Computer Science Technical Report Archive

USC Computer Science Technical Reports, no. 619 (1995)

(USC DC Other)

USC Computer Science Technical Reports, no. 619 (1995)

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content An Ev aluation of A TM cards for HP W orkstations
Shahram Ghandeharizadeh and Dongho Kim
Departmen t of Computer Science
Univ ersit y of Southern California
Los Angeles California August In tro duction
W e conducted exp erimen ts to measure the p erformance of A TM adaptors on HP Series w ork
stations Our target en vironmen t consisted of t woHP w orkstations named Bogota and
T aip ei and a CISCO Hyp erSwitc hA A TM switc h Eac hHPw orkstations consists of a Gb yte
in ternal disk a MHz P ARISC CPU an A TM card and an Ethernet card While Bogota is
congured with Mb ytes of memoryT aip ei has Mb ytes of DRAM Both mac hines participate
in a lo cal area net w ork consisting of ten HP Series w orkstations a Sun SparcStation a PC
equipp ed with P en tium MHz CPU and t w o prin ters
W e measured the p erformance of our A TM cards using TCP UDP and a reliable transmission
proto col using UDP implemen ted b y the Nugget op erating system In order to observ e the impact of
the A TM switc h on the observ ed transfer rates w e also conducted exp erimen ts using congurations
consisting of t wow orkstations connected p oin ttop oin t using A TM cards alone The maxim um
observ ed transfer rate w as marginally higher than megabits p er second Mbs Our A TM man ual
adv ertises a Mbs transfer rate for the cards There could b e sev eral explanations for whyw e did
not observ e the adv ertised transfer rate W ebeliev e the primary factor for this is the latency incurred
b y the system to execute the soft w are that implemen ts the in terface b et w een A TM and TCPUDP T o test this h yp othesis and in order to ensure that neither the CPU pro cessing capabilit y nor the
system bus are the b ottlenec k resources w e conducted exp erimen ts in a lo opbac k mo de In these
exp erimen ts b oth the sender and receiv er execute on one mac hine The maxim um transfer rate
observ ed b y the receiv er w as Mbs Ho w ev er the maxim um transfer rate of a w orkstation is Mbs b ecause b oth the sender and receiv er are executing on one CPU

ATM
Loopback on Bogota
Loopback on Bogota
Nugget Between Bogota and Taipei
Loopback on Bogota
Ethernet
TCP
Between Bogota and Taipei
UDP Between Bogota and Taipei
Loopback
TCP
UDP
Nugget
TCP Between Bogota and Taipei
TCP Between Bogota and Taipei
UDP Between Bogota and Taipei
UDP Between Bogota and Taipei
UDP Between Bogota and Taipei
TCP Between Bogota and Taipeiw/ load
w/o load
w/ switch
w/ load
w/o load
w/ switch
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 10
Figure 8
Figure 9
Figure 11
Figure 12
Figure 13
Figure Summary of P erformed Exp erimen ts
F or comparison purp oses w e conducted exp erimen ts using Ethernet In these exp erimen ts the
participating mac hines w ere temp orarily separated from the rest of the net w ork to eliminate com
p eting trac
Mac hine Name CPU T yp e Memory In ternal Storage
Bogota HP P ARISC MHz MB GB
T aip ei HP P ARISC MHz MB GB
T able HP w orkstations used
Exp erimen ts
V ariables
In eac h proto col a p acket is a unit of transmission Unix op erating system reserv es sp ecic amoun tof
memory space termed buer in order to send and receiv e pac k ets using a so c k et In our exp erimen ts
w e manipulated b oth the buer size and the pac k et size of a so c k et The analyzed pac k et sizes w ere
and Kb ytes The a v ailable buer size cannot b e smaller than the pac k et size The lo w er

