storage density over time. IBM plans for these IBM Power
Systems enhancements to include both SAS-bay-based
and PCIe-based SSD product offerings that will leverage
IBM's investments in its SSD optimized Enterprise Class
RAID Storage Controllers.

All statements regarding IBM's future direction & intent are subject to change or withdrawal without notice, and represents goals &
objectives only. Information regarding potential future products is intended to outline our general product direction and it should not be
relied on in making a purchase decision. The information mentioned regarding potential future products is not a commitment, promise, or
legal obligation to deliver any material, code or functionality. Information about potential future products may not be incorporated into any
contract. The development, release, and timing of any future features or functionality described for our products remains at our sole
discretion.

Power Solid State Drives (SSD) Roadmap
2009
• Apr/May - Introduction SAS-bay-based
• July/Oct/Nov - Misc enhancements
• Nov/Dec - Huge price action – matching memory price action
2010
• Feb - support in Power 750, 755, 770, 780
• April - IBM i 7.1 enhancements
• August - Support in new POWER7 servers & AIX 7.1
enhancements
• August - Introduce additional SSD config option … PCIe-based
2011
• Enhancements to both SAS-bay-based and PCIe-based SSD
More capacity in SSD devices
More performance in SAS controllers
2012 … more enhancements

See SODs made August 2010. All statements regarding IBM's future direction & intent are subject to change or withdrawal
without notice, and represents goals & objectives only.


Roadmap: SSD Configuration 2011 Options

SAS-bay-based
SAS Bays
PCI SAS
SSD
SSD
SSD
SSD

SSD
SSD
SSD
SSD
SSD

SSD

controller

New PCIe card Larger capacity SSD
(pair for redundancy) Continue Hot plug SSD
Large cache, high performance More SSD modules / controller

PCIe-based
Double-wide PCIe card
PCIe SAS
SSD

(need pair for hot plug)
SSD

controller
SSD

Up to 4 SSD modules
SSD

See SODs made August 2010. All statements regarding IBM's future direction & intent are subject to change or withdrawal
without notice, and represents goals & objectives only.


Agenda







SSD Analyzer Tool for IBM i
• Quick, easy, no-charge analysis looks at standard performance report output
• Provides “probably yes”, “probably no”, or “maybe
• Provides rough estimate of quantity of SSD to recommend

SSD ANALYSIS TOOL (ANZSSDDTA)
Type choices, press Enter.
PERFORMANCE MEMBER . . . . . . . *DEFAULT__ Name, *DEFAULT
LIBRARY . . . . . . . . . . . __________ Name
Additional Parameters
REPORT TYPE . . . . . . . . . . *SUMMARY *DETAIL, *SUMMARY, *BOTH
TIME PERIOD::
START TIME AND DATE::
BEGINNING TIME . . . . . . . . *AVAIL__ Time, *AVAIL
BEGINNING DATE . . . . . . . . *BEGIN__ Date, *BEGIN
END TIME AND DATE::
ENDING TIME . . . . . . . . . *AVAIL__ Time, *AVAIL
ENDING DATE . . . . . . . . . *END____ Date, *END
NUMBER OF RECORDS IN REPORT . . 50__ 0 - 9999

Bottom
F3=Exit F4=Prompt F5=Refresh F12=Cancel F13=How to use this display
F24=More keys

Available via www.ibm.com/support/techdocs in “Presentations & Tools”.
Search using keyword SSD

Additional Marketing Materials

Brochure ….

