top - download
⟦64e234dea⟧ Wang Wps File
Length: 12234 (0x2fca)
Types: Wang Wps File
Notes: Spelunked
Names: »~ORPHAN74.00«
Derivation
└─⟦17da89677⟧ Bits:30006229 8" Wang WCS floppy, CR 0126A
└─ ⟦this⟧ »~ORPHAN74.00«
WangText
T̲A̲B̲L̲E̲ ̲O̲F̲ ̲C̲O̲N̲T̲E̲N̲T̲S̲
4.7 SYSTEM PERFORMANCE ......................
4.7.1 Message Traffic Flow .................
4.7.2 Throughput ............................
4.7.3 Storage ................................
4.7.4 Timing .................................
…86…1 …02… …02… …02… …02…
4.7 S̲Y̲S̲T̲E̲M̲ ̲P̲E̲R̲F̲O̲R̲M̲A̲N̲C̲E̲ ̲
This section presents the evaluations of the porposed
MPF system in terms of performance. The system is
fully compliant with the IFB requirements (sectin 5.1.4)
for
- throughput
- storage
- timing
with the message traffic characteristics presented
in the IFB(section 5.1.4.1) as underlaying assumption.
Detailed information on the requirements has been given
in section 1.2 and will be repeted again here in each
subsection together with the proof of compliance.
System characteristics and performance calculations
will be presented to the extent that sufficient evidence
is available for judgement of performance compliance.
In orderto perform these calculations it has been necessary
in some cases to introduce assumptions concerning message
characteristics and the message flow: Such assumptions
are merely used in order to provide a rough estimate
for proof of system performane compliance; deviations
from these additional assumptions, which are of conservative
nature, are not expected to degrade the performance
appreciably and never to the extent of non-compliance.
Whenever additional assumptions have been made this
fat is made clear in the text.
4.7.1 M̲e̲s̲s̲a̲g̲e̲ ̲T̲r̲a̲f̲f̲i̲c̲ ̲F̲l̲o̲w̲ ̲
Refer to fig. 4.7.1-1.
The message traffic flow has been derived as follows:
- number of incoming messages equal to number of outgoing
is a direct consequence of the figures from th IFB
reflecting the fact that the MPF is a relay facility.
- 50% ACP 127 formatted incoming messages which are
broadcasted, and 50% ACP126 formatted incoming
messages which are transmitted on TARE and TRC
is a correct case assumption givig the maximum
processing load for message analysis and conversion.
- 10% of incoming messages need service (garbled
etc.) is a Christian Rovsing assumption. 3 of
4 supervisory positions are service positions.
- 20% of incoming message for loca delivery is a
conservative Christian Rovsing and supported by
the fact that 10% of all incoming messages are
service messages out of which a certain part are
directed to the supervisors attention.
- 10% of outgoing messages need routing assignmet
is a Christian Rovsing assumption.…86…1 …02…
…02… …02… …02…
Fig. 4.7.1-1…86…1 …02… …02… …02… …02…
- 10% of outgoing messages are locally prepared is
a Christian Rovsing assumption which is supported
by the fact that 10% of all messages are service
messages out of which a certan part is prepared
at supervisory positions. This is also the reasonable
the amount of messages which may be prepared considering
that message reception will take place simultaneously.
- external channel capabilities are taken from the
IFB.
- haracter flows for the internal terminals are calculated
from the above flow figures assuming maximum external
character input rate and the following assumptions
plus derivations from the IFB:
- messages are presented in full with annotation
formessage service, and returned in full (1500
char.)
- a guiding format is presented to the user for
message preparation and full message is returned
(400 char for supervisory positions, 1500 char.
for the MCSF).
- delivery of messages as receied:
400 char. for supervisory positions and 1500
char for the MCSF.
- The outgoing character rate for external channels
has an upper limit given by the capacity; the ratio
between the actual number of outgoing messages
and the possible number f outgoing messages is
the maximum possible average number of multiple
transmissions per outgoing message. This number
is found as the ratio between the incoming character
capacity and the outgoing character capacity of
external channels for the pak traffic flow and
is 240/200 = 1.2.
- input character rate on the MC channel (from standby
system) is low and disregarded in these calculations.