limit on the buer size is the same as pac k et size The upp er b ound on the buer size as w ell as the
pac k et size is b ytes When w e tried to emplo y a buer size larger than b ytes using
setsockopt function
the system fails to reserv e desired buer space
W e analyzed buer sizes that are and times larger than the pac k et sizes In addition
wev aried the n um b er of concurren t streams in order to determine the scalabilityof eac h proto col
The n um b er of streams analyzed w ere and streams
TCP and UDP
The rate at whic h the sender can transmit data and the rate of data arriv al observ ed b y the receiv er
is not the same for UDP These t w o data rates are iden tical for TCP This is b ecause with UDP the
reliabilit y of a pac k et loss is in excess of W e measured transfer rate as the rate of data arriv al
observ ed b y the receiving Unix pro cess W e start the timer once the receiv er has receiv ed its rst
pac k et
The pseudoco des of the algorithms used for the exp erimen ts can b e found in app endix A
System load and A TM switc h
Exp erimen ts for A TM with TCP and UDP proto cols w ere conducted in v arious situations in order
to determine the b ottlenec k resource and inuence of eac h resource
In Figures and w e measured the transfer rates with mo derate system loads whereas the
transfer rates in Figures and w ere measured with virtually no system loads
T aip ei and Bogota
w ere connected directly except in Figures and In these t w o exp erimen ts w e connected t w o
mac hines via CISCO A TM switc h
Results
The results are sho wn in Figures Except for Figures and that resp ect on the p erformance
of Nugget op erating system eac h Figure has six graphs The rst v e graphs denote the transfer rate
with alternativ e buer sizes In these gures w ev aried the pac k et size from and Kb ytes
A HPUX function used for setting the buer size of so c k et Refer app endix
Systems w ere running essen tial system pro cesses only
xaxis while observing the system transfer rate yaxis The buer size is a m ultiple of pac k et
size v aried from to times larger than the pac k et size The last graph fo cuses on the
exp erimen t where the buer size is t wice larger than the pac k et size Eac h curv e denotes the factor
of impro v emen t relativ e to the a v erage transfer rate of eigh t concurren t streams The curv e denoting
eigh t streams is the base line b ecause relativ e to itself it results in a factor of one impro v emen t
In general for a xed buer size the system pro vides a higher bandwidth with larger pac k et sizes
F or example with a Kb yte buer size a single stream observ es a Mbs transfer rate with a Kb yte pac k et see Figure b as compared to Mbs with a Kb yte pac k et In sev eral cases w e
observ ed that Kb yte pac k et sizes pro vide a higher bandwidth see Figure a and b This migh t
b e b ecause of the underlying proto col that transmits Kb yte c h unks If b oth the application and
the underlying proto col use the same pac k et size the p erformance w ould b e enhanced b ecause the
system do es not need to comp ose or decomp ose the pac k ets as the alternativenet w ork la y ers are
transfered
W e also observ ed p erformance impro v emen t with larger buer sizes when the pac k et sizes w ere
unc hanged see Figures and This is b ecause the underlying proto col could pro cess the pac k ets
in a more ecien t manner with larger buer space
The p erformance scales do wn almost linearly as w e increase the n um b er of sim ultaneous streams
W e observ ed sev eral discrepancies and sp eculate that they could b e eliminated if the exp erimentw ere
conducted using mac hines in a singleuser mo de The scalabilit y graphs fo cused on the scenario where
the buer size is t wice the pac k et size b ecause these exp erimen ts sho w ed the b est p erformance
in general they con tained the en tire range of pac k et sizes
The p erformance of A TM is exp ected to b e higher with UDP as compared to TCP This is b ecause
UDP is a simple and unreliable transmission proto col as opp osed to TCPHo w ev er the p erformance
measurementw as almost the same due to follo wing reasons First receiv er of UDP exp ects the same
n um b er of pac k ets as TCP Second the transfer rate w as measured at the receiv ers side Th us ev en
if the transfer rate for sender in UDP w ould b e a lot faster than TCP the transfer rate for receiv er
in UDP w as around the same as that of TCP b ecause of pac k et losses
W e observ ed v ariable transfer rates with the UDP proto col when the system load v aries compare
Figures and Ev en though the systems w ere idle Figure b sho ws a lo w er p erformance when
compared with Figure b with KByte pac k et size and one stream This is b ecause w e cannot
determine when UDP proto col loses pac k ets
T o observ e the impact of A TM switc h w e conducted the exp erimen ts with TCP and UDP