Power System SSD Web pages
Positioning, educating, linking to other materials
www.ibm.com/systems/power/hardware/peripherals/ssd/index.html

Additional white paper information being worked on by Development –
availability dates TBD

List of existing SSD white papers follow on separate slides


SSD Brochure

Sales personnel can
access brochure through
SSI / PW web pages


Power Systems SSD Web Pages

Several informative
pages on SSD
positioned from a
Power Systems
perspective

http://www.ibm.com/
systems/power/hard
ware/peripherals/ss
d/index.html


Five Power Systems SSD White Papers

IBM Power SSD vs Consumer SSD (posted Nov 2009)
“Advantages of True Enterprise Solid State Drives (SSDs) in Enterprise
Systems”
AIX-specific (posted Apr 2009)
“Driving Business Value on Power Systems with Solid State Drives”
IBM i-specific (posted May 2009)
“Performance Value of Solid State Drives using IBM i”
First published May 2009
More SSD technology specific – AIX/IBM i/Linux appropriate (posted
Jun 2009)
“Performance Impacts of Flash SSDs Upon IBM Power Systems”
Above papers in Power Systems web site under “Resources/Literature”
http://www.ibm.com/common/ssi/apilite?infotype=SA&infosubt=WH&lastdays
=1825&hitlimit=200&ctvwcode=US&pubno=POW*USEN&appname=STGE_PO_
PO_USEN_WH&additional=summary&contents=keeponlit

5th paper for an SAP environment
http://www.sdn.sap.com/irj/sdn/db4?rid=/library/uuid/90a1637e-065f-2c10-
3ab7-bea9375fc88d


i Oriented References/Tool

IBM - Performance Management on IBM i Resource Library
http://www.ibm.com/systems/i/advantages/perfmgmt/resource.html
Performance Value of Solid State Drives using IBM i
http://www.ibm.com/systems/resources/ssd_ibmi.pdf
Performance Impacts of Flash SSDs Upon IBM Power Systems
http://www.ibm.com/common/ssi/cgi-
bin/ssialias?infotype=SA&subtype=WH&htmlfid=POW03028USEN&attachme
nt=POW03028USEN.PDF&appname=STGE_PO_PO_USEN_WH
Driving Business Value on Power Systems with Solid State Drives
http://www.ibm.com/common/ssi/cgi-
bin/ssialias?infotype=SA&subtype=WH&htmlfid=POW03025USEN&attachme
nt=POW03025USEN.PDF&appname=STGE_PO_PO_USEN_WH
IBM Systems Lab Services and Training
http://www.ibm.com/systems/services/labservices
IBM Power Systems(i) Benchmarking and Proof-of-Concept Centers
http://www.ibm.com/systems/i/support/benchmarkcenters


Agenda






Questions?


IBM

Tak

Dank je Gracia Merci Thanks
s

Grazie
Danke
Takk


IBM


Performance and Scalability Services
The IBM i Performance and Scalability Services Center can provide
facilities and hardware IN ROCHESTER to assist you in testing
hardware or software changes

“Traditional” benchmarks
Release-to-release upgrades
Assess application performance when migrating to a new release of IBM I
Stress test your system
Determine impact of application changes
Proofs of Concept (e.g. HA alternatives; SSD analysis, external storage, etc.)
Evaluate application scalability
Capacity planning

… all with the availability of Lab Services IBM i experts and
development personnel

63 To request any of these services, submit at:

IBM Systems Lab Services and
Training
Mainframe Systems
Our Mission and Profile
Support the IBM Systems Agenda and accelerate the adoption of
new products and solutions Power Systems

Maximize performance of our clients’ existing IBM systems
Deliver technical training, conferences, and other services tailored to System x & Bladecenter
meet client needs
Team with IBM Service Providers to optimize the deployment of IBM
solutions (GTS, GBS, SWG Lab Services and our IBM Business System Storage
Partners)

