slow disk-to-disk transfers: devb-aha8

Any way to speed up disk transfers? I started a move of about 32GB of
data early this morning from one 72GB disk to another on the same chain.
It’s still going at 4:41PM… I would think that it would take less time
than that! The system is not busy at all, the disk transfers seem to be
happening quietly in the background.

The SCSI adapter is listed as an “AIC-7892A Ultra160/m PCI SCSI
Controller” in the pci outout. Disks are IBM 72GB Ultra SCSI units. CPU
is a Pentium III 1.0GHz with 256MB of RAM.

I’ve tried to increase the cache size in the kernel build file, but all
I get is “Unable to allocate resource” at boot time, and the size of the
devb-aha8 process in memory stays at the default 12MB.

Mathew Kirsch <mkirsch@ocdus.jnj.com> wrote:

Any way to speed up disk transfers? I started a move of about 32GB of
data early this morning from one 72GB disk to another on the same chain.
It’s still going at 4:41PM… I would think that it would take less time
than that! The system is not busy at all, the disk transfers seem to be
happening quietly in the background.

Yes, I’d expect that volume of data to take half an hour or so? Were
you just using “cp”? What did “cp -V” report as the throughput rate?

The SCSI adapter is listed as an “AIC-7892A Ultra160/m PCI SCSI
Controller” in the pci outout. Disks are IBM 72GB Ultra SCSI units. CPU
is a Pentium III 1.0GHz with 256MB of RAM.

Hopefully our disk guy can comment on configuring these disks (things
like sync, disconnect, etc) and what rate would be expected from them.

You could try this … I’ve attached a disk throughput benchmark. Run
it on each disk and see if they perform similarly and at expected
levels. If not, then some sort of basic configuration issue. Otherwise,
allowing for the fact that a disk-to-disk copy is going to use twice
the disk cache, twice the message-passing and other overheads, and allow
that cp will only use a 16k buffer, then if the actual copy is not at
that level, then there is maybe a disk-to-disk interaction that we need
to look into? Any help at all?