proto cols when t wow orkstations w ere connected via A TM switchasw ell as when they w ere connected
using a p oin ttop oin t conguration see Figures through W e observ ed the same p erformance
for b oth congurations using TCP see Figures and With UDP the switc hpro vided a sup erior
p erformance Figures and W e sp eculate that this is b ecause the switc h reduces the n um ber of
lost pac k ets using its in ternal memory Ev en though the A TM man ual claims that a Mbs data transfer rate the maxim um observ ed
transfer rate w as sligh tly higher than Mbs In order to determine if the CPU is the b ottlenec k
resource w e also p erformed an exp erimen t that transmits data to the sending mac hine itself using
the comm unication proto col As sho wn in Figure w e observ ed a Mbs transfer rate The CPU
can supp ort up to Mbs transfer rate b ecause b oth the sender and receiv er executing on one
CPU
Both TCP and UDP w ere designed indep endentof A TM and their curren t implemen tation is
supp orted for bac kw ard compatibilit y with the programming in terface W e sp eculate that the im
plemen tation of the in terface b et w een the underlying A TM proto col and TCPUDP is inecien t
resulting in a lo w p erformance This h yp othesis is bac k ed b y the result of exp erimen ts conducted
using Ethernet Figures and The transfer rates observ ed with Ethernet w ere v ery close
to its sp ecication Mbs Moreo v er it scales almost linearly as a function of the n um ber of
sim ultaneous streams
Conclusion
This pap er presen ts p erformance of A TM and Ethernet using TCPUDP proto cols and an imple
men tation using Nugget In general larger pac k et and buer sizes result in higher transfer rates with
A TM The same is not true with Ethernet b ecause it cannot pro vide a transfer rate higher than Mbs

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for A TM using TCP with load

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for A TM using TCP without load

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for A TM using TCP
via CISCO Hyp erSwitc h A A TM switch

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for A TM using UDP with load

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for A TM using UDP without load

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for A TM using UDP
via CISCO Hyp erSwitc h A A TM switch

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for Lo opbackusing TCPonBogota

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for Lo opbac k using UDP on Bogota

5 10 15 20 25 30
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
Figure T ransfer rate of A TM using Nugget Buer Size Kb ytes
5 10 15 20 25 30
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
Figure T ransfer rates of Lo opbac k using Nugget Buer Size Kb ytes

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for Ethernet card using TCP

2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
a Buer Size P ac k et Size b Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
c Buer Size x P ac k et Size d Buer Size x P ac k et Size
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
20
40
60
80
100
Mb/s
One Stream
Two Streams
Four Streams
EightStreams
2 4 6 8 10 12 14 16
Packet Size (KBytes)
0
2
4
6
8
10
Times
One Stream
Two Streams
Four Streams
EightStreams
e Buer Size x P ac k et Size f F actor of Impro v eme n ts relativeto Eigh t Streams
Buer Size x P ac k et Size
Figure T ransfer rates and F actor of Impro v emen ts for Ethernet card using UDP

A Pseudo Co des
F ollo wings are the pseudoco des of the algorithms used for the exp erimen ts The v alue of LOOPS
used w as A Pseudoco de for TCP
sender
socket
bind
setsockopt set the buffer size of socket
connect
forLOOPS
sendPACKETSIZE
shutdown
close
receiver
socket
bind
setsockopt set the buffer size of socket
connect
forLOOPS
start timer after receiving the first packet
recvPACKETSIZE
stop the timer
shutdown
close
A Pseudoco de for UDP
sender
socket
bind
setsockopt set the buffer size of socket
forLOOPS
sendtoPACKETSIZE
receiver
socket
bind
setsockopt set the buffer size of socket
forLOOPS
start timer after receiving the first packet
recvfromPACKETSIZE
stop the timer