Our Competitive Advantage IT Infrastructure Optimization
Leverage relationships with the IBM development labs to build deep
technical skills and exploit the expertise of our developers
Combined expertise of Lab Services and the Training for Systems Data Center Services
team
Skills can be deployed worldwide to assure all client needs can be
Successful worldwide history: Training Services
met 17 years in Americas, 9 years in Europe/Middle East/Africa,
5 years in Asia Pacific

www.ibm.com/systems/services/labservices stgls@us.ibm.com

Special notices
This document was developed for IBM offerings in the United States as of the date of publication. IBM may not make these offerings available in
other countries, and the information is subject to change without notice. Consult your local IBM business contact for information on the IBM
offerings available in your area.
Information in this document concerning non-IBM products was obtained from the suppliers of these products or other public sources. Questions
on the capabilities of non-IBM products should be addressed to the suppliers of those products.
IBM may have patents or pending patent applications covering subject matter in this document. The furnishing of this document does not give
you any license to these patents. Send license inquires, in writing, to IBM Director of Licensing, IBM Corporation, New Castle Drive, Armonk, NY
10504-1785 USA.
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives
only.
The information contained in this document has not been submitted to any formal IBM test and is provided "AS IS" with no warranties or
guarantees either expressed or implied.
All examples cited or described in this document are presented as illustrations of the manner in which some IBM products can be used and the
results that may be achieved. Actual environmental costs and performance characteristics will vary depending on individual client configurations
and conditions.
IBM Global Financing offerings are provided through IBM Credit Corporation in the United States and other IBM subsidiaries and divisions
worldwide to qualified commercial and government clients. Rates are based on a client's credit rating, financing terms, offering type, equipment
type and options, and may vary by country. Other restrictions may apply. Rates and offerings are subject to change, extension or withdrawal
without notice.
IBM is not responsible for printing errors in this document that result in pricing or information inaccuracies.
All prices shown are IBM's United States suggested list prices and are subject to change without notice; reseller prices may vary.
IBM hardware products are manufactured from new parts, or new and serviceable used parts. Regardless, our warranty terms apply.
Any performance data contained in this document was determined in a controlled environment. Actual results may vary significantly and are
dependent on many factors including system hardware configuration and software design and configuration. Some measurements quoted in this
document may have been made on development-level systems. There is no guarantee these measurements will be the same on generally-
available systems. Some measurements quoted in this document may have been estimated through extrapolation. Users of this document
should verify the applicable data for their specific environment.

Revised September 26, 2006


Special notices (cont.)
IBM, the IBM logo, ibm.com AIX, AIX (logo), AIX 5L, AIX 6 (logo), AS/400, BladeCenter, Blue Gene, ClusterProven, DB2, ESCON, i5/OS, i5/OS (logo), IBM Business
Partner (logo), IntelliStation, LoadLeveler, Lotus, Lotus Notes, Notes, Operating System/400, OS/400, PartnerLink, PartnerWorld, PowerPC, pSeries, Rational, RISC
System/6000, RS/6000, THINK, Tivoli, Tivoli (logo), Tivoli Management Environment, WebSphere, xSeries, z/OS, zSeries, Active Memory, Balanced Warehouse,
CacheFlow, Cool Blue, IBM Systems Director VMControl, pureScale, TurboCore, Chiphopper, Cloudscape, DB2 Universal Database, DS4000, DS6000, DS8000,
EnergyScale, Enterprise Workload Manager, General Parallel File System, , GPFS, HACMP, HACMP/6000, HASM, IBM Systems Director Active Energy Manager,
iSeries, Micro-Partitioning, POWER, PowerExecutive, PowerVM, PowerVM (logo), PowerHA, Power Architecture, Power Everywhere, Power Family, POWER
Hypervisor, Power Systems, Power Systems (logo), Power Systems Software, Power Systems Software (logo), POWER2, POWER3, POWER4, POWER4+, POWER5,
POWER5+, POWER6, POWER6+, POWER7, System i, System p, System p5, System Storage, System z, TME 10, Workload Partitions Manager and X-Architecture are
trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. If these and other IBM trademarked
terms are marked on their first occurrence in this information with a trademark symbol (® or ™), these symbols indicate U.S. registered or common law trademarks
owned by IBM at the time this information was published. Such trademarks may also be registered or common law trademarks in other countries.

