Sei sulla pagina 1di 7

Job Schedule Model Based Introduction

on Grid Environment In the mid 1990s the term “Grid” was coined
to denote a proposed distributed computing
infrastructure for advanced science and
engineering . In grid environment, users may
S.Saranya, access the computationalresources at many
B.Tech IT, III year, sites . The functions of information systems
Sri Sai Ram Engineering College, based on grid computing architectures are
E-mail:ssaranyainfotech@gmail.com resources (e.g., CPUs, memory, storages,
etc.) sharing, collaborative processing,
reliable and secure connection, etc.
G.Sangeetha, However, each resource of coordinate node
B.Tech IT, III year, in the grid environment, (e.g., CPU loading,
Sri Sai Ram Engineering College, memory usage, etc.) changes dynamically.
E-mail:sangeinfotech@gmail.com Therefore, how to optimizresources usages
is an important issue. Foster et al. presented
Grid Resource Allocation and Management
Abstract: (GRAM). GRAM simplifies the usage of
remote systems by providing a single
Grid computing architecture was defined to standard interface for requesting and using
be a complete physical layer. The functions remote system resources for the execution of
of information systems based on grid "jobs". The most common usage of
architecture are resources sharing, GRAM is a remote job submission and
collaborative processing, etc. System control. Lee et al. presented an optimal
performance is calculated from resource analyzing resources model that can receive
usages of all grid nodes. If we have load- the information about CPU usage,
balancing in all grid nodes, we will have number of running jobs of each grid
good performance in the system. All the grid resource node to achieve load balancing and
nodes are divided to supervisor, supervisor make the plans and allocations of the
backup and execute grid nodes. Each grid resources of collaborated nodes
node has computing cost level to charge optimize. Lee et al. proposed a process
each running job. Transfer jobs from high schedule analyzing model to do jobs transfer
CPU utilization grid nodes to low utilization when the load of node is high. Iosup et al.
grid nodes through supervisor scheduling, proposed synthetic workloads to evaluate
the system performance will be better. In performance on grid nodes. In the
this paper, we propose a job schedule model scheduling criteria, Silberschatz et
(JSM) base on grid environment. By this al.proposed that it includes CPU utilization,
model, we can make all grid nodes be throughput, turnaround time, waiting time
loading balance. When the load of the node and response time. From the above
is high, it sends the request to the supervisor statements, we know that if we want
and it selects the most suitable node to to have good performance in the system
execute. Via implementing this model, we based on grid environment, we should have
can improve the performance of grid load-balancing of all grid nodes in the
environment efficiently. system. When a new job is entered and the
load of this node is high, we should schedule (1)Supervisor job schedule module
some jobs to the other nodes to process. (SJSM):
Through the supervisor ‘s evolution, we can
find the most suitable grid node to do job (a) Collects the latest computing cost level
transfer. In this paper, we propose a job of all grid nodes.
schedule model (JSM) base on grid (b) Collects period the current state
environment. By this model, we can information of all grid nodes.
supervise the states (e.g., CPU usage, (c) Process jobs
memory usage, etc.) of grid resource nodes, (d) Process job transfer request from execute
and analyze thisinformation to find the most grid nodes.
suitable node to do job transfer. Through (e) Evaluates the most suitable transfer node
this model, we can achieve the best when receives job transfer request from
resources allocation, and enhance the overall high-loading node.
grid computing performance. (f) Transfers jobs to other execute grid node
when the load of supervisor grid node is
high.
II. Framework of the Proposed Model (g) Assign the most suitable transfer node
when receive job transfer request.
We divide all grid nodes to supervisor grid
node, supervisor backup grid node and
execute grid nodes. To have load-balancing (2) Execute job schedule module (EJSM):
for all grid nodes, we propose a job schedule
model (JSM) based on grid environment. (a) Sends grid node information to the
There are two modules in JSM, which are supervisor when modification.
supervisor job schedule module (SJSM) (b) Processes jobs.
built on the supervisor grid node, and (c) When loading is high, request supervisor
execute job schedule module (EJSM) to do job transfer.
built on the execute grid node and (d) Requests to do job transfer to the
supervisor backup grid node as shown in selected grid node through supervisor’s
Fig. 1. evaluation.
(e) Receives the completion of job transfer
from the selected grid node.
(f) Receives job transfer from other grid
node.
(g) Returns the completion of job to the
request grid node.
(h) Sends the current state information to
supervisor grid node.

II.1. Supervisor Job Schedule Module


(SJSM)

In supervisor grid node, we install


The functions of these two modules are as supervisor job schedule module (SJSM).
follows: There are three sub-modules
in SJSM, which are supervisor receive sub- grid node.
module (SRSM), supervisor evaluation sub- (3) The load of supervisor is too high.
module (SESM) and supervisor assign sub-
module (SASM),as shown in figure. (4) Supervisor assign sub-module
(SASM)

The functions of this sub-module are as


followings:
(a) Find the most suitable grid node.
(b) Assign this selected grid node to do job
transfer request.

II.1.1 Supervisor Receive Sub-Module


(SRSM)
There are four components in SRSM, which
are,

supervisor collect resource component


(SCRC),
supervisor collect current state component
(SCCSC),
supervisor request transfer component
(SRTC) and
supervisor receive job component (SRJC) as
(1