table
!"#$%&’()+,-./0123456789:;<=>?
@ABCDEFGHIJKLMNOPQRSTUVWXYZ[]^_
begin 644 rw
M?T5,1@$! 0 ( P ! -(T$"#0 #%1 #0 ( &z
M “@ &P : 8 T -( $” # P 4 $ P /0 y
M #T@ 0( !0 4 ! $ ! " ! @ @ 0(Y"T x
M .0M % ! $ #D+0 Y+T$".2]! B 0 Q , 8 $ w
M @ -PN #<O@0( (@ “( !@ 0 $ 9”\ v
M 3 ! "]U<W(O;&EB+VQD<6YX+G-O+C( )0 #\ u
M 2 +P # )P (P !X .0 t
M 0 Q "H R / S + #T ^ s
M 0 #8 ) ; &@ #@ -0 P 8r
M (0 #0 0 q
M , H " p
M T 5 "P 4 $0 !< " o
M " & (@ !\ 9 !, %n
M ) "8 K '0 I H %@ "4 m
M N #P < W “T . ’ #L Z l
M L !<B@0(0P !( 9 ;(H$”"D 2 k
M( 'R
! AC $@ “< “,B@0(-P !( P G(H$”$H 2j
M .@ -R^! @ $0#Q_T, “LB@0(:stuck_out_tongue_winking_eye: !( !, @+$” 0 i
M 1 !( ZP$ BH! @ $0#Q_U( “\B@0(#@ !( !9 A+$h
M” 0 1 !( 8 ,R
! C= $@ &@ “(OP0(! !$ $@!U g
MC+$” 0 1 !( ?@ -R*! BL $@ (4 “0OP0(! !$ $@”,f
M [(H$“T 2 DP #BG! @2 0 $@ * )T #\B@0()0 !( e
M “D #(L$”)T 2 J@ !R+! A@ $@ + LBP0(( d
M !( “X /(L$”$4 2 Q0 $R+! A@ $@ ,T “4OP0(c
M! !$ $@#4 7(L$” X 2 Y &R+! C; $@ .L !\b
MBP0(3@ !( #R C(L$"$H 2 ]P )R+! A6 $@ /\ a
M “LBP0(50 !( & 0 O(L$”"< 2 Y $ “R-! @ $0 z
M T! #,BP0(!0$ !( 3 0 W(L$"/T 2 ‘@$ .R+! @B y
M$@ /D! !DOP0( !$ ?\D 0 (L$",<" 2 *P$ -“8! AEx
M#@ $@ * # ! ,C 0(AP$ !( W 0 '(P$”# 2 /@$ “R,w
M! @] $@ % ! \C 0(-P !( !8 0 3(P$”#T 2 7P$ v
M %R,! @U $@ &8! !LC 0(+ !( !Q 0 F+$"% 1 !( u
M>0$ 'R,! A# $@ '! #HOP0(! !$ $@"& 0 C(P$"%0! 2t
M @$ &2
! @ $0#Q_XT! <O@0( !$ ?% @ J,$$" s
M 1 /‘HP$ )R,! A? $@ *H! “LC 0(( !( “O 0 O(P$r
M” L 2 M@$ ,R,! AK $@ +X! #<C 0()P !( ## 0 q
M[(P$”“4 2 R0$ /R,! A] $@ ,! ,C00(80 !( #8p
M 0 '(T$”.$ 2 &QI8F,N<V\N,@!C;&]C:U]G971T:6UE ‘-T0!P<FEN=&8 <W1R97)R;W( 9G!A=&AC;VYF %]$64Y!34E# ‘!W<FET938Tn
M &5R<FYO &=E=’!I9 !?0W1Y<&4 <V5T=F)U9@!?<WES<&%G95]P=’( 7U1Om
M=7!T86( ;6%L;&]C &]P=&%R9P!F9FQU<V@ 7U]U9&EV9&DS ‘-T<G1O; !Ul
M;F%M90!W<FET90!S=’)I8VUP &UE;5]O9F9S970V- !F<’)I;G1F &]P=&ENk
M9 !?7V=E=%]E<G)N;U]P=’( 9&5V8W1L '-I9VYA; !R96%D '!R96%D-C0 j
M=6YL:6YK ‘-T<G1O:P!F<F5A9 !?:6YI=%]L:6)C &9O<&5N &=E=&]P= !Mi
M86EN &9C;&]S90!S=’)C;7 0V]N;F5C=%-E<G9E<DEN9F\ <W!R:6YT9@!Ah
M=&5X:70 ;6UA<#8T &9S=&%T=F9S-C0 7U-T9&5R<@!F<WEN8P!O<'1O<'0 g
M9G=R:71E %]‘3$]"04Q?3T9&4T547U1!0DQ%7P!R97=I;F0 97AI= !F:6QEf
M;F\ ;’-E96LV- !O<&5N &9C;G1L &-L;W-E ‘9F<’)I;G1F &9T<G5N8V%Te
M938T %]B=&5X= !?971E>'0 7V5D871A %]?8G-S7W-T87)T %]E;F0 " d
MOP0(!0@ (2
! @%“P B+$” 4- “,OP0(!0X )”! @%$ E+$" 49c
M "8OP0(!2\ .B
! @%,0 +X$" <! LO@0(!P( #"^! @’ P -+X$b
M" <$ XO@0(!P4 #R^! @’!P 0+X$" <
!$O@0(!PP $B^! @’#P a
M3+X$" <1 !0O@0(!Q, %2^! @’% 6+X$" <5 !<O@0(!Q8 &"^! @‘z
M%P 9+X$" <8 !HO@0(!QH &R^! @’&P <+X$" << !TO@0(!QT ‘B^y
M! @’’@ ?+X$" <? " O@0(!R (2^! @’(@ B+X$" <C “,O@0(!R0 x
M )”^! @’)@ E+X$" <H "8O@0(!RD )R^! @’
@ H+X$” <K “DO@0(w
M!RP B^! @’+0 K+X$" <N “PO@0(!S +2^! @’,@ N+X$” <V “\v
MO@0(!S< ,”^! @’. Q+X$" <Y #(O@0(!SH ,R^! @’.P T+X$" <\u
M #4O@0(!ST -B^! @’/@ Z ,> #" #-2"^! C)22^! @ R4Ht
MO@0(: #IX/
___\E++X$"&@( Z=#)3"^! AH$ .G s
M_R4TO@0(:!@ #IL/
\E.+X$"&@@ Z:#
)3R^! AH
.F0r
M_____R5 O@0(:# #I@/\E1+X$"&@X Z7#_____)4B^! AH0 q
M .E@_____R5,O@0(:$@ #I4/
\E4+X$”&A0 Z4#)52^! AHp
M6 .DP_____R58O@0(:& #I(/____\E7+X$"&AH Z1#
)6"^o
M! AH< .D R5DO@0(:’@ #I/[\E:+X$"&B Z>#^n
M)6R^! AHB .G0_O___R5PO@0(:slight_smile: #IP/[
\E=+X$"&B8 Z;#^m
M
_)7B^! AHH .F@_O___R5\O@0(:
@ #ID/[\E@+X$"&BP l
MZ8#^
)82^! AHN .EP_O___R6(O@0(:, #I8/[\EC+X$"&C(k
M Z5#^
)9"^! AHT .E O___R64O@0(:-@ #I,/[\EF+X$j
M”&C@ Z2#^)9R^! AHZ .D0_O___R6@O@0(:confused: #I /[\Ei
MI+X$"&CX Z?#]
):B^! AH $ .G@?R6LO@0(: @! #IT/W_h
M
\EL+X$"&@0 0 Z<#]
);2^! AH& $ .FP?R6XO@0(:" ! #Ig
MH/W
\EO+X$”&@H 0 Z9#])<"^! AH, $ .F ?R7$O@0(:#@!f
M #I</W
\ER+X$"&A 0 Z6#]
)<R^! AH2 $ .E0
?R70O@0(e
M:% ! #I0/W
\EU+X$"&A8 0 Z3#])=B^! AH8 $ .D@?__3DE!d
M3="8! B)TXMT) "-3"0$C02QC7@$C5 (@W@$ '0,C78 BP*#P@2%P’7W4U)7c
M45;H>?[__VBHJ 0(Z-
^^#Q 3HS_S<%@+$" #H3 L %#H(O__b
M
\R058GE@^P44^@ 6X’#A# .L6C78 BX/(C5 $B9/(BP#a
MT(N#R/X,X ‘7B6\G#C78 58GE@^P44^@ 6X’#2# %O)PY"058GEz
M@^P(@\3:->H! AHVJ@$"&B8OP0(Z$
]
^-10R#Q/Q0_W4(:)B
! CH^
[y
M_X/$((/$^&C?J 0(:)B
! CH)OW__X/$]&H!Z’S^__]5B>6#[ B#Q/3
-0# x
M! CH:/W__X/$]&H!Z%[^__^)]E6)Y8/L%%.+70B*10R(1?N#Q/QJ (U%
%!3w
MZ(W_
^)P8/$$(7)?D Y7?QT.XM%
^^$&,OP0(B@10/$MU!<'A"NL3/$UUv
M",'A%.L
C78 /$=U!L’A’O]%
(M%
( X '4&B<CK&8GV@\3X#[Y%^U!HX:@$u
M".@6____N/
^+7>C)PU6)Y8/L$%93BW4(BTT,N_BH! B)RL’Z"HG0P> t
M.<AU+8G1N_FH! C!^@J)T,’@"CG(=1J)T;O[J 0(P?H
B=#!X HYR’4’B=&[s M_:@$"%-1:/^H! A6Z W]__^)(UEZ%M>R<-5B>6#[!13BUT(@SV P00( '09r
M@\3X:@#
,^B%
?@\00B4,@C="?K,(/$_&BV 0 H43 ! @- @, %#q
M,^A>
?
@\00B4,@C=0OHSOO__XL ZP2)]C’ BUWHR<.058GE@^P05E.+p
