(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No.4, 2010
but could out of system earlier than Re time if all of workshave been done on computing nodes.The system model in this paper is considered as a cluster of nodes that connected by a single and shared media backbone, similar to a LAN network. A cluster consist of one coordinator node and n agent nodes A1, A2, ..,An. thecoordinator node receives requests, reservations, and possibly plans to schedule request on agent nodes by itsscheduler module. In a different way, each agent node alsohas two major parts: local scheduler and processor frequencycontroller. The coordinator's scheduler dispatches schedulingtimetables and requests that should be ran on node, to agentschedulers. According received timetables local
scheduler give control of processing unit to request, the reservation.Figure 1 shows structure of cluster of nodes with a master or coordinator for managing several agent nodes that allconnected to single backbone.
According to this model of computation, there are tworesource, computing resource and network resource. Basedon these two types of resources, there are conflicts onaccessing and utilization them. First conflict appears whenany two or more request want exclusively access thenetwork media for communicating and deploying workloadto destination node. Only one of them could access thenetwork and transfer its data to destination node. Another resource is computing power of the nodes. When a requestwants completely access to the node, intended for uses it for processing purposes in some time interval, other requestscould not access it until end of processing time of currentrequest on it.
Figure 1. Topology of cluster of nodes with a coordinator and many agentnodes.
Relation between processor speed and thermal behavior of any chip can be approximated by the followingequation:
C RT(t)C (t)
Where T(t) is temperature at time
is speed of processor at time
. the parameters
are the thermalresistance and capacitance of chips, respectively (with fan or any peripheral attached to chip, like heat sink).The parameter
relate power consumption of processor to its speed. The
parameter has a value of roughly 3.0 [1,3]. For safety of system, processor temperature should not reach to critical point of temperature,due to damaging effects on chip operation.According to thermal model in the (1), we can derivefollowing (2) for calculating temperature at any point of time[1,3]:
t F F E
)eT (T +T =T
Where in general
is steady state temperature atoverclocking speed of
is temperature at between times with speed of
after elapsed t unit of time,and
is the temperature at lowest level at the start time.Parameter
is equal to
and t is elapsed time of time thattemperature was
.By this equation, we can calculate the t value:
)T T T T (
H E H o
To avoiding complex and time consuming computationsat run time on scheduler, we utilize simple and effectivestrict overclocking schema. Consequently, in this schema,we exploited three phases in support of CPU frequencyscaling, under-clocked phase, normal clocked phase andoverclocked phase. In under-clocking phase (i.e. idle mode)frequency of processor is reduced to minimum availablevalue which results in reduced temperature to near theminimum possible value. In the over-clocking phasetransiently frequency of processor is increased to maximumvalue until temperature reach to normal point. Finally in thenormal-clocking phase frequency backs to nominal it tocontinue probably reminded workload of request.Considering the temperature is not above normal, reliabilityand continuity of computing operations are preserved. Alsowe cover two working modes in the schema, normal loadmode and idle load mode. To reducing temperature morequickly in idle mode we never deploy any workload to the processor that keeps temperature and frequency in lowestlimit, i.e. under-clocking phase. We exploit this situation dueto expanding succeeding overclocking interval to themaximum possible value. Using the (3) we can calculate tand ratio of under-clocking to over-clocking periods.
In this section we introduce a scheduling algorithm thatuses described strict overclocking schema in situationswhere conflicts are appeared between current reservationrequest and previous guaranteed and scheduled requests,reservations parts, is discovered.As previously described, for overclocking any time period of the processors, we elaborate the three step strictoverclocking schema: in first step, node processor get under-clocking frequency with idle workload, in the second, thenode get overclocking frequency, and last, the node getnormal clocking frequency. Only the timeslots of processor