RFP 198156 Clarificaton Questions and Answers, Round 3

stuckwarmersMobile - sans fil

14 déc. 2013 (il y a 3 années et 3 mois)

211 vue(s)

1


RFP 198156

HIGH PERFORMANCE COMPUTER SYSTEM NO. 4 (HPCS
-
4)

Questions & Answers for Clarification


1.1

Q

Will the 400 ton chiller units be GFE on the HPCS
-
4? Please advise.

1.2

A

The 400 tons of chiller capacity will be provided by the facility, not by the offeror.

2.1

Q

With regards to the RFP, I wanted to confirm that I understood some information as written on the
Technical Requirement document, Attachment 1.


On section 11


A
䝇剅䝁R䔠卙協SM P䕒䙏前䅎䍅 䅎M⁃䅐䅃I呙Ⱐ瑨敲W⁡牥 3⁲ qu楲im敮瑳e瑨慴⁉W
hop攠you⁣慮⁨敬 m攠捬慲c晹f慳⁨agh汩gh瑥T⸠䤠瑨Wn欠瑨敳e my⁳ mp汹lbe⁴ypog牡rh楣慬⁥牲r牳⸠䍡r you
捯n晩fm⁩映 h敹⁳ ou汤⁢攠ch慮g敤e

剥晥牥r捥c

M
-
11⸳.周攠sy獴敭us琠h慶攠a⁳桡牥 Ⱐ桩Hh
-
p敲eorm慮捥cp慲慬汥氠晩l攠sy獴敭⁴h慴a楳⁡i捥c獩s汥⁦牯m 慬a
noT敳e楮⁴h攠sy獴sm⁡湤⁨慳⁡琠汥慳琠2
P楢楢y瑥s

捡c慣a瑹⸠周.⁷物瑥⁰敲 orm慮捥 o映fh楳⁦楬敳e獴sm⁳桡 l
獵獴慩n 敳猠瑨慮‱00
䝩b楂祴e
sI獥挮c


M
-

⸴.周攠獨sr敤Ⱐ桩gh⁰ 牦rrm慮捥cp慲慬l敬⁦楬敳y獴em⁳桡 氠獵l瑡楮‱00
䝩G楂iWe
猯s散efo爠r敡摩ng
晩f敳⸠


M
-
11.7 The system must be able to transfer files to and from EMSL’s HPSS archive at an aggregate
b慮T睩摴w映200
M楢楢y瑥
猯獥挠⡅mu污l敤⁌楮u砠晩f敳y獴
em moun琬W千P⁡湤⁓䙔P⁰牯瑯捯汳⁡牥
慶慩a慢汥⁩渠慤T楴ion⁴o 瑨攠n慴楶攠HP卓⁩湴 牦慣攩r

2⸲

A

周敲攠慲攠Wypog牡rh楣慬 敲eo牳r


P楢椠獨ou汤⁢攠睲w瑴敮 慳⁰敢a⁡湤 m楢椠慳im敢e.



Mu攠Wo

T楦i敲敮捥猠扥s睥en⁤楳欠捡灡捩Wy⁡湤⁰敲 orm慮捥⁳灥捩晩捡瑩Wn猬⁷e

楮瑥nT敤⁴o⁳灥捩晹cT楳欠
捡c慣楴楥a⁡湤⁰敲景牭慮c敳e楮⁳ 捴con⁍
-
11⁵獩 g⁢楮慲礠mu汴楰汥猠慳⁤敦楮敤⁢y⁉䕅 ‱541
-
2002.


Fo爠
數amp汥⁴h攠捡c慣a瑹⁩渠M
-
11⸳⁳桯u汤⁢攠2⁰敢 byW敳eo爠2砲⨪50⁢y瑥猠牡瑨敲⁴h慮′⁰ 瑡批W敳eo爠
2砱0*⨱5.


卩m楬慲ayH⁴h攠pr敦e
砠m敢e⁩湤楣慴敳e2⨪20⁡nT⁇楢椠i敡湳′⨪30.



周敳T 捯u汤⁢ or攠捬敡牬y⁷物瑴敮⁡猺



M
-
11⸳.周攠sy獴敭us琠h慶攠a⁳桡牥 Ⱐ桩Hh
-
p敲eorm慮捥cp慲慬汥氠晩l攠sy獴敭⁴h慴a楳⁡i捥c獩s汥⁦牯m 慬a
noT敳e楮⁴h攠sy獴sm⁡湤⁨慳⁡琠汥慳琠2

P敢ebyW敳

⠲x2
50

bytes
)capacity. The write performance of this
filesystem shall sustain no less than 100

GibiByte
s/sec (100x2
30

bytes/sec).



M
-
11.4 The shared, high performance parallel filesystem shall sustain 100

GibiByte
s/sec
(100x2
30

bytes/sec) for reading files.



M
-
11.7
The system must be able to transfer files to and from EMSL’s HPSS archive at an aggregate
b慮T睩摴w映200
M敢ebyWe
猯獥挠(200砲
20

bytes/sec )(Emulated Linux filesystem mount, SCP and SFTP
protocols are available in addition to the native HPSS interface).