A full list of U.S. trademarks owned by IBM may be found at: http://www.ibm.com/legal/copytrade.shtml.

Adobe, the Adobe logo, PostScript, and the PostScript logo are either registered trademarks or trademarks of Adobe Systems Incorporated in the United States, and/or
other countries.
AltiVec is a trademark of Freescale Semiconductor, Inc.
AMD Opteron is a trademark of Advanced Micro Devices, Inc.
InfiniBand, InfiniBand Trade Association and the InfiniBand design marks are trademarks and/or service marks of the InfiniBand Trade Association.
Intel, Intel logo, Intel Inside, Intel Inside logo, Intel Centrino, Intel Centrino logo, Celeron, Intel Xeon, Intel SpeedStep, Itanium, and Pentium are trademarks or registered
trademarks of Intel Corporation or its subsidiaries in the United States and other countries.
IT Infrastructure Library is a registered trademark of the Central Computer and Telecommunications Agency which is now part of the Office of Government Commerce.
Java and all Java-based trademarks and logos are trademarks or registered trademarks of Oracle and/or its affiliates.
Linear Tape-Open, LTO, the LTO Logo, Ultrium, and the Ultrium logo are trademarks of HP, IBM Corp. and Quantum in the U.S. and other countries.
Linux is a registered trademark of Linus Torvalds in the United States, other countries or both.
Microsoft, Windows and the Windows logo are registered trademarks of Microsoft Corporation in the United States, other countries or both.
NetBench is a registered trademark of Ziff Davis Media in the United States, other countries or both.
SPECint, SPECfp, SPECjbb, SPECweb, SPECjAppServer, SPEC OMP, SPECviewperf, SPECapc, SPEChpc, SPECjvm, SPECmail, SPECimap and SPECsfs are
trademarks of the Standard Performance Evaluation Corp (SPEC).
The Power Architecture and Power.org wordmarks and the Power and Power.org logos and related marks are trademarks and service marks licensed by Power.org.
TPC-C and TPC-H are trademarks of the Transaction Performance Processing Council (TPPC).
UNIX is a registered trademark of The Open Group in the United States, other countries or both.

Other company, product and service names may be trademarks or service marks of others.

Revised December 2, 2010


Notes on benchmarks and values
The IBM benchmarks results shown herein were derived using particular, well configured, development-level and generally-available computer systems. Buyers should
consult other sources of information to evaluate the performance of systems they are considering buying and should consider conducting application oriented testing. For
additional information about the benchmarks, values and systems tested, contact your local IBM office or IBM authorized reseller or access the Web site of the benchmark
consortium or benchmark vendor.

IBM benchmark results can be found in the IBM Power Systems Performance Report at http://www.ibm.com/systems/p/hardware/system_perf.html .

All performance measurements were made with AIX or AIX 5L operating systems unless otherwise indicated to have used Linux. For new and upgraded systems, the latest
versions of AIX were used. All other systems used previous versions of AIX. The SPEC CPU2006, LINPACK, and Technical Computing benchmarks were compiled using
IBM's high performance C, C++, and FORTRAN compilers for AIX 5L and Linux. For new and upgraded systems, the latest versions of these compilers were used: XL C
for AIX v11.1, XL C/C++ for AIX v11.1, XL FORTRAN for AIX v13.1, XL C/C++ for Linux v11.1, and XL FORTRAN for Linux v13.1.

For a definition/explanation of each benchmark and the full list of detailed results, visit the Web site of the benchmark consortium or benchmark vendor.