- output rate to the MC channel (to standby system)
is assumed to be of the same size as the total
outgoing character rate for other external channels
(which reflects the message reation rate for the
system), i.e. 200 char./sec. minus 10% due to CTS
& ATOMAL messages which are deleted = 180 char/sec.
Fig. 4.7.1-1 presents the peak traffaic flow in terms
of characters; the equivalent number of messages is
simply derived frm the size of the average message.
1500 x 0.84 + 400 x 0.1 + 7000 + 0.04 = 1610 char.
where the distribution of operational, service, and
data messages corresponds to the figures from the IFB.
The p̲e̲a̲k̲ ̲n̲u̲m̲b̲e̲r̲ ̲o̲f̲ ̲i̲n̲c̲o̲m̲i̲n̲g̲ ̲a̲n̲d̲ ̲h̲e̲n̲c̲e̲ ̲o̲u̲t̲g̲o̲i̲n̲g̲ ̲m̲e̲s̲a̲g̲e̲s̲
̲f̲o̲r̲ ̲t̲h̲e̲ ̲p̲r̲e̲s̲e̲n̲t̲ ̲c̲o̲n̲f̲i̲g̲u̲r̲a̲t̲i̲o̲n̲ is found as
…01…0.12 msg/sec. = 432 msg/hour-
According to the IFB, section 5.1.3(01) a 25% increase
in local (internal) and external communication lines
(terminals) combined with a 30% increase of message
traffc shall be possible without hardware or software
changes. A 30% increase will result in a p̲e̲a̲k̲ ̲n̲u̲m̲b̲e̲r̲
̲o̲f̲ ̲i̲n̲c̲o̲m̲i̲n̲g̲ ̲a̲n̲d̲ ̲h̲e̲n̲c̲e̲ ̲o̲u̲t̲g̲o̲i̲n̲g̲ ̲m̲e̲s̲s̲a̲g̲e̲s̲ ̲f̲o̲r̲ ̲t̲h̲e̲ ̲e̲x̲p̲a̲n̲d̲e̲d̲
̲c̲o̲n̲f̲i̲g̲u̲r̲a̲t̲i̲o̲n̲ of
0.16 msg/sec. = 562 msg/hour
Taking into account the maximum amountof multiple transmissions
(factor 1.2) the number of outgoing messages for the
expanded configuration is 674 msg/hour. It is seen
that the busy hour rate of 500 messages specified in
the IFB is included in these figures.
A peak (character) throuhput rate of
600 char/sec incoming, and
800 char/sec. outgoing
has been specified in the IFB, section 5.1.4.2(g).
The sytem character throughput is discussed in section
4.7.2 and it is shown how the system also will accept
this peak rate whih is well above the maximum possible
rate in a system with a 30% traffic expansion:
242 x 1.3 = 315 char/sec incoming, and
478 x 1.3 = 631 char/sec outgoing
Refer to fig. 4.7.1-1.
4.7.2 T̲h̲r̲o̲u̲g̲h̲p̲u̲t̲
The expanded configuration flow, i.e. the configuration
corresponding to 30% traffic expansion and with continous
maximum character transfer rates is used for troughput
calculation. The corresponding traffic flow is explained
in section 4.7.1.
With the addition of one function-retrieval for rerun
of outgoing messages - the functions of the flow have
been analysed and the resource consumption in terms
o PU processing time and number of disk reads and writes
has been calculated.
The result is presented in the tables 4.7.2-1 to 9,
the last table containing the summation.
The following assumptions have been made in addition
to section 4.7.1:
-there is one edit session per prepared message
- retrieval is based on request for rerun and corresponds
to 5% of the transmitted messages (including the
effect of multiple transmissions). Worst case
retrieval (DTG and TO PLA) is assumed.
- 10 additional resource needs have been added to
message analysis and conversion to cover reprocessing
after message service (worst case corresponding
to 10% of messages going to message service)
The PU time needs correspond to instruction times witout
CACHE HITS and is thus a conservative estimate. Since
the direct CPU time for an application process is known
to be small from experience with appliations of this
kind a fairly accurate estimate is possible, based
on number of system calls (I/ access, disk access,
table access etc.)