M=0B#/8#!! @ =""#Q/AH!D$"/\VZ"W__^#Q!")1@R%P’0@,=OK+8UV (/$o
M^&@'J00(S;H#?S__X/$$(E&#(7 =0OH;OO__XL8ZP2)]C’;A=MU?H,]@,$$n
M" !U/(,]1, $" !T,X/$
%$P 0(#0(# !0:@2#Q/3_=@SHE?S__X/$$%#Hm
MO/S__X/$$(/X_W4’Z![__^+&(7;=36#/:3!! @ ="S-7C!! B#/43 ! @ l
M#Y7 )?\ !P%!J /]V#.A>^O__@C=06[# (G8C67H6U[)PXUV %6)k
MY8/L"(M%"(/$]/]P#.AD_/R<.)]E6)Y8/L"(M%"(/$]/]P#.A<^@\00j
MA<!T">B@^O__BP#K C’ R<-5B>6#[!B+30B#/8S!! @ ?BJAA,$$")F#Q/Q2i
M4/]Q#.@S
/@\00A<!T#.AG^O__BP#K+8UV #’ ZR:#Q/1J &H(H83!! B)h
M1?C!^!^)1?R-1?A0: X""(#_<0SH1OK
\G#58GE@^P05E.+=0B#/8S!! @ g
M?CB#Q/3_=@SH=?O__X/$$(G!BQ6$P00(B=“9@\3\4E!1Z+S[^#Q!"%P’0)f
MZ/#Y
^+&.L[,=OK-X/$]&H :@BAA,$$”(E%^,‘X’XE%(U%^%!H#@((@(/$e
M]/]V#.@?^
@\004.C&^?_B<.#Q""#Q/3=@SH-OG__XG8C67H6U[)PY!5d
MB>6#[ B+10AJ &H :@#
< SH]?K__S’ R<.058GE@^P(BT4(@\3T_W ,ZSZc
M
_\QP,G#58GE@^P(BT4(@\3T_W ,Z’3Z__^#Q!"%P’0)Z$CY__^+ .L",<#)b
MPU6)Y8/L%%.+70B#Q/3_<PSHN_C__X/$$(7 =1Z#Q/2#Q/3_<PSH=OK__X/$a
M$%#H+?K__X/$$(7 = KH ?G__XL ZP.0,<"+7>C)PY!5B>6#[ B+10B#Q/S_z
M-7C!! C_< 3_< SHE_C__XG"@\00@K=0GHR/C__XL ZP\QP#L5>,$$"'0%y
MN 4 #)PXUV %6)Y8/L"(M%"(/$](M0$(M(%%%2_S5XP00(W $W ,Z-OWx
M
^)PH/$((/Z_W4)Z’SX__^+ .L/,< [%7C!! AT!;@% R<.-=@!5B>6#w
M[ B+10C_< QJ ?\U>,$$"/]P!.AX^?__@\00@_@!= SH._C__XL ZP6-=@ Qv
MP,G#58GE@^PHBTT(QT7T @ (M!$(M1%(E%Z(E5[
%XP00(B47PBT$(B47Xu
MBT$$B47@\3T:@!J&(U%Z%!H#P(8@/]Q#.CX]R<.)]E6)Y8/L"(M%"(/$t
M
/\U>,$$"/]P!/]P#.CW]B<*#Q!"#^O]U">BX]BP#K#S’ .Q5XP00(s
M= 6X!0 ,G#C78 58GE@^P(BT4(@\3TBU 0BT@445+
-7C!! C
< 3
< SHr
MN
?__XG"@\0@@K=0GH;/?_XL ZP\QP#L5>,$$"'0%N 4 #)PXUV %6)q
MY8/L"(M%"/]P#&H!S5XP00(W $Z*CW^#Q!"#^ %T#.@K]BP#K!8UVp
M #’ R<-5B>6#[“B+30C’1?0! BT$0BU$4B47HB57LH7C!! B)1?”+00B)o
M1?B+002)1?R#Q/1J &H8C47H4&@/ AB W$,Z.CV
)PXGV58GE@>S, n
M5U93BW4,BUT0@\3\C84_4(M%"%#H#;U#H>??X/$$(/X_W1%@[U$m M____ 74\BU48QP( B?<QP/RY*@ /.K@\3XH0NI! B)1CR@#ZD$"(A&l M0(U&/%!3Z’;U__^#Q!#I,P$ (GVC;UL____,<#\N24 #SJ\>%;/___V k M “)A7S___]FQT6” ")183&18 QD60$L9%E&#&18$&QT6(- ,=%C& j
M "#Q/1J &B4 C85L____4&@’##3 BT4(4.@&]O
@\0@A<!U>XJ%<i
M_R0
/ %U;XM%L(D#BT6TB4,$BT6XB4,(BT6\B4,QD,0 (U+#SG9=B$/OE,/h
MZPZ-=@#& 0!).=EV$ ^^$:&$OP0(9O<$4$0!=>@YV74/H1"I! B) Z 4J00(g
MB$,$BT6@@.2 9CT%@ ^4P"7
BU48B0+K&*$0J00(B0.@%*D$"(A#!(M-f
M&,<! (/$^%:+10A0Z%V
^#Q!"#^/]U’(GW,<#\N2H #SJZ$5J00(e
MB48\H!FI! B(1D"#Q/2+1110Z,T
^#^/]U.8M]%#’ +E! 0 \ZNJH1JId
M! B+512) J$>J00(B8(" @ H2I! B)@@0$ “A)JD$”(F"" 0 (VE/c
M_UM>7\G#B?95B>6![.P !75E/'A1S
\ @WT(P^$D (V](/b
M_XN%’/RY. /.K,=O’A23! .P5TP 0(#XWY OWC ! B-a
M=@"+10C!X >—B#Q/1J &C@ C84@4&@’".# S5PP 0(Z&;T^#z
MQ""Z +D A<!U!HM5R(M-S(D4/HE,/@1#X4D
.QUTP 0(?*GIy
ME0 (-]# /A(L “^ +\ ,=L['73 ! A]*XL-=, $”(L$W?C x
M! B+%-W\P 0(P3=>, $"!L4W7S ! @!QA’70SG+?-MJ &CH P 5U;HJ0\ w
M (/$$(G!B=.A=, $"(T$@ ’ #Z]%#)E24%-1Z(H/ ")P87 ?0(QR8/Y9’X%v
MN60 "X9 "G(B84<____BX4<____C:4(6UY?R<.)]E6)Y8M%",< u
M 0 (U5$.L,C78 QP @($BT+\A<!U,G#B?95B>6#[ Q75E,Q_X/$t
M^&@JJ00(W4,ZS2+70CK!HUV (/#"(,[ '02@\3X5O\SZ/'R
^#Q!"%P’7Fs
M"WL$@
=!B#Q/AH
JD$"&H Z&/S
^)QH/$$(7V=;Z)^(UEZ%M>7\G#58GEr
M@^P,5U93BW40B?>P /RY
NB<CWT(U!O^Z^@$"(M="(,[ '0KC78 q
MBT4,(T,$=!C
,U)H+*D$"%?HBO/__P’‘NC&I! B#Q!"#PPB#.P!UV(GPC67Hp
M6UY?R<.)]E6)Y8’L/ 8 %=64XM="(M]#,<%>,$$" @ #’!83!! @ 0o
MQP5 P 0( 0 ,<%1, $" #’!8S!! @ QP6 P00( ,<%D,$$n
M" #’!3C ! @ QP6(P00( ,<%I,$$" #’!4C ! @ m
MZ8$! “-=@”#P,:#^#H/AW(! #
)(7XJP0(D,<%D,$$" $ #I6P$ )!Jl
M &A(P 0(:#C ! AHB,$$".A._O__QP6DP00( 0 .DR 0 C78 :@!H2, $k
M"&B(P00(:#C ! CH)O[
^D4 0 D,<%0, $" #I!P$ )#’!8S!! @!j
M Z?< "0QP6,P00(
^GG D(/$^/\UD+$"&CHO00(Z 7^+i
M!43 ! BC1, $"(/$$(/X_P^%O (/$^/\UZ+$"&A$J00(Z:$ “-=@!Jh
M &B(P00(:#C ! AH2, $”.B:?__Z8@ “0@\3X#[X%Z+$”%#-9"! CHg
M[O/_Z-XP00(ZVJ-=@#’!8#!! @! ZUYJ &A(P 0(:#C ! AHB,$$".A2f
M
?ZT.#Q/@/O@7HOP0(4/\UD+$".BJ_HX3!! CK)HUV (/$^/\UZ+$e
M"&A?J00(ZPZ#Q/C
->B
! AH>JD$".@-_
@\00@\3:#.I! A74^@7?_d
M@\00@C#X5G_O__H83!! @[!7C!! A]#KX! HWC!! CK(XGVBS6$P00(c
M3HGPF?<]>,$$"(UP 8GP#Z%>,$$"*.$P00(H8B
! @/MT &HW3 ! B#Q/QJb
M 6B/J00(:&# ! CH[?#__X/$$(/$^&H :&# ! CH>
'X/$$-PP 0(@_C_a
M=0KHZ>___XL ZP.0,<"%P’0?@\3@\3T4.@"[___4&B;J00(:2I! CH1O+z
M_X/$(#L=E+$"'X+H92
! B+!(?K!I"XLZD$"
, P 0(N#B/! B#/8C!! @ y
M= 6XG($"
,8P 0(N(20! B#/8C!! @ = 6XG) $",<P 0(@SV P00( '4Ix
M@SV,P00( '4)@STXP 0( '07N,20! B#/8C!! @ = NX
)$$".L$B?8QP*,@w
MP 0(@SV P00( '48N,“1! B#/8C!! @ = RXW)$$”.L%C78 ,<“C), $”(,]v
M@,$$" !U((,]0, $" !T%[CTD00(@SV(P00( '0+N!R2! CK!(GV,<“C*, $u
M”(,]@,$$" !U.(,]., $" !U)X,]B,$$" !U%KAHD@0(@SU(P 0( '0:NR2t
M! CK$Y"X^)($".L+D+@LDP0(ZP.0,<“C+, $”(,]., $" !U