TPC http://www.tpc.org
SPEC http://www.spec.org
LINPACK http://www.netlib.org/benchmark/performance.pdf
Pro/E http://www.proe.com
GPC http://www.spec.org/gpc
VolanoMark http://www.volano.com
STREAM http://www.cs.virginia.edu/stream/
SAP http://www.sap.com/benchmark/
Oracle, Siebel, PeopleSoft http://www.oracle.com/apps_benchmark/
Baan http://www.ssaglobal.com
Fluent http://www.fluent.com/software/fluent/index.htm
TOP500 Supercomputers http://www.top500.org/
Ideas International http://www.ideasinternational.com/benchmark/bench.html
Storage Performance Council http://www.storageperformance.org/results

Revised December 2, 2010


Notes on HPC benchmarks and values
The IBM benchmarks results shown herein were derived using particular, well configured, development-level and generally-available computer systems. Buyers should
consult other sources of information to evaluate the performance of systems they are considering buying and should consider conducting application oriented testing. For
additional information about the benchmarks, values and systems tested, contact your local IBM office or IBM authorized reseller or access the Web site of the benchmark
consortium or benchmark vendor.

IBM benchmark results can be found in the IBM Power Systems Performance Report at http://www.ibm.com/systems/p/hardware/system_perf.html .

All performance measurements were made with AIX or AIX 5L operating systems unless otherwise indicated to have used Linux. For new and upgraded systems, the latest
versions of AIX were used. All other systems used previous versions of AIX. The SPEC CPU2006, LINPACK, and Technical Computing benchmarks were compiled using
IBM's high performance C, C++, and FORTRAN compilers for AIX 5L and Linux. For new and upgraded systems, the latest versions of these compilers were used: XL C
for AIX v11.1, XL C/C++ for AIX v11.1, XL FORTRAN for AIX v13.1, XL C/C++ for Linux v11.1, and XL FORTRAN for Linux v13.1. Linpack HPC (Highly Parallel
Computing) used the current versions of the IBM Engineering and Scientific Subroutine Library (ESSL). For Power7 systems, IBM Engineering and Scientific Subroutine
Library (ESSL) for AIX Version 5.1 and IBM Engineering and Scientific Subroutine Library (ESSL) for Linux Version 5.1 were used.

For a definition/explanation of each benchmark and the full list of detailed results, visit the Web site of the benchmark consortium or benchmark vendor.
SPEC http://www.spec.org
LINPACK http://www.netlib.org/benchmark/performance.pdf
Pro/E http://www.proe.com
GPC http://www.spec.org/gpc
STREAM http://www.cs.virginia.edu/stream/
Fluent http://www.fluent.com/software/fluent/index.htm
TOP500 Supercomputers http://www.top500.org/
AMBER http://amber.scripps.edu/
FLUENT http://www.fluent.com/software/fluent/fl5bench/index.htm
GAMESS http://www.msg.chem.iastate.edu/gamess
GAUSSIAN http://www.gaussian.com
ANSYS http://www.ansys.com/services/hardware-support-db.htm
Click on the "Benchmarks" icon on the left hand side frame to expand. Click on "Benchmark Results in a Table" icon for benchmark
results.
ABAQUS http://www.simulia.com/support/v68/v68_performance.php
ECLIPSE http://www.sis.slb.com/content/software/simulation/index.asp?seg=geoquest&
MM5 http://www.mmm.ucar.edu/mm5/
MSC.NASTRAN http://www.mscsoftware.com/support/prod%5Fsupport/nastran/performance/v04_sngl.cfm
STAR-CD www.cd-adapco.com/products/STAR-CD/performance/320/index/html
NAMD http://www.ks.uiuc.edu/Research/namd Revised December 2, 2010
HMMER http://hmmer.janelia.org/
http://powerdev.osuosl.org/project/hmmerAltivecGen2mod


Practical experiences and best practices for SSD and IBM i

Practical experiences and best practices for SSD and IBM i

Recomendados

Recomendados

Más contenido relacionado

La actualidad más candente

La actualidad más candente (20)

Similar a Practical experiences and best practices for SSD and IBM i

Similar a Practical experiences and best practices for SSD and IBM i (20)

Más de COMMON Europe

Más de COMMON Europe (20)

Último

Último (20)

Practical experiences and best practices for SSD and IBM i