The disk access time is given as
t…0f…access…0e… = t…0f…HANDLER…0e… + t…0f…channel…0e…+ t…0f…rot…0e…
+ t…0f…SEEK…0e… + t…0f…TRANSFER
t…0f…HANDLER…0e… = Disk handler CPU time = 6 ms.
t…0f…CHANNEL…0e… = transfer time to/from disk controller
of data and instructions, time is negligible
t…0f…ROT…0e… = rotational delay of disk = 8.3 ms. average
t…0f…SEEK…0e… = time for locating the proper rack; the
seek time is caluclated in fig. 4.7.2-1.
As a worst case data have been assumed
located randomly on the whole disk; on
this point improvements may be gained:
Data which are often required, such
as tables, may be constrained to smaller
aeas. t…0f…SEEK…0e… = 34.7 ms.
…0f…TRANSFER…0e… = time for transfer of data to/from disk
from/to the controller; time is negligible
Consequently
t…0f…acess …0e… = 49 ms.
The access time ths calculated is the time for executing
a READ to any of the two mirrored disks. since in
average any of the two disks may used half this time
shall be used for load calculations.
Disk WRITE accesses are executed to both disks and
are in both cass followed by a READ for check. The
READ after WRITE will experience the average rotational
delay plus some validation time in the PU: A total
of 10 ms is added per WRITE access to each disk: hence
the resulting access time for disk writes to beused
for load calculations is
t…0f…WRITE…0e… = 59 ms.
In order to include the effect to the system load of
the support functions - log, statistics, storage catalog
creation, status preparation, report generation, security
functions -experience factors shall be used which have
been taken from a similar project (CAMPS). The factors
are
PU 1.26
DISK READS 1.11
DISK WRITES 1.43
The final result thus obtained is then according to
table 4.7.2-9.
PU 63%
CPU load (2 CPUs) 31.5%
DISK load 53%
The throghput is thus secured in the absolute worst
case traffic load situation.
Finally, we will consider the peak character rate specified
in the IFB to last for up to 5 min. in any one hour:
600 char./sec. incom,ing traffic
800 char./sec. outgoingtraffic
Comparing these character rates to the expanded configuration
rates used for the throughput calculations above, the
following factors of increase are found
incoming traffic: 600/315 = 1.91
outgoing traffic: 800/631 = 1.29
These chracter rates are only obtained with the introduction
of considerably more external channels and local terminals
than the number corresponding to the expanded configuration
.
The loads corresponding to the high character rates
are
Each CPU 42%
Disk 75%
It has here been assumed thatall S/W modules, except
the reception which works at the higher input rate,
work at the outgoing traffic rate.
Concerning the connectivity of internal terminals and
external channels a discussion is presented in section
1.2.1.…86…1 …02… …02…
…02… …02…
M̲S̲G̲,̲ ̲R̲E̲C̲E̲P̲T̲I̲O̲N̲
ITEM PU(ms) NUMBER
OF
DISK
R W
Read from Channel 60
Create item 16
Prelim. Analysis 30
Write to disk 140 4
Trans.Acknowledgement 25
Checkpoint to dis 3̲5̲ 1̲
306 5
Unload to off line disk
̲6̲7̲
̲
̲2̲
̲
̲3̲
̲
373 2 8
MSG. TRANSMISSION (1.2 transmissions per msg.)
NUMBER OF DISK
I̲T̲E̲M̲ P̲U̲(̲m̲s̲)̲ R W
Read item 144 4.8
Table access for serial
no. plus direct CPU time
36
Transmission to channel
72
Receive acknowledgement
25
Checkpoint to disk
42 1.2
̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲ ̲
3̲1̲9̲ 4̲.̲8̲ 1̲.̲2̲ ̲ ̲
T̲R̲A̲N̲S̲M̲I̲S̲S̲I̲O̲N̲ ̲T̲O̲ ̲S̲T̲A̲N̲D̲B̲Y̲ ̲S̲Y̲S̲T̲E̲M̲ ̲(̲V̲I̲A̲ ̲M̲C̲)̲
Approximately as for MSG TRANSMISSION divided by the
transmission mu