(,]B,$$" !Us
M%[AXDP0(@SU(P 0( '06N+R3! CK#XGVN B4! CK!I"X/)0$"*,PP 0(@STXr
MP 0( '05@\3:@!J &K
: “P!H L .L<@SV0P00( '0W@\3:@!J &K_q
M: ( " !H , /\U>,$$”&H Z!?O
^#Q"“C!, $”(/X_W51Z 7N
_^+ .LEp
MD(/$]/\U>,$$".AR[?@\00B<*)%03 ! @QP(72=06X# (7 =!^#Q/R#o
MQ/10Z/SL
]0:+JI! AHID$".A /@\0@C87L^?@\3T:@!0_S5XP00(n
M:O__-03 ! CH>NW__X/$(#'2A<!U!HN5[/G__XD5", $",<%$, $" #'m
M!13 ! @ @\3XC87D^?4&H Z&#L^#Q/1H , $"
$8P 0(]"#Q""%l
MP’0N@\3@\3T4.AN[/4+C1J00(@SV P00( '0%N,RI! A0:2I! CHH^k
M_X/$((/$](V%X/G__U"-A5#Z
]0C84^O__4(V%6/__U"#/8C!! @ =!2#j
MQ/3
-0S ! CH1N[__X/$$.L&D
$,P 0(4.@!]O
@\0@@SV P00( ‘59@\3Xi
M:#".! AJ NC:[/@\00@SV,P00( '4)@STXP 0( '1W@\3T: # ! BA(, $h
M"/0@\00A<!T88/$(/$]%#HL^O__U!HW:D$"&BDJ00(Z/?N___K0)"#O>#Yg
M
\ =#F#Q/1J &H :@!H.0P (,]B,$$" !T%(/$]/\U#, $".BB[?@\00f
MZP:0H0S ! A0Z$'L
^#Q""#Q/B-A=CY__]0:@#H’>O__XN%Y/G__XN5V/G_e
M_RG"B=")Q’G!2G’C3RXC9WD^?BT,$BXW<^?*<&X@]X;0
?IB=#!^!+!d
M^1\IR(T4^(D5F,$$"(/$$(,]@,$$" /A:T! "#Q/AJ &H Z.CV__^#Q/A3c
M:@#HL>K__S’;@\0@.?-]3(/$]&@ P 0(H2S ! C_T(/$$(7 =!^#Q/R#Q/10b
MZ+;J__]0:.6I! AHID$".CZ[?@\0@H7C!! B9 040P 0($144P 0(0SGSa
M?+2#/4# ! @ ='F#Q/AJ#H,]B,$$" !T%8/$]/\U#, $".B7[/
@\00ZP>)z
M]J$,P 0(4.AEZO__@\00A< /G\ E_P - P 0(A<!T-8/$]&@ P 0(H2C y
M! C_T(/$$(7 =!^#Q/R#Q/10Z!WJ__]0:.NI! AHI
D$".AA[?@\0@@\3Xx
MC878^?4&H Z,GI^#Q/AJ &H!Z.GU
^+A>3Y__^+C=CY__\IP8G(P>$%w
M
<&-#(B+A>CY__^+G=SY__\IP[^#WAM#B?CWZXG7B?C!^!)VL’Z’RG0C0S(v
MB0V4P00(@\0@@\3X46K_Z)7U__^C?,$$",<%$, $" #’!13 ! @ u
M@\3T: # ! BA), $"/0@\0@A<!T’X/$(/$]%#H7NG__U!H:D$"&BDJ00(t
MZ
+L__^#Q""#Q/AJ &H Z#OU__^#Q/B-A>3Y__]0:@#H_NC__S’;@\0@.?-]s
M39"#Q/1H , $"$PP 0(_]"#Q!"%P’0?@\3@\3T4.@"Z?4&CXJ00(:2Ir
M! CH1NS__X/$(
%XP00(F0$%$, $"!$5%, $"$,Y\WRT@\3XC878^?4&H q
MZ)?H
^#Q/AJ &H!Z+?T
^+A>3Y__^+C=CY__\IP8G(P>$%
<&-#(B-G>3Yp
M__^+0P2+O=SY__\IQ[B#WAM#]^^)T,'X$HGZP?H?="-#,B)#:#!! B#Q""#o
MQ/A1:O_H9/3__Z,\P 0(@\3X4VH Z"CH__^#Q""#Q/1H , $"
$<P 0(]"#n
MQ!"%P’0?@\3@\3T4.@SZ/4&C]J00(:*2I! CH=^O__X/$((,]@,$$" !Um
M4X/$]/\U , $".@IZ?
@\00A<!T"NC-Z/BP#K Y QP(7 =!^#Q/R#Q/10l
MZ.;G
]0: .J! AHID$".@JZ__@\0@@\3X:@!J NBWZ/@\00@\3XC878k
M^?4&H Z(/G^+A>3Y
^+C=CY__\IP8G(B</!XP4IPXT<F(N%Z/G__XN-j
MW/G__RG!OX/>&T.)^/?IB=>)^,'X$L’Y’RG(C138B16<P00(@\3X@\00N!2Ji
M! B#/8#!! @ = 6X#ZH$"%!H’ZH$".@YY___C950^O__C854_O__4(V%4OS_h
M_U!2:#VJ! CH&N?__X/$((N%8/__PN%9/P^$B0 (/$](N-8/XN=g
M9/RN-</QN==/XG(B=H/I,(“P> " <@1VHG!B=,/I,L"P>$” <@1f
MV@^DP@+!X (#A6#
\3E63
^#P/^#TO^+C6#
^+G63
]345)0Z%H#e
M “#Q!!24(U%E%”-A3SZ
]0:%"J! CHA.;X/$(.LUC78 @\3\C5644HV%d
M//K__U"#Q/A04N@$Z/
@\00NG^J! B%P’0%NFZJ! A2Z$WF
^#Q!"#Q/2#c
M/3C ! @ =0F#/9#!! @ = FXC
H$".L’B?:XD:H$"%"#/8#!! @ =1*#/8S!b
M! @ =0F#/3C ! @ = >XFH$".L%N/BH! A0N/BH! B#/8#!! @ =0Z#/4# a
M! @ = 6XHZH$"%"#/8#!! @ =42#/43 ! @ =#N#Q/P/MP6KJ@0(9HF%]/G_z
M_Z"MJ@0(B(7V^?__C87T^?4/\U1, $"&CHO00(Z(SS^#Q!#K"(UV +CXy
MJ 0(4+CXJ 0(@SV(P00( '0.@SVDP00( '0%N
ZJ! A0@STXP 0( '4H@SV(x
MP00( '47N+RJ! B#/4C ! @ =!:XNH$".L/B?:XOZH$".L&D+C%J@0(4(/$w
M^/\U>,$$"(V%)/K__U#H?>G__X/$$%"#Q/C_-83!! B-A3#Z__]0Z&3I__^#v
MQ!!0:.“J! CH_N3__X/$,(,]@,$$” /A;X “#Q/C_-9C!! AH#JL$”.C;u
MY/__@\00@SV4P00( '03H83!! B9]SV4P00(B<‘K!8UV #’)A?9T&HL5E,$$t
M"(G0P> %
="-!(+!X .9]Z)P^L",=N#Q/B)R(7)?0:-@?# E /S__XGs
M
<)T,’@!2G0C02"C3S% +IG9F9FB?CWZL’Z#(GXP?@?<)2B<B%R7T&r
MC8’
P P?@4/\U?,$$"%/_-93!! AH0L$".@VY/@\0@ZR"0@[W@^?q
M ‘06@\3X_S68P00(:frowning:K! CH%.3__X/$$(,]H,$$" !T$&$P00(F?<]H,$$p
M"(G!ZP(QR87V=!J+%:#!! B)T,’@!2G0C02"P> #F??^B</K C’;@\3XB<B%o
MR7T&C8’
P )0#_^)RBG"B=#!X 4IT(T$@HTTQ0 "Z9V9F9HGP]^K!n
M^@R),'X’RG"4HG(A<!]!07
P P?@4/\U/, $"%/_-:#!! AHHL$".ASm
MX
@\0@@SV P00( ‘43@\3X_S6<P00(:.K! CH5./__S’ C:6X^?__6UY?l
MR<.0D)!5B>6#[#!75HM%"(M5#(M-%(M]$(E-Z(E%(E5Y(7)=3\YUW8+]>)k
M1>#IS@ )"#?1 =0V^ 0 (GP,=+W]XG’BT7DBU7H]>)P8E5Y(M%/?Wj
MB47@Z:( “-=@”+5>0Y5>AV#,=%X #IB0 ^]=>B#]A]U&XM-Z#E-i
MY’<’.7W<MZ)]L=%X $ #K:(UV ,=%^" “+1>B)\2EU^-/@B7WHBTWXh
MTVWHBU7H”=")1>")=/GBT7DB47TBTWXTVWTB?'3X(M5_(M-^-/J"=")\8E%g
MZ(M5]/=UX–E_(E%X(G1]^>)QCG
=P=U"#MU_‘8#TW@,<F+=>")=>R)3?"+f
M1>R+5?!>7\G#D)!5B>6#[!13Z !;@</$%0 BX/L
___@SC_="&)]HN#e
M[/XL ]"+@^S^-4/R)D^S__^#>/S_=>%;R<.)]E6)Y8/L%%/H d
M %N!PX 5 !;R<.0D.C?Y/P@ $]?4E-93D, c
M3U]$4UE.0P!/7U-93D, <G< )7,Z( * &EN=F%L:60@<VEZ92!F;W(@)RTEb
M8R< &L 30!’ “5D)7, <F( =RMB ‘!R;V, 9&ES:P!F<WES %%.6 V+G@ a
M=6YK;F]W;@ L "5S)7, ? Z86)D9F=’;SIP<CI2<W0Z &EN=F%L:60@87)Gz
M=6UE;G0@9F]R(”<M)6,G &UI<W-I;F<@87)G=6UE;G0@9F]R("<M)6,G ‘5Ny
M:VYO=VX@;W!T:6]N("<M)6,G “]P<F]C+R5D+V%S &-P=2!L;V%D “5S(&9Ax
M:6QE9” M(“5S ')W+G1M< !B=69F97(@86QL;V-A=&EO;@!O<&5N &9I;&4@w