A Pseudoco de for Nugget
send
initnetwork
InitDisk
receivewindowid getwinid
destinationportid getportid
forLOOPS
getnetpacket
sendmsgPACKET
recv
initnetwork
InitDisk
forLOOPS
start timer after receiving the first packet
recvmsgPACKET
relenetpacket
stop the timer

Linked assets

Computer Science Technical Report Archive

Conceptually similar

PDF

USC Computer Science Technical Reports, no. 634 (1996)

PDF

USC Computer Science Technical Reports, no. 615 (1995)

PDF

USC Computer Science Technical Reports, no. 625 (1996)

PDF

USC Computer Science Technical Reports, no. 865 (2005)

PDF

USC Computer Science Technical Reports, no. 650 (1997)

PDF

USC Computer Science Technical Reports, no. 623 (1995)

PDF

USC Computer Science Technical Reports, no. 610 (1995)

PDF

USC Computer Science Technical Reports, no. 602 (1995)

PDF

USC Computer Science Technical Reports, no. 612 (1995)

PDF

USC Computer Science Technical Reports, no. 766 (2002)

PDF

USC Computer Science Technical Reports, no. 659 (1997)

PDF

USC Computer Science Technical Reports, no. 627 (1996)

PDF

USC Computer Science Technical Reports, no. 618 (1995)

PDF

USC Computer Science Technical Reports, no. 622 (1995)

PDF

USC Computer Science Technical Reports, no. 600 (1995)

PDF

USC Computer Science Technical Reports, no. 666 (1998)

PDF

USC Computer Science Technical Reports, no. 601 (1995)

PDF

USC Computer Science Technical Reports, no. 587 (1994)

PDF

USC Computer Science Technical Reports, no. 578 (1994)

PDF

USC Computer Science Technical Reports, no. 685 (1998)

Description

Shahram Ghandeharizadeh and Dongho Kim. "An evaluation of ATM cards for HP 9000/735 workstations." Computer Science Technical Reports (Los Angeles, California, USA: University of Southern California. Department of Computer Science) no. 619 (1995).

Asset Metadata

Creator Ghandeharizadeh, Shahram (author), Kim, Dongho (author)

Core Title USC Computer Science Technical Reports, no. 619 (1995)

Alternative Title An evaluation of ATM cards for HP 9000/735 workstations (title)

Publisher Department of Computer Science,USC Viterbi School of Engineering, University of Southern California, 3650 McClintock Avenue, Los Angeles, California, 90089, USA (publisher)

Tag OAI-PMH Harvest

Format 18 pages (extent), technical reports (aat)

Language English

Unique identifier UC16269780

Identifier 95-619 An Evaluation of ATM cards for HP 9000735 Workstations (filename)

Legacy Identifier usc-cstr-95-619

Format 18 pages (extent),technical reports (aat)

Rights Department of Computer Science (University of Southern California) and the author(s).

Internet Media Type application/pdf

Source 20180426-rozan-cstechreports-shoaf (batch), Computer Science Technical Report Archive (collection), University of Southern California. Department of Computer Science. Technical Reports (series)

Access Conditions The author(s) retain rights to their work according to U.S. copyright law. Electronic access is being provided by the USC Libraries, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name USC Viterbi School of Engineering Department of Computer Science

Repository Location Department of Computer Science. USC Viterbi School of Engineering. Los Angeles\, CA\, 90089

Repository Email csdept@usc.edu

Inherited Values

Title Computer Science Technical Report Archive

Description Archive of computer science technical reports published by the USC Department of Computer Science from 1991 - 2017.

Coverage Temporal 1991/2017

Repository Email csdept@usc.edu

Repository Name USC Viterbi School of Engineering Department of Computer Science

Repository Location Department of Computer Science. USC Viterbi School of Engineering. Los Angeles\, CA\, 90089

Publisher Department of Computer Science,USC Viterbi School of Engineering, University of Southern California, 3650 McClintock Avenue, Los Angeles, California, 90089, USA (publisher)