A

revision to Attachment 1 to reflect this edit will be forthcoming.

3.1

Q

The RFP states that the proposed HPCS
-
4 system must have a value not
-
to
-
exceed $12 million. That
amount is the Total Aggregate System Price (TASP) that includes the purchase or leas
e price and all
maintenance options price, including the optional maintenance for years 2




2



Attachment 1 states that the HPCS
-
4 procurement will consist of two systems two years apart: HPCS
-
4A, to be installed in early 2013, and HPCS
-
4B, to be installe
d in early 2015.


Please clarify whether it is your intention to spend a total of $12 million spread across two years.

3.2

A

T
he
TASP of
$12M includes the initial purchase plus year 1 maintenance and years 2
-
4 optional
maintenance.


HPCS
-
4B is not part of

this solicitation.

4.1

Q

Is there a list of registered bidders, my product line is for a portion of the infrastructure and we may
want to supply material bids to General Contractors.

4.2

A

We do not publish
the
bidders

list.

5.1

Q

We
've been investigating the source code to ga_shift (file ga_shift.F).


If
we

make the following changes



The two lines beginning with “!djp” have been commented out of the source code.

The single line ending with “!djp” has been added.


坯u汤 you⁰汥慳攠
捯n晩fm 睨整h敲⁴h攠sou牣攠coT攠ch慮g敳e慮Tu瑰W琠慲攠a捣数瑡e汥l


c

挪c*†䥮瑩瑩慬Wze⁡ m敳獡e攠p慳獩ag楢牡ry⁣ ⍩晤敦eMPI

††
楮瑥g敲⁩敲

††
捡汬⁧慟楮i瑩慬Wz攨⤠††‡ Tjp

††
瑢敧W㴠g慟睴im攨)

Ⅴ橰†† ⁣慬a mp楟楮楴⡩敲爩

⍥汳l

††
瑢敧

㴠g慟睴im攨)

††
捡汬⁰ 敧enf

⍥nT楦

††
瑟m獧彩_楴‽⁧ 彷瑩me(⤠
-

瑢敧

c

挪c*†䥮楴楡i楺攠䝁

c

挠† ⁔h敲攠慲a′ 捨c楣敳e⁧ 彩_楴楡i楺攠o爠条彩_楴楡i楺e彬_T.

挠† ⁉ ⁴h攠晩f獴s捡獥Ⱐ瑨敲W⁩猠湯 數p汩捩琠汩m楴n m敭o特ru獡s攮

挠† ⁉ ⁴h攠獥sonTⰠ畳
敲⁣慮 獥s m楴i⡰敲⁰(o捥c獯爩⁩渠by瑥献

c

††
瑢敧W㴠g慟睴im攨)

Ⅴ橰†† ⁣慬a⁧慟楮楴楡i楺攨)

周攠b敮捨m慲欠牵as⁧ v楮g 瑨攠景汬o睩湧u瑰W琠⡡(慩aⰠ灬敡獥H瑲敡琠瑨W猠摡瑡W愠䅰p牯⁰牯p物整慲a⤺


n㴱

呩Te⁳灥湴⁩ ⁲畮瑩W攠楮楴i慬az慴楯n: ††† ‰.000
000

呩Te⁳灥湴⁩ ⁇䄠楮楴楡i楺慴aon: ††††† †0.000000


䝁⁩湩瑩慬Wz敤

呩Te⁳灥湴⁩ ⁍䄠楮楴i慬az慴楯n:†† ††††‰.000017

啳楮g†† † 1⁰ o捥cs⡥猩


呥獴楮s⁢ o捫⁳楺攠o映f† † 4

呩Te⁳灥湴 捲敡瑩Wg⁇䄺 †† †† ††† ††‰.000028

3


Time spent
in put: 0.000018

Bandwidth for put (MB/s): 1.766023

Time spent in sync: 0.000000

Bandwidth for put with sync (MB/s): 1.766023

Time spent in get: 0
.000010

Bandwidth for get (MB/s): 3.195660

Time spent destroying GA: 0.000013


Testing block size of 16

Time spent creating GA: 0.000002

Time spent in put: 0.000
001

Bandwidth for put (MB/s): 107.374182

Time spent in sync: 0.000000

Bandwidth for put with sync (MB/s): 107.374182

Time spent in get: 0.000001

Bandwidth for get (MB/s):

134.217728

Time spent destroying GA: 0.000001


Testing block size of 64

Time spent creating GA: 0.000001

Time spent in put: 0.000001

Bandwidth for put (MB/s):

536.870912

Time spent in sync: 0.000000

Bandwidth for put with sync (MB/s): 536.870912

Time spent in get: 0.000001

Bandwidth for get (MB/s): 536.870912

Time spent destro
ying GA: 0.000000


Testing block size of 256

Time spent creating GA: 0.000001

Time spent in put: 0.000001

Bandwidth for put (MB/s): 2147.483648

Time spent in sync:

0.000000

Bandwidth for put with sync (MB/s): 2147.483648

Time spent in get: 0.000001

Bandwidth for get (MB/s): 2147.483648

Time spent destroying GA: 0.000000


T
esting block size of 1024

Time spent creating GA: 0.000004

Time spent in put: 0.000002

Bandwidth for put (MB/s): 4294.967296

Time spent in sync: 0.000001

Bandwi
dth for put with sync (MB/s): 2863.311531

Time spent in get: 0.000002

Bandwidth for get (MB/s): 3817.748708

Time spent destroying GA: 0.000000


Testing block size of 4096

Time spent

creating GA: 0.000004

Time spent in put: 0.000003

Bandwidth for put (MB/s): 10572.227190

Time spent in sync: 0.000001

4


Bandwidth for put with sync (MB/s): 808
4.644322

Time spent in get: 0.000004

Bandwidth for get (MB/s): 8589.934592

Time spent destroying GA: 0.000000


Testing block size of 16384

Time spent creating GA: 0.00
0004

Time spent in put: 0.000011

Bandwidth for put (MB/s): 11696.932210

Time spent in sync: 0.000000

Bandwidth for put with sync (MB/s): 11696.932210

Time spent in get:

0.000012

Bandwidth for get (MB/s): 10779.525763

Time spent destroying GA: 0.000000


Testing block size of 65536

Time spent creating GA: 0.000015

Time spent in put:

0.000042

Bandwidth for put (MB/s): 12494.450316

Time spent in sync: 0.000000

Bandwidth for put with sync (MB/s): 12494.450316

Time spent in get: 0.000055

Bandwidth for ge
t (MB/s): 9519.581193

Time spent destroying GA: 0.000001


Testing block size of 262144

Time spent creating GA: 0.000005

Time spent in put: 0.000158

Bandwidth for put (M
B/s): 13267.108631

Time spent in sync: 0.000001

Bandwidth for put with sync (MB/s): 13187.545760

Time spent in get: 0.000175

Bandwidth for get (MB/s): 11983.777959

T
ime spent destroying GA: 0.000000


Testing block size of 1048576

Time spent creating GA: 0.000005

Time spent in put: 0.000729

Bandwidth for put (MB/s): 11505.680866

Time
spent in sync: 0.000000

Bandwidth for put with sync (MB/s): 11505.680866

Time spent in get: 0.000714

Bandwidth for get (MB/s): 11751.627284

Time spent destroying GA:

0.000001

n=2

Time spent in runtime initialization: 0.000000

Time spent in GA initialization: 0.000000


GA initialized

Time spent in MA initialization: 0.000019

Using 2 process(es)


Testing block size of
4

Time spent creating GA: 0.000049

5


Time spent in put: 0.000029

Bandwidth for put (MB/s): 1.104673

Time spent in sync: 0.000096

Bandwidth for put with sync (MB/s)
: 0.255897

Time spent in get: 0.000092

Bandwidth for get (MB/s): 0.347714

Time spent destroying GA: 0.000005


Testing block size of 16

Time spent creating GA:

0.000029

Time spent in put: 0.000003

Bandwidth for put (MB/s): 42.949673

Time spent in sync: 0.000088

Bandwidth for put with sync (MB/s): 1.399924

Time spent in get:

0.000054

Bandwidth for get (MB/s): 2.354697

Time spent destroying GA: 0.000003


Testing block size of 64

Time spent creating GA: 0.000022

Time spent in put:

0.000003

Bandwidth for put (MB/s): 171.798692

Time spent in sync: 0.000082

Bandwidth for put with sync (MB/s): 6.057782

Time spent in get: 0.000102

Band
width for get (MB/s): 5.041041

Time spent destroying GA: 0.000003


Testing block size of 256

Time spent creating GA: 0.000042

Time spent in put: 0.000003

Bandwidt
h for put (MB/s): 592.409282

Time spent in sync: 0.000115

Bandwidth for put with sync (MB/s): 17.300976

Time spent in get: 0.000066

Bandwidth for get (MB/s):

31.010594

Time spent destroying GA: 0.000016


Testing block size of 1024

Time spent creating GA: 0.000022

Time spent in put: 0.000010

Bandwidth for put (MB/s): 827.
945503

Time spent in sync: 0.000100

Bandwidth for put with sync (MB/s): 74.533055

Time spent in get: 0.000027

Bandwidth for get (MB/s): 298.780334

Time spent destroying GA:

0.000017


Testing block size of 4096

Time spent creating GA: 0.000031

Time spent in put: 0.000008

Bandwidth for put (MB/s): 3871.519816

Time spent in sync:

0.000051

6


Bandwidth for put with sync (MB/s): 550.857529

Time spent in get: 0.000041

Bandwidth for get (MB/s): 808.464432

Time spent destroying GA: 0.000003


Testing block

size of 16384

Time spent creating GA: 0.000027

Time spent in put: 0.000029

Bandwidth for put (MB/s): 4451.464080

Time spent in sync: 0.000068

Bandwidth for put
with sync (MB/s): 1345.791466

Time spent in get: 0.000069

Bandwidth for get (MB/s): 1885.954765

Time spent destroying GA: 0.000003


Testing block size of 65536

Time spent creating GA
: 0.000035

Time spent in put: 0.000085

Bandwidth for put (MB/s): 6203.168563

Time spent in sync: 0.000134

Bandwidth for put with sync (MB/s): 2399.370710

Tim
e spent in get: 0.000208

Bandwidth for get (MB/s): 2520.370493

Time spent destroying GA: 0.000004


Testing block size of 262144

Time spent creating GA: 0.000036

Time
spent in put: 0.000477

Bandwidth for put (MB/s): 4392.555816

Time spent in sync: 0.000202

Bandwidth for put with sync (MB/s): 3088.515808

Time spent in get:

0.000573

Bandwidth for get (MB/s): 3660.463180

Time spent destroying GA: 0.000003


Testing block size of 1048576

Time spent creating GA: 0.000035

Time spent in put:

0.001844

Bandwidth for put (MB/s): 4549.017013

Time spent in sync: 0.000200

Bandwidth for put with sync (MB/s): 4103.133771

Time spent in get: 0.002359

Bandwidth for get (MB/s):

3556.671427

Time spent destroying GA: 0.000004

n=4

Time spent in runtime initialization: 0.000001

Time spent in GA initialization: 0.000000


GA initialized

Time spent in MA initialization:

0.000020

Using 4 process(es)


Testing block size of 4

Time spent creating GA: 0.000071

7


Time spent in put: 0.000024

Bandwidth for put (MB/s): 1.345541

Time spent in sync:

0.000102

Bandwidth for put with sync (MB/s): 0.255046

Time spent in get: 0.000090

Bandwidth for get (MB/s): 0.354604

Time spent destroying GA: 0.000013


Testing block size of 16

Time spent creating GA: 0.000074

Time spent in put: 0.000006

Bandwidth for put (MB/s): 22.369621

Time spent in sync: 0.000078

Band
width for put with sync (MB/s): 1.528458

Time spent in get: 0.000064

Bandwidth for get (MB/s): 1.999519

Time spent destroying GA: 0.000006


Testing block size of 64

Time spe
nt creating GA: 0.000041

Time spent in put: 0.000003

Bandwidth for put (MB/s): 171.798692

Time spent in sync: 0.000088

Bandwidth for put with sync (MB/s):

5.647557

Time spent in get: 0.000061

Bandwidth for get (MB/s): 8.396808

Time spent destroying GA: 0.000018


Testing block size of 256

Time spent creating GA: 0.
000057

Time spent in put: 0.000003

Bandwidth for put (MB/s): 636.291451

Time spent in sync: 0.000104

Bandwidth for put with sync (MB/s): 19.014797

Time spent in get:

0.000035

Bandwidth for get (MB/s): 58.040099

Time spent destroying GA: 0.000011


Testing block size of 1024

Time spent creating GA: 0.000051

Time spent in put:

0.000005

Bandwidth for put (MB/s): 1655.891006

Time spent in sync: 0.000060

Bandwidth for put with sync (MB/s): 126.438780

Time spent in get: 0.000014

Bandwidth for
get (MB/s): 572.662306

Time spent destroying GA: 0.000005


Testing block size of 4096

Time spent creating GA: 0.000042

Time spent in put: 0.000008

Bandwidth for put
(MB/s): 4102.655328

Time spent in sync: 0.000050

8


Bandwidth for put with sync (MB/s): 567.343461

Time spent in get: 0.000020

Bandwidth for get (MB/s): 1616.928864

Time spent destroying GA: 0.000005


Testing block size of 16384

Time spent creating GA: 0.000051

Time spent in put: 0.000029

Bandwidth for put (MB/s): 4524.739209

Tim
e spent in sync: 0.000064

Bandwidth for put with sync (MB/s): 1405.126681

Time spent in get: 0.000067

Bandwidth for get (MB/s): 1954.687338

Time spent destroying GA:

0.000013


Testing block size of 65536

Time spent creating GA: 0.000044

Time spent in put: 0.000081

Bandwidth for put (MB/s): 6496.375940

Time spent in sync:

0.000142

Bandwidth for put with sync (MB/s): 2351.267849

Time spent in get: 0.000201

Bandwidth for get (MB/s): 2611.666574

Time spent destroying GA: 0.000006


Testing block size of

262144

Time spent creating GA: 0.000053

Time spent in put: 0.000484

Bandwidth for put (MB/s): 4330.917293

Time spent in sync: 0.000180

Bandwidth for put with sync
(MB/s): 3154.983150

Time spent in get: 0.000599

Bandwidth for get (MB/s): 3503.721578

Time spent destroying GA: 0.000008


Testing block size of 1048576

Time spent creating GA:

0.000058

Time spent in put: 0.001817

Bandwidth for put (MB/s): 4616.160075

Time spent in sync: 0.000217

Bandwidth for put with sync (MB/s): 4123.692120

Time spent in

get: 0.002282

Bandwidth for get (MB/s): 3675.567729

Time spent destroying GA: 0.000008

n=8

Time spent in runtime initialization: 0.000001

Time spent in GA initialization:

0.000000


GA initialized

Time spent in MA initialization: 0.000019

Using 8 process(es)


Testing block size of 4

Time spent creating GA: 0.000105

9


Time spent in put: 0.000024

Bandwi
dth for put (MB/s): 1.360890

Time spent in sync: 0.000103

Bandwidth for put with sync (MB/s): 0.252289

Time spent in get: 0.000097

Bandwidth for get (MB/s):

0.328462

Time spent destroying GA: 0.000016


Testing block size of 16

Time spent creating GA: 0.000087

Time spent in put: 0.000003

Bandwidth for put (MB/s): 4
1.698712

Time spent in sync: 0.000080

Bandwidth for put with sync (MB/s): 1.540519

Time spent in get: 0.000086

Bandwidth for get (MB/s): 1.484607

Time spent destroying GA:

0.000020


Testing block size of 64

Time spent creating GA: 0.000083

Time spent in put: 0.000003

Bandwidth for put (MB/s): 168.430090

Time spent in sync:

0.000086

Bandwidth for put with sync (MB/s): 5.768929

Time spent in get: 0.000090

Bandwidth for get (MB/s): 5.696243

Time spent destroying GA: 0.000014


Testing blo
ck size of 256

Time spent creating GA: 0.000082

Time spent in put: 0.000005

Bandwidth for put (MB/s): 411.493873

Time spent in sync: 0.000107

Bandwidth for pu
t with sync (MB/s): 18.369280

Time spent in get: 0.000035

Bandwidth for get (MB/s): 57.747459

Time spent destroying GA: 0.000008


Testing block size of 1024

Time spent creating
GA: 0.000074

Time spent in put: 0.000005

Bandwidth for put (MB/s): 1686.367527

Time spent in sync: 0.000084

Bandwidth for put with sync (MB/s): 91.932410

T
ime spent in get: 0.000032

Bandwidth for get (MB/s): 259.318780

Time spent destroying GA: 0.000010


Testing block size of 4096

Time spent creating GA: 0.000077

Time
spent in put: 0.000008

Bandwidth for put (MB/s): 3969.356057

Time spent in sync: 0.000048

10


Bandwidth for put with sync (MB/s): 586.093618

Time spent in get:

0.000021

Bandwidth for get (MB/s): 1570.730897

Time spent destroying GA: 0.000008


Testing block size of 16384

Time spent creating GA: 0.000082

Time spent in put:

0.000030

Bandwidth for put (MB/s): 4437.988407

Time spent in sync: 0.000076

Bandwidth for put with sync (MB/s): 1240.633713

Time spent in get: 0.000064

Bandwidth for get (MB/s):

2062.873598

Time spent destroying GA: 0.000008


Testing block size of 65536

Time spent creating GA: 0.000074

Time spent in put: 0.000083

Bandwidth for put (MB/s):

6285.168290

Time spent in sync: 0.000120

Bandwidth for put with sync (MB/s): 2579.121250

Time spent in get: 0.000205

Bandwidth for get (MB/s): 2560.353085

Time spent
destroying GA: 0.000008


Testing block size of 262144

Time spent creating GA: 0.000078

Time spent in put: 0.000475

Bandwidth for put (MB/s): 4417.372516

Time spent in s
ync: 0.000145

Bandwidth for put with sync (MB/s): 3381.162030

Time spent in get: 0.000574

Bandwidth for get (MB/s): 3654.380150

Time spent destroying GA:
0.000008


Testing block size of 1048576

Time spent creating GA: 0.000081

Time spent in put: 0.001804

Bandwidth for put (MB/s): 4651.248872

Time spent in sync: 0.00
0162

Bandwidth for put with sync (MB/s): 4268.847183

Time spent in get: 0.002275

Bandwidth for get (MB/s): 3686.881613

Time spent destroying GA: 0.000012

n=16

Time spent in runtime init
ialization: 0.000001

Time spent in GA initialization: 0.000000


GA initialized

Time spent in MA initialization: 0.000021

Using 16 process(es)


Testing block size of 4

Time spent creating GA:

0.000167

11


Time spent in put: 0.000024

Bandwidth for put (MB/s): 1.356591

Time spent in sync: 0.000104

Bandwidth for put with sync (MB/s): 0.250816

Time spent in get:

0.000089

Bandwidth for get (MB/s): 0.358631

Time spent destroying GA: 0.000019


Testing block size of 16

Time spent creating GA: 0.000131

Time spent in put:

0.000003

Bandwidth for put (MB/s): 41.497269

Time spent in sync: 0.000087

Bandwidth for put with sync (MB/s): 1.417949

Time spent in get: 0.000086

Bandwidt
h for get (MB/s): 1.492604

Time spent destroying GA: 0.000017


Testing block size of 64

Time spent creating GA: 0.000154

Time spent in put: 0.000003

Bandwidth fo
r put (MB/s): 156.180629

Time spent in sync: 0.000087

Bandwidth for put with sync (MB/s): 5.672732

Time spent in get: 0.000081

Bandwidth for get (MB/s): 6.
320776

Time spent destroying GA: 0.000026


Testing block size of 256

Time spent creating GA: 0.000127

Time spent in put: 0.000004

Bandwidth for put (MB/s): 572.6623
06

Time spent in sync: 0.000097

Bandwidth for put with sync (MB/s): 20.346255

Time spent in get: 0.000026

Bandwidth for get (MB/s): 79.906368

Time spent destroying GA:

0.000013


Testing block size of 1024

Time spent creating GA: 0.000115

Time spent in put: 0.000005

Bandwidth for put (MB/s): 1660.893698

Time spent in sync:

0.000035

Bandwidth for put with sync (MB/s): 206.364795

Time spent in get: 0.000014

Bandwidth for get (MB/s): 580.523563

Time spent destroying GA: 0.000012


Testing block siz
e of 4096

Time spent creating GA: 0.000112

Time spent in put: 0.000008

Bandwidth for put (MB/s): 3864.715739

Time spent in sync: 0.000048

12


Bandwidth for put with

sync (MB/s): 578.081823

Time spent in get: 0.000024

Bandwidth for get (MB/s): 1347.440720

Time spent destroying GA: 0.000011


Testing block size of 16384

Time spent creating GA:

0.000117

Time spent in put: 0.000030

Bandwidth for put (MB/s): 4404.653491

Time spent in sync: 0.000059

Bandwidth for put with sync (MB/s): 1483.070818

Time sp
ent in get: 0.000058

Bandwidth for get (MB/s): 2275.243927

Time spent destroying GA: 0.000010


Testing block size of 65536

Time spent creating GA: 0.000118

Time spent
in put: 0.000081

Bandwidth for put (MB/s): 6446.385505

Time spent in sync: 0.000074

Bandwidth for put with sync (MB/s): 3385.391330

Time spent in get:
0.000202

Bandwidth for get (MB/s): 2600.855418

Time spent destroying GA: 0.000011


Testing block size of 262144

Time spent creating GA: 0.000118

Time spent in put: 0.00
0477

Bandwidth for put (MB/s): 4392.144567

Time spent in sync: 0.000076

Bandwidth for put with sync (MB/s): 3790.500373

Time spent in get: 0.000581

Bandwidth for get (MB/s):

3607.266138

Time spent destroying GA: 0.000025


Testing block size of 1048576

Time spent creating GA: 0.000128

Time spent in put: 0.001867

Bandwidth for put (MB/s):

4492.028163

Time spent in sync: 0.000131

Bandwidth for put with sync (MB/s): 4196.546699

Time spent in get: 0.002298

Bandwidth for get (MB/s): 3650.991001

Time spent destr
oying GA: 0.000027


5.2

A

The suggested modification shouldn't cause the benchmark to produce bad results, but should not be
needed if compilation environment and options are correct. The results appear to be in the expected
format.

6.1

Q

I have a question on Infiniband interconnect. Do we have to provide 1:1 100% bi
-
sectional bandwidth
or can we provide 2:1 oversubscription?

6.2

A

Benchmark performance will be a major determinant for selection. If 1:1 100% bisectional bandwidth
doe
s not help performance on our benchmarks, then it is not needed.

13


7.1

Q

* Can you define the expected [Minimum and Maximum] Service Level agreement for system
maintenance for this procurement?


Answer:
We don't have an expected Service Level for system
maintenance.


It really depends on the
reliability of the system and the expected maintenance requirements needed to provide the 97.5%
system availability.


We expect that the system supplier will be able to provide the necessary
replacement equipment or r
edundancy equipment necessary to maintain the required availability.


* For requirement M
-
3.1 could we ask for clarification of the meaning of “system availability”?


䅮獷s爺†
卥S瑩Wn⁆映瑨攠Wo汩捩瑡瑩on⁨慳⁡ 獥捴con⁴楴i敤•䅣e数瑡e捥⁔敳瑩Wg∠numb敲e4⁩渠瑨 琠獥捴楯n
T敳e物r敳e瑨攠Wv慩a慢楬楴i⁴e獴楮s⁤ 物rg⁴h攠30
-
T慹a慶a楬慢楬楴i⁴e獴sby⁣oun瑩Wg T攠To睮瑩W攮


䥴⁡汳
獴慴敳⁴h慴a≴h攠佦晥Oo爠may⁰牯po獥⁡渠慬 敲湡瑥 me
WhoT⁦ r⁴牡r歩kg⁳ s瑥m⁡v慩a慢楬楴i⸢


⨠䙯爠牥煵楲im敮瑳⁍
-
5⸷. M
-
5⸸⁡湤⁍
-
5.9, these capabilities are listed for “All nodes”. It could be
慲au敤⁴桡琠瑨敳e⁲敱e楲im敮瑳⁡牥 noW⁡灰汩捡c汥⁦o爠捯mpu瑥 T敳e⁆ 爠mo獴⁡牣桩瑥c瑵牥猠
慵瑨敮瑩捡瑩on⁩猠Ton
攠on汹 g楮爠桥慤oT敳e慮T⁩ no琠Ton攠⡡湤⁩渠獯me⁣慳敳e捡cno琠be⁤ n攩n
捯mpu瑥oT敳⸠坯u汤 PNNL⁣on獩s敲e牥污硩xg⁴h敳攠牥煵楲敭敮瑳eWo on汹⁩湣 uT攠Log楮⁡湤⁓敲 i捥c
noT敳e


䅮獷s爺†
䙯r⁡湹 noT敳⁴ha琠牥煵楲攠灡i獷o牤⁡畴 敮e楣慴楯nⰠ瑨敳W⁲ q
u楲im敮W猠su獴⁢攠
m整.


卥捵S楴y⁦敡eu牥rf⁴h攠獹s瑥m mu獴sb攠慢汥l瑯業楴⁡捣c獳⁴o⁴h攠compu瑡瑩WnoT敳e楮 瑨攠
獹獴smH⁳畣 ⁴h慴⁵ 慵瑨Wr楺敤⁵獥牳e睩汬oW⁨慶攠慣a敳猠WooT敳e牵nn楮g瑨敲Wu獥爧猠橯b献


周攠u獥物sⰠ畳敲湡m攠慮T g牯up楤Ⱐ杲Hupn慭攠
m慰p楮ng⁦ 爠rhe⁍千⁩猠m慩a瑡Wn敤⁩e⁡渠數瑥牮慬a
LMAP⁤慴慢慳攮⁁汬⁳y獴sm 獥牶楣敳er敱e楲ing⁵獥物 Igroup楤慰p楮g⁦o爠畳敲r⁷i汬⁵獥 瑨W⁩湦orm慴aon
p牯v楤敤⁢y⁌MAP.


⨠*
-
7⸷.楳⁡ r敱e楲敭敮琠eo爠rPI
-
2⁳異po牴⸠Mo敳e瑨Ws⁩n捬cT攠瑨攠灲o捥c猠sr敡eion⁡
nT慮慧em敮e
po牴楯nf⁴h攠s瑡湤慲a?⁉映獯 睯u汤 PNNL⁣on獩s敲⁲ l慸楮g⁴h攠r敱e楲敭敮琠eoo琠楮捬cTe⁰牯c敳猠
捲敡瑩on⁡湤 m慮慧em敮e?


䅮獷s爺†
M
-
7⸷⁳桯u汤⁢ ⁣污l楦i敤⁦牯m:



M
-
7⸷›

MP䤠汩b牡r楥猠獨ou汤⁦畬汹l獵spo牴⁴h攠MPI
-
2⁳W慮T慲a



呯:



M
-
7
⸷›

MP䤠汩b牡r楥猠獨ou汤⁳異po牴r瑨攠MPI
-
2⁳ 慮Ta牤⸠䥦⁡⁦畬氠業pl敭敮瑡瑩on o映fhe⁍PI
-
2⁳ 慮T慲a
楳i琠prov楤敤Ⱐ瑨o獥⁰慲瑳映瑨攠MPI
-
2⁳ 慮T慲ao琠b敩eg晦敲敤emu獴sbe⁤ cum敮e敤⁩e⁴h攠
牥獰rn獥s†P牯v楤楮g敳 瑨慮⁡ 晵汬⁍PI
-
2⁩ p汥m敮瑡瑩Wn
睩汬⁤散 敡獥 瑨攠v慬aa瑩Wn o映fhe⁳ 獴sm.


⨠兵敳瑩Wn⁰敲瑡楮猠eo 瑨敳攠楮獴牵捴son猠楮s瑨攠W䕁MM䔠晩fe⁦牯m⁐NNL⁦ r⁴h攠坒䘠扥湣Pm慲a:

2⸠創.楮g

ⴭⴭⴭⴭ-

†† †

周攠nam敬楳e⁰牯v楤敤e景爠扯瑨⁴桥‹歭⁡湤″km 捡獥猠獨ou汤⁢攠敤e瑥T

†† †

Wo⁲敦汥e琠a

牵n o映5
fo爠rh攠9km⤠慮T 1
fo爠瑨攠3歭)⁳業u污瑥T

14




minutes, by changing the run_hours and run_minutes inputs. Also note



that the namelist and restart files may have to be renamed.




Use a power of two processors within a node, and

a power of two nodes,



for a scalability run. For each processor count, use a


minimum number



of sockets and nodes within the testsystem (i.e. packed within



sockets and nodes). The larger (3km) can start at a higher number of




cores (e.g. 1 to 2 nodes) due to large iteration times.


Question for PNNL is:


In the WRF README instructions PNNL says to run the 2 benchmark cases

on 1 to 2 nodes.


Given the grid size of the 3KM case we feel that this is not

enough nodes.


Can we sta
rt at a higher number of nodes for this case?

Answer:
Yes. You can start with a higher number of nodes if necessary.


Also, since the requested forecast duration for both tests is quite short (1 and 5 minutes)

can we report average integration time step t
imings to avoid I/O overheads?