M8W)E871E '!R96=R;W< =W)I=&4 9G-Y;F, <F5W:6YD ‘)E860 8VQO<V4 v
M9FEL92!D96QE=&4 4F5A9 !7<FET92]296%D %-E<75E;G1I86P@1FEL92 Eu
M<R!“96YC:&UA<FL* $]3.B @(” @("5S("5S(“5S”@!&:6QE<WES.B E<RP@t
M)7,L(“5L;&0E)2!F=6QL”@!&:6QE<WES.B E<RP@)7,* $9I;&5S>7,Z(“5Ss
M”@!M;6%P &UA;&QO8P L(’!R96=R;W=N “P@9G-Y;F, +” +”!S971V8G5Fr
M '!F9 !F9 !S=&1I;P!D:7)E8W1)3P !#;VYFq
M:6<Z(” E<R!F:6QE+" E<R!R96-O<F0@“5S)7,E<R5S)7,I+” E<PH 0W)Ep
M871E.B @)39L9"!M<V5C"@ o
M !7<FET93H@(" E-FQD(&US96,L(“4T9”!U<V5C+W=R:71E
"DL(“4S9"4En
M($-052P@)3)D+B4P,F0@36(O<V5C”@!3<&EN57 Z(" E-FQD(&US96,* m
M !296%D.B @(" E-FQD(&US96,L(“4T9”!U<V5C+W)E860H2P@l
M(“4S9"4E($-052P@)3)D+B4P,F0@36(O<V5C”@!$96QE=&4Z(" E-FQD(&USk
M96,
L)H$"-::! C6F@0(UIH$"-::! C F@0(UIH$"-::! C6F@0(UIH$j
M"-::! C6F@0(UIH$".“9! C6F@0(UIH$”-::! C6F@0(UIH$"-::! C6F@0(i
MUIH$"-::! C6F@0(;)H$"-::! C6F@0(UIH$"-::! C6F@0(UIH$"-::! C6h
MF@0(UIH$"-::! C6F@0(UIH$"-::! C6F@0(;)D$"‘R9! C6F@0(I)D$"-::g
M! C F00(T)D$"-::! C6F@0(UIH$"-::! C6F@0(UIH$"-::! CPF00(,)H$f
M"-::! A,F@0(>)H$")":! @ 0(" P,# P0$! 0$! 0$!04%!04%!04%!04%e
M!04%!08&!@8&!@8&!@8&!@8&!@8&!@8&!@8&!@8&!@8&!@8&!P<’!P<’!P<‘d
M!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<’!P<‘c
M!P<’!P<’!P<’!P@(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(b
M" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(a
M" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(" @(z
M" @(&+X$"-“H! @@ R*@$”! # J 0(0 #
___#+X$"/y
M
\ _____P #<O@0( !BB@0(<HH$"(! B2B@0(HHH$x
M"+
! C"B@0(THH$".! CRB@0( HL$"!+! @BBP0(,HL$"$+! A2BP0(w
M8HL$"’*+! B"BP0(DHL$"
+! BRBP0(PHL$"-+! CBBP0(\HL$" ,! @2v
MC 0((HP$"#
,! A"C 0(4HP$"&
,! ARC 0(@HP$"),! BBC 0(LHP$",,u
M! C2C 0(XHP$"/,! @“C00($HT$”"-! @! 0 P !$B@0(#0 t
M BH! @$ “($$” 4 “0A@0(!@ “”! @ “@( L 0 s
M%0 # '+X$” ( !H 0 % !$ 7 W(@$”!$ "<r
MB 0($@ $ 3 " ! #L ! 44Y8 “]Hq
M;VUE+VIG87)V97DO<V-R871C:”]C=G,O<V]F=’=A<F4O=71I;’,O<B]R=R]Np
M=&\O>#@V+V\O<G< 1T-#.B H1TY5
2 R+CDU+C(@,3DY.3$P,C0@o
M*’)E;&5A<V4I !‘0T,Z("A’3E4I(#(N.34N,R R,# Q,#,Q-2 H<F5L96%Sn
M92D $=#0SH@$=.52D@,BXY-2XS(#(P,#$P,S$U("AR96QE87-E0 1T-#m
M.B H1TY52 R+CDU+C(@,3DY.3$P,C0@’)E;&5A<V4I ‘P 1 !( G $ l
M #@ +BXO+BXO9V-C+VQI8F=C8S(N8P!8 D=.52!#(#(N.34N,R R,# Q,#,Qk
M-2 H<F5L96%S92D -@$! $0$XIP0((0%J 0(N $O<W)C+W1O;VQS+V=Cj
M8R]N=&\M>#@V+6\M;G1O>#@V+V=C8P E % 2 .4 X %]?=61I=FUOi
M9&1I- ( @!X @!5 B"% 4 $@“U . !N %4 “((4 !0 2 ,D h
M X &0 50 (@A@ % !( X0 #@ <G 8P # $(@@0 W !@ 2g
M ((! X %]?=61I=F1I,P!X @!5 B”$0$XIP0((0%J 0(08 XIP0(48!f
MJ 0('0 4 $@ Y 0 . !N %4 "((C 4 0 C !0 2 %P! Xe
M &0 50 (@B, "P “!0 00 !QX = !( ?@$ +(”? !$!.
<$d
M""$!-Z@$" 0 $ %@ !, $@“8 0 . !R='A?9&5F 0 #E c
M7U]U9&EV9&DS ]( $” # $ b
M B!! @ P " (@00( , P H(($" # 0 a
M )"&! @ P % “<B 0( , !@ W(@$” #z
M < $2
! @ P ( !,B@0( , “0 +(T$” y
M # H BH! @ P + # J 0( , # Y+T$x
M" # T R^! @ P . 4O@0( , #P w
M’+X$" # ! -R^! @ P 1 " OP0( , $@ v
M # !, P 4 , u
M%0 # !8 P 7 t
M , & # !D 0 ! #Q_PP ! s
M ?\4 C(T$" H (P .2]! @ 0 - “< 4r
MO@0( $ #P U C(T$” " H 2P ,B-! @ @ * %8 q
M #HO00( $ #0!D #+X$" ! X # $ #Qp
M_PP ! ?]R $ /’_# $ o
M #Q_Q0 !,J 0( “@ C “+X$” ! T >P !”^! @ n
M 0 . (@ !,J 0( ( “@”> D
@$” " H 5@ R^m
M! @ 0 - D 8O@0( $ #P"V $ /’_% l
M ."-! @ * +L 0 ?\4 .
<$" H k
MQ0 .2L! @ 0 0 , ,\ !<B@0(0P !( #= , $"#@ 1j
M !( Y0 &R
! @I $@ .P !\B@0(8P !( #S C(H$"#< i
M 2 _ )R*! A* $@ 8! #$D 0(9 !( “@ - 0 W+X$h
M” 1 /’%@$ R! AO $@ !! " OP0(! !$ $@ E 0 g
MJ*@$" 1 /’
+ $ #C ! @$ $0 2 #,! “\B@0(#@ !( Zf
M 0 A+$” 0 1 !( 00$ ,R*! C= $@ $D! # D00(&P !( e
M"@!2 0 B+$" 0 1 !( 7P$ (R_! @$ $0 2 &@! #<D00(& d
M !( “@!S 0 W(H$“P 2 >@$ )"_! @$ $0 2 ($! <D@0(c
M2P !( “@”
0 4(X$”(0 2 H E $ .R*! BM $@ )L! Xb
MCP0(8P !( “@“B 0 /, $” 0 1 !( JP$ #BG! @2 0 $@ * +4!a
M ! P 0(! !$ $@”^ 0 1(H$” 2 @ Q $ R2! A) $@ z
M ,T! #\B@0()0 !( #4 0 #(L$")T 2 V@$ $3 ! @$ y
M$0 2 .$! !(P 0(! !$ $@#H 0 )$$")< 2 H \0$ !R+! A@x
M $@ /<! LBP0(( !( #_ 0 /(L$"$4 2 # ( $R+w
M! A@ $@ !0" (E 0(- !( “@ = @ ^)($”#0 2 H )P( v
M &B2! A! $@ * “” !@P 0(& $ !$ $@ T @ >,$$" 0 1 !( u
M/P( )2_! @$ $0 2 $8" #@C00(4 !( “@!, @ 7(L$” X 2t
M 7 ( &R+! C; $@ &," “(E@0(5@$ !( “@!H @ -(T$”%< s
M 2 H ;P( 'S!! @$ $0 2 'D” !\BP0(3@ !( " @ C(L$r
M"$H 2 A0( )R+! A6 $@ (T" #TD00(
!( “@“4 @ q
MK(L$”%4 2 FP( B8! AD $@ * 0" " P00(! !$ $@“Mp
M @ A,$$” 0 1 !( MP( +R+! @G $@ +X" LC00( !$ o
M"@#% @ S(L$" 4! 2 RP( -R+! C] $@ -8" #LBP0((@ n
M !( #< @ 9+$" 1 /’_Z ( /R+! C’ @ $@ ." #@EP0(m
M)@ !( “@#Y @ T)@$”&4. 2 H _@( #".! @> $@ * 8# “(l
MP00(! !$ $@ / P #(P$”(<! 2 %@, "R3! A
$@ * !#k
M “\DP0(20 !( “@ G P '(P$”# 2 +@, (B4! C^ 0 $@ j
M #0# LC 0(/0 !( !& P J
@$” 2 L 3 , #R,! @W i
M$@ %0# !LF 0(8@ !( “@!< P 3(P$”#T 2 8P, -2.! ADh
M $@ * &X# !<C 0(-0 !( !U P ;(P$”"P 2 @ , )B_g
M! A0 $0 2 (@# !\C 0(0P !( “. P Z+$” 0 1 !( E0, f
M (S!! @$ $0 2 )T# "0P00(! !$ $@“I P E,$$” 0 1 !( e
MLP, (R,! A4 0 $@ +H# !DOP0( !$ ?! P Z+T$"" 1d
M T S@, !R^! @ $0#Q
^0# “HP00( !$ ?_I P G(P$”%\ c
M 2 \ , #R4! A
$@ * /@# “LC 0(( !( #] P O(P$b
M” L 2 ! 0 )C!! @$ $0 2 $ “<CP0(Y0 !( “@ 8! a
MS(P$”&L 2 ( 0 -R,! @G $@ “4$ #LC 0()0 !( Kz
M! G) $””@ 2 H -00 (20! @6 $@ * #T$ #\C 0(?0 !( y
M !#! G,$$" 0 1 !( 3@0 #!! @$ $0 2 %<$ ,C00(80 x
M !( !@! >),$"$$ 2 H 9P0 !R-! CA $@ ‘,$ “DP00(w
M! !$ $@ 8W)T8F5G:6XN8P!?7U!)0U]? &=C8S)?8V]M<&EL960N ’ Nv
M,P!?7T143U)?3$E35%]? %]?9&]?9VQO8F%L7V1T;W)S7V%U> !F:6YI7V1Uu
M;6UY &9O<F-E7W1O7V1A=&$ 7U]#5$]27TQ)4U1?7P!C<G1E;F0N8P!?7T-4t
M3U)?14Y$7U\ 7U]D;U]G;&]B86Q?8W1O<G-?875X &EN:71?9’5M;7D 7U]$s
M5$]27T5.1%]? ')W+F, ;&EB9V-C,BYC %]?8VQZ7W1A8@!C;&]C:U]G971Tr
M:6UE $-O;G1E>'0 <W1R8W!Y ‘!R:6YT9@!S=’)E<G)O<@!F<&%T:&-O;F8 q
M9W)O=T9$ %]$64Y!34E# '!W<FET938T &5R<FYO %]E=&5X= !5<V5$;6$ p
M9V5T<&ED %]#='EP90!S971V8G5F ')E=VEN9$9$ %]S>7-P86=E7W!T<@!?o
M5&]U<'1A8@!R97=I;F1&24Q% &UA;&QO8P!O<'1A<F< <WEN8T9)3$4 <&%Rn
M<V5S:7IE &9F;'5S: !O<&5N1D0 3&]A9%)E860 7U]U9&EV9&DS %5S949Sm
M>6YC %]I;FET '=R:71E<$9$ ‘-T<G1O; !U;F%M90!/9FQA9W, 57-E4’)Wl
M &=R;W=&24Q% '=R:71E '-T<FEC;7 ;65M7V]F9G-E=#8T &9P<FEN=&8 k
M<F5A9$9)3$4 =W)I=&5&24Q% ‘=R:71E1D0 261L90!296-O<F13:7IE &]Pj
M=&EN9 !F871A; !?7V=E=%]E<G)N;U]P=’( 9&5V8W1L &ED;&4 7W-T87)Ti
M $QO8617<FET90!S:6=N86P <F5A9 !P<F5A9#8T '-Y;F-&1 !U;FQI;FL h
M;6%T8VAO<'0 4F5A9$]N;'D 5&]T86Q3:7IE '-TG1O:P!?8G1E’0 9G)Eg
M860 7VEN:71?;&EB8P!F;W!E;@!?7V)S<U]S=&%R= !G971O<‘0 97AC;‘5Sf
M:79E &UA:6X 8VQE86YU< !5<V53=&1I;P!F8VQO<V4 =W)I=&5$34$ <F5Ae
M9’!&1 !S=’)C;7 <75E<GD 0V]N;F5C=%-E<G9E<DEN9F\ 7V9I;FD <W!Rd
M:6YT9@!T97AT;W!T &%T97AI= !P<F5T='ES:7IE &UM87 V- !F<W1A='9Fc
M<S8T %]3=&1E<G( 9G-Y;F, ;W!T;W!T %!R96=R;W< 06QI9VY”=69F97( b
M5&EM95=R:71E &9W<FET90!?961A=&$ 3W!E;D9L86=486=S %]‘3$]"04Q?a
M3T9&4T547U1!0DQ%7P!?96YD ‘)E=VEN9 !R96%D1$U! &5X:70 9FEL96YOz
M %1I;65#<F5A=&4 ;W!E;D9)3$4 ;’-E96LV- !O<&5N &9C;G1L &-L;W-Ey
M1DE,10!C;&]S949$ &-L;W-E %1I;65$96QE=&4 5&EM95)E860 =F9P<FENx
M=&8 <F5A9$9$ &9T<G5N8V%T938T %5S94)U9@ E0R M(&)E;F-H;6%R:R!Fw
M:6QE(’-E<75E;G1I86P@<F5A9"]W<FET92!T:’)O=6=H<'5T(" H:F=A2D
"B5#(%MO<'1I;VYS72H@6V9I;&5="@I/<'1I;VYS.@H@(“UA(” @(%5Su
M92!P86=E+6%L:6=N960@;6UA<"=D(&UE;6]R>2!F;W(@8G5F9F5R<R H9&5Ft
M875L="!I<R!M86QL;V,G9"D
(" M8B @("!3970@=&AE(’-T9&EO(&)U9F9Es
M<B!S:7IE('1O('1H92!R96-O<F0@<VEZ92 H9&5F875L="!I<R!“549325H@r
M*#%K2D(” M9" @("!5<V4@1$–1%]&4UE37T1)4D5#5%])3R H=7-E<B!$q
M34$I(&9I;&4@86-C97-S"B @+68@(" @1&@;F]T(&9S>6YC*“D@=&AE(&9Ip
M;&4@8F5T=V5E;B!W<FET92!A;F0@<F5A9”!P:&%S97,(" M9R @("!0<F5Go
M<F]W(‘1H92!F:6QE(’!R:6]R(‘1O(’=R:71E(&)E;F-H;6%R:VEN9R H9&5Fn
M875L="!T;R!A<’!E;F0I"B @+4<@(" @4’)E9W)O=R!T:&4@9FEL92 H;F@m
M,"UF:6QL
2!P<FEO<B!T;R!W<FET92!B96YC:&UA<FMI;F<(" M;R @("!3l
M970@97AT<F$@9FEL92!O<&5N(&9L86=S(“A/7U-93D,L($]?1%-93D,L($]?k
M4E-93D,I"B @+7 @(” @57-E(’!R96%D+W!W<FET92!F:6QE(&%C8V5S<R Hj
M9&5F875L="!I<R!R96%D+W=R:71E
0H@(“UR(” @(%)E8V]R9"!S:7IE(&9Oi
M<B!E86-H($DO3R!O<&5R871I;VX@&1E9F%U;'0@.&LI"B @+5(@(" @3W!Eh
M<F%T92!I;B!R96%D+6]N;'D@;6]D92 H9F]R(&5X86UP;&4@=&@=&5S="!#g
M1"U23TT@86-C97-S
0H@(“US(” @(%5S92!S=&1I;R]&24Q%B!F:6QE(&%Cf
M8V5S<R H9&5F875L="!I<R!F:6QE(&1E<V-R:7!T;W(I"B @+70@(" @5&]Te
M86P@<VEZ92!O9B!T:&4@9FEL92 H9&5F875L=" R-39-0H +G-Y;71A8@ Nd
M<W1R=&%B "YS:’-T<G1A8@ N:6YT97)P "YN;W1E, N:&%S: N9’EN<WEMc
M "YD>6YS=’( +G)E;"YB<W, +G)E;"YP;'0 +FEN:70 +G!L= N=&5X= Nb
M9FEN:0 N<F]D871A "YD871A "YC=&]R<P N9’1O<G, +F=O= N9’EN86UIa
M8P N8G-S "YC;VUM96YT "YD96)U9P N9&5B=6=?<'5B;F%M97, +FYO=&4 z
M44Y87W5S86=E y
M ; 0 ( #T@ 0(] !0 $ x
M(P < " “($$” @! ! "H %w
M @ B!! @( 0 F $ 0 ! 0 P “P ( v
M “@@@0(H ( / # % 0 0 0 . , " D(8$u
M”) & * @ ! $ ) @ )R(! B<” t
M0 0 2 ! @ !) "0 ( #<B 0(W @ &@! $s
M “0 0 ( 4@ $ & 1(H$”$0
( r
M ! %@ ! !@ $R
! A,"@ X ( ! q
M 0 != 0 8 LC00(+ T ‘P; 0 p
M8P $ & J*@$"*@H ( ! &D !o
M @ ,“H! C * ) 4 ( !Q 0 , n
M #DO00(Y"T “@ 0 =P $ # #+X$m
M” PN ( $ 'X ! P !2^! @4+@ l
M” ! "% 0 , <O@0(’“X , k
M 0 $ B@ 8 # W+X$”-PN “( !0 j
M $ " ), ( P (”_! B +P * ( ( i
M “8 0 M”\ )@ $ h
MH0 $ $PP “< 0 $ *@ !g
M #H,0 #@ 0 “X !P f
M 9”\ $P $ 0 ( e
M /8Q P”@ & #, $ $ D # F/ d
M>@0 0 "^ 0 H$ %T# c
M $0 , /U# #( b