Answer:
You are welcome to report the average integration time step timings in addition to the
information requested.

7.2

A

Included above in
RED

after each question.

8.1

Q

I was wondering if this solicitation will be acce
pting of quotations of configurations that are not Dell
branded?

8.2

A

We do not have a requirement

for a particular brand of HPC equipment.

9.1

Q

C
an you confirm if I am able to dial
-
in to this Q
&
A meeting via tele
-
conference?

9.2

A

The meeting’s primary purpose is to allow for a site visit. The meeting will not be available via tele
-
捯n晥f敮捥⸠e⁍敥瑩Wg m楮u瑥猠s楬l⁢攠po獴sT on 瑨攠睥b⁰慧攮

10⸱

Q

1.

Can you provide any additional Ch
inook related HW and SW details?

Beyond what is

available on the provided link:
http://www.emsl.pnl.gov/capabilities/computing/msc/supercomputing/supercomputer.js
p


Answer: Chinook has 2321 compute
nodes. The fabric is fully interconnected within each
of 12 computational units (CU). The node
-
level switches from each CU are aggregated
with four (4) top level IB switches at a 2:1 or 3:1 oversubscription ratio, depending on
which fabric blade the link
s go through.


Each compute node had 32G of memory with 2 sockets of quad
-
core AMD Opteron
“Barcelona” processors running at 2.2GHz. Each node also contains 8 SCSI disks configured
in a RAID configuration offering 400MB/sec of I/O bandwidth to the local /
scratch
filesystem for jobs to use.


The compute nodes do a network boot to an NFS mounted root filesystem, storing some
15


local logs on a partition of the local disks. We currently use xcat for provisioning
computational nodes, using a “statelite” configur
ation. The NFS filesystem is mounted
over the GigE connection and each node also has a GigE connection to an IPMI/control
network for console and power control.


A lustre /home filesystem contains user directories and shared applications and system
tools,

including compilers, MPI libraries etc.


Four login nodes are available to users. Two of them are accessible from outside of the
PNNL network via ssh. A parallel lustre /dtemp filesystem is available on the system to
provide high performance parallel I/O

at 30GB/sec.


2.

Should the M
-
9.11 reference 64A instead of 640A.

M
-
9.11 Maximum rack current shall be 80A at 480V 3 phase and each 800A buss duct shall
not exceed 640A of continuous power.



Answer: Each 800A buss duct shall not exceed 640A of continuous p
ower. However, the
maximum rack current shall be 80A at 480V 3 phase. The 640A is correct.


3.

Would it be possible 50 page limit be relaxed for the Technical Proposal?

This would allow
for a more cohesive and readable proposal, with the description of how

the requirement is
to be met within the MR and VRF tables supplied in Attachment 1, as well as a narrative
response that may include figures and/or diagrams as supporting documentation in the
page number referenced within the tables.


Answer: Our intent
in limiting the technical proposal to 50 pages was to avoid the
inclusion of long brochures and product description documents. By limiting the proposal
to 50 pages we limit the extraneous information that may not apply to the project
requirements and valu
e related factors. We welcome diagrams and figures that reduce
the need for long explanations, but reaffirm our desire that the technical proposal be
limited to 50 pages.


10.2

A

Included above in
RED

after each question.

11.1

Q

I'm trying to access the

ARSC website for the WRF benchmark datasets as stated in the HPCS
-
4
benchmark instructions and also the link from your MSC benchmark website:


http://www.emsl.pnl.gov/r
oot/capabilities/computing/msc/msc_benchmark/




Both links point to
http://weather.arsc.edu/BenchmarkSuite/


but the site appears to be unavailable:


http://weather.arsc.edu/BenchmarkSuite/




Not Found


The requested URL /BenchmarkSuite/ was not found on this server.

16


11.2

A

PNNL
tried the link for
http://weather.arsc.edu/BenchmarkSuite


and
http://weather.arsc.edu/BenchmarkSuite/


from Internet Explorer, Chrome, Firefox and Safari, from
Windows, Linux and MacOS X, through work and home conn
ections
,

and even tried
my
Android phone
through the 3G connection and they all seem to find the page.

12.1

Q

We request a three (3) week extension for the due date of the subject Proposal to May 30, 2012. In
today's supplier environment we have seen sig
nificant delays

in getting favored pricing for large scale bids. The extension will allow the lowest price possible to be
submitted to PNNL.

12.2

A

The original RFP due date of 5/9/12 remains firm. As provided in Section A , Clause “Submission,
MoT楦ic
ation, Revision, and Withdrawal of Proposals” Part 3, it states that late modifications of an
o瑨敲睩獥w獵捣敳獦s氠灲opo獡氠瑨慴sm慫攠瑨攠W敲e猠mo牥r晡vo牡r汥l瑯⁂慴 敬汥e睩汬⁢攠con獩s敲敤⁡琠慮y
瑩We⁩琠楳⁲ c敩e敤⁰e楯r⁴o 慷慲a⁡湤 m慹ab攠慣捥灴cT.