  • ! a
    z
    end

I am getting about 30-40Mb/sec (cp -V says) on Adaptec 29160 with Seagate
U160 drives. Check your cabling, it my experience it is the most common
problem with SCSI. Watch for SCSI BIOS messages… sometimes it can degrade
the interface if there is something wrong.

“Mathew Kirsch” <mkirsch@ocdus.jnj.com> wrote in message
news:b8pbil$ooj$1@inn.qnx.com

Any way to speed up disk transfers? I started a move of about 32GB of
data early this morning from one 72GB disk to another on the same chain.
It’s still going at 4:41PM… I would think that it would take less time
than that! The system is not busy at all, the disk transfers seem to be
happening quietly in the background.

The SCSI adapter is listed as an “AIC-7892A Ultra160/m PCI SCSI
Controller” in the pci outout. Disks are IBM 72GB Ultra SCSI units. CPU
is a Pentium III 1.0GHz with 256MB of RAM.

I’ve tried to increase the cache size in the kernel build file, but all
I get is “Unable to allocate resource” at boot time, and the size of the
devb-aha8 process in memory stays at the default 12MB.

John Garvey wrote:

Yes, I’d expect that volume of data to take half an hour or so? Were
you just using “cp”? What did “cp -V” report as the throughput rate?

cp -V is giving me about 2300 kb/s.

Oh, and thanks a BUNCH for that so-called “disk benchmark” program. I
ran it, and it just wiped out the boot disk on that server!!! There’s
nothing that tells you how to use it (“use rw” gives impossibly cryptic
help text), so I ran it figuring it would tell me what to do. But no, it
just merrily started munching on the first disk it found, and now the
whole system is toast.

If you can’t tell, I’m thoroughly PI$$ED. Thanks a bunch QNX. Now I get
to all night rebuilding this server AGAIN.

Mathew Kirsch <mkirsch@ocdus.jnj.com> wrote:

John Garvey wrote:
Yes, I’d expect that volume of data to take half an hour or so? Were
you just using “cp”? What did “cp -V” report as the throughput rate?
cp -V is giving me about 2300 kb/s.

That seems far too slow; Igor claimed “cp -V” speeds of >10x that.

Oh, and thanks a BUNCH for that so-called “disk benchmark” program. I
ran it, and it just wiped out the boot disk on that server!!! There’s
nothing that tells you how to use it (“use rw” gives impossibly cryptic
help text), so I ran it figuring it would tell me what to do. But no, it
just merrily started munching on the first disk it found, and now the
whole system is toast.

It operates in the current directory by writing and reading a 256M
file in 8k units as the default operation. You can override any of
this as described in the usage message, but with no options is the
normal/expected method. It does nothing much more unusual than
open/write/read/close with a few system config queries, so I doubt it
itself trashed your disk. I run it on my system all the time for
performance analysis (I wrote all the filesystems), we run it in our
labs, and I’ve given it to other customers, and nowhere has it done
this. Perhaps Igor is brave enough to also try it?!? I suspect your
disk h/w is misconfigured (accounting also for the slow throughput),
and the intensive writing just showed it up (this is quite common, for
example if DMA cycle timings are incorrect, or cabling issues, etc).
I will see if our disk driver guy has similar h/w and can set up a
system to verify this …

John Garvey wrote:

It operates in the current directory by writing and reading a 256M
file in 8k units as the default operation. You can override any of
this as described in the usage message, but with no options is the
normal/expected method. It does nothing much more unusual than
open/write/read/close with a few system config queries, so I doubt it
itself trashed your disk. I run it on my system all the time for
performance analysis (I wrote all the filesystems), we run it in our
labs, and I’ve given it to other customers, and nowhere has it done
this.

Well, it did it here. As far as I can see, all I did was run “rw,” and
it munched the boot disk. Before I ran it, everything was fine. There
was no output from the command after several minutes, so I hit CTRL-C.
After I cancelled the command, none of the commands I typed in worked.
All indications point to this “rw” command overwriting everything on the
boot disk starting at sector 0. When I booted from CD, there wasn’t even
a partition table on the disk.

Perhaps Igor is brave enough to also try it?!? I suspect your
disk h/w is misconfigured (accounting also for the slow throughput),
and the intensive writing just showed it up (this is quite common, for
example if DMA cycle timings are incorrect, or cabling issues, etc).
I will see if our disk driver guy has similar h/w and can set up a
system to verify this …

The disk it munched was EIDE, primary master. It didn’t go anywhere near
the SCSI disks.

I believe you’re right about the misconfigured SCSI disks, but I don’t
know HOW they can be misconfigured. Everything’s on the default
settings, and I see no “go faster” setting, let alone any settings that
would change the behaviour of anything. They’re LVD SCSI disks, and I
believe the controller is LVD. We purposely bought an LVD external case,
and an LVD terminator. During the rebuild, I noticed when the system
boots up, the SCSI BIOS says it’s running at 16MB/s. So, I guess it’s
not a software issue. Too bad it took trashing the system to find that out.

Mathew Kirsch <mkirsch@ocdus.jnj.com> wrote:

Well, it did it here. As far as I can see, all I did was run “rw,” and
it munched the boot disk. Before I ran it, everything was fine. There

You didn’t, as root, run “rw /dev/hd0”, by any chance?

John Garvey <jgarvey@qnx.com> wrote:

I will see if our disk driver guy has similar h/w and can set up a
system to verify this …

We’ve just tested/benchmarked this controller with LVD, and got
38MB/s throughput (and no trashed disk). Have you enabled “Ultra160”
in the SCSI BIOS?

“John Garvey” <jgarvey@qnx.com> wrote in message
news:b9bqk3$9p4$1@nntp.qnx.com

John Garvey <> jgarvey@qnx.com> > wrote:
I will see if our disk driver guy has similar h/w and can set up a
system to verify this …

We’ve just tested/benchmarked this controller with LVD, and got
38MB/s throughput (and no trashed disk). Have you enabled “Ultra160”
in the SCSI BIOS?

I don’t think using U160 mode matters much, bus speed is hardly the
bottleneck here. He might have write cache disabled in the SCSI BIOS - I’ve
found that dramatically decreases performance with QNX, at least on older
2940 adapters.

– igor

John Garvey wrote:

John Garvey <> jgarvey@qnx.com> > wrote:

I will see if our disk driver guy has similar h/w and can set up a
system to verify this …


We’ve just tested/benchmarked this controller with LVD, and got
38MB/s throughput (and no trashed disk). Have you enabled “Ultra160”
in the SCSI BIOS?

The SCSI BIOS is set to the defaults. There is no specific “Ultra160”
setting that I can see. You can manually adjust the transfer speeds for
each SCSI ID, but they’re all set to 160 by default.

cp -V reports a transfer speed of ~5200kb/s right after a reboot, but it
slowly degrades to about 2200kb/s as time goes on. I don’t dare try that
“rw” utility again until you tell me how to use it. Yes, after three
unsuccessful attempts to run “rw” with no options, I tried “rw /dev/hd0”
because it seemed logical. It’s a BENCHMARK utility, so it probably
takes a device name as an argument, right? There’s nothing in the usage
message that says otherwise.

I’ve fiddled with all of the settings in the SCSI BIOS, and the problem
still exists.

We need to move about 200GB a day on these disks, once things get going.
If that’s going to take 36 hours, that’s unacceptable.

Igor Kovalenko wrote:

I don’t think using U160 mode matters much, bus speed is hardly the
bottleneck here. He might have write cache disabled in the SCSI BIOS - I’ve
found that dramatically decreases performance with QNX, at least on older
2940 adapters.

I don’t know what “U160 mode” you’re talking about. There are no
settings like that in the SCSI BIOS. There is no setting for “write
cache” either.

rw reports a dismal 4.3MB/s for writes, but 45MB/s for reads???

Mathew Kirsch <mkirsch@ocdus.jnj.com> wrote:
MK > cp -V reports a transfer speed of ~5200kb/s right after a reboot, but it
MK > slowly degrades to about 2200kb/s as time goes on. I don’t dare try that
MK > “rw” utility again until you tell me how to use it. Yes, after three
MK > unsuccessful attempts to run “rw” with no options, I tried “rw /dev/hd0”
MK > because it seemed logical. It’s a BENCHMARK utility, so it probably
MK > takes a device name as an argument, right? There’s nothing in the usage
MK > message that says otherwise.

I have to agree. Any utility that is destructive like that should:

  1. have a usage message
  2. put out a run time warning forcing the user to hit OK

Bill Caroselli <qtps@earthlink.net> wrote:

Mathew Kirsch <> mkirsch@ocdus.jnj.com> > wrote:
MK > cp -V reports a transfer speed of ~5200kb/s right after a reboot, but it
MK > slowly degrades to about 2200kb/s as time goes on. I don’t dare try that
MK > “rw” utility again until you tell me how to use it. Yes, after three
MK > unsuccessful attempts to run “rw” with no options, I tried “rw /dev/hd0”
MK > because it seemed logical. It’s a BENCHMARK utility, so it probably
MK > takes a device name as an argument, right? There’s nothing in the usage
MK > message that says otherwise.

I have to agree. Any utility that is destructive like that should:

2) put out a run time warning forcing the user to hit OK

Hm…

dd if=/dev/fd0 of=/tmp/image
(swap floppies)
dd if=/tmp/image of=/dev/fd0

Is fine.

dd if=/dev/fd0 of=/tmp/image
(swap floppies)
dd if=/tmp/image of=/dev/hd0

or

cp /bin/ls /dev/hd0

Those look bad… should cp warn you? Should dd?


-David

QNX Training Services
http://www.qnx.com/support/training/
Please followup in this newsgroup if you have further questions.

Bill Caroselli <qtps@earthlink.net> wrote:

Mathew Kirsch <> mkirsch@ocdus.jnj.com> > wrote:
MK > cp -V reports a transfer speed of ~5200kb/s right after a reboot, but it
MK > slowly degrades to about 2200kb/s as time goes on. I don’t dare try that
MK > “rw” utility again until you tell me how to use it. Yes, after three
MK > unsuccessful attempts to run “rw” with no options, I tried “rw /dev/hd0”
MK > because it seemed logical. It’s a BENCHMARK utility, so it probably
MK > takes a device name as an argument, right? There’s nothing in the usage
MK > message that says otherwise.

I have to agree. Any utility that is destructive like that should:

  1. have a usage message
  2. put out a run time warning forcing the user to hit OK

rm -rfv /

I don’t see an “OK” message…

shutdown -f

no “OK” message…

mv /bin /junk

no “OK” message…

gzip -9v /bin/*

no “OK” message…

cp /dev/null /.boot

no “OK” message…

I’m sure there are an infinite number of other examples…

Summary: You are given a gun. Shoot yourself in the foot and enjoy already.

Contrary opinion:

DOS> del *.c

no “OK” message…

DOS> del .

Big message saying “are you sure?”

:slight_smile:

Cheers,
-RK


Robert Krten, PARSE Software Devices +1 613 599 8316.
Realtime Systems Architecture, Books, Video-based and Instructor-led
Training, Consulting and Software Products at www.parse.com.

David Gibbs <dagibbs@qnx.com> wrote:

Bill Caroselli <> qtps@earthlink.net> > wrote:
Mathew Kirsch <> mkirsch@ocdus.jnj.com> > wrote:
MK > cp -V reports a transfer speed of ~5200kb/s right after a reboot, but it
MK > slowly degrades to about 2200kb/s as time goes on. I don’t dare try that
MK > “rw” utility again until you tell me how to use it. Yes, after three
MK > unsuccessful attempts to run “rw” with no options, I tried “rw /dev/hd0”
MK > because it seemed logical. It’s a BENCHMARK utility, so it probably
MK > takes a device name as an argument, right? There’s nothing in the usage
MK > message that says otherwise.

I have to agree. Any utility that is destructive like that should:

2) put out a run time warning forcing the user to hit OK

Hm…

dd if=/dev/fd0 of=/tmp/image
(swap floppies)
dd if=/tmp/image of=/dev/fd0

Is fine.

dd if=/dev/fd0 of=/tmp/image
(swap floppies)
dd if=/tmp/image of=/dev/hd0

or

cp /bin/ls /dev/hd0

Those look bad… should cp warn you? Should dd?

Aww ya beat me to it, Dave!

Cheers,
-RK


Robert Krten, PARSE Software Devices +1 613 599 8316.
Realtime Systems Architecture, Books, Video-based and Instructor-led
Training, Consulting and Software Products at www.parse.com.

Big difference: “rw” is supposed to be a BENCHMARK utility, not a file
manipulation utility. There is a reasonable expectation that a file
manipulation utility will do something destructive if you do something
stuipd.

Mathew Kirsch <mkirsch@ocdus.jnj.com> wrote:

Big difference: “rw” is supposed to be a BENCHMARK utility, not a file
manipulation utility. There is a reasonable expectation that a file
manipulation utility will do something destructive if you do something
stuipd.

The benchmark is doing a series of file operations like a file manipulation
util. (it just measures how efficiently it can do them) If you tell it
to do the operations on the raw disk (removing the intermediate layers is
a valid thing to try and do) then it will happily do so if you have the appropriate
permissions to operate on the raw disk.

Sorry that you lost your data, but I hope that you did get some of the answers
that you were looking for.

There was a use message with the util:

peterm-1$ use rw
rw - benchmark file sequential read/write throughput (jgarvey)

rw [options]* [file]

Options:
-b Set the stdio buffer size to the record size (default is BUFSIZ (1k))
-d Use DCMD_FSYS_DIRECT_IO (user dma) file access
-f Do not fsync() the file between write and read phases
-p Pregrow the file prior to write benchmarking (default to append)
-P Pregrow the file (no 0-fill) prior to write benchmarking
-r Record size for each I/O operation (default 8k)
-s Use stdio/FILE* file access (default is file descriptor)
-S Specify O_SYNC semantics (default is to use mount default)
-t Total size of the file (default 256M)

Peter

Peter Martin wrote:

The benchmark is doing a series of file operations like a file manipulation
util. (it just measures how efficiently it can do them)

Exactly. The purpose of a BENCHMARK utility is to measure how
efficiently it can perform file operations, not to destroy data. File
operations should be restricted such that they are non-invasive.

Any benchmark I’ve ever dealt with in the past has always asked for a
DEVICE as a command line argument. Where the use message simply says
[file], I took that to mean a file in /dev, as in a hard disk device
node. This utility was also sold to me as a DISK benchmark, when it’s
actually a FILESYSTEM benchmark.

Sorry that you lost your data, but I hope that you did get some of the answers
that you were looking for.

Nope. Write performance is horrible, only 4MB/s. The SCSI chain is
properly configured, properly cabled, and properly terminated, using
good quality cables and terminator. The four drives are the only devices
on the chain.

There was a use message with the util:

I know. However, the usage message doesn’t tell you how to properly use
the utility… “rw” by itself didn’t seem to do anything, and the [file]
command line argument is ambiguous. I wanted to test a disk, therefore I
reasoned that it took a disk device node as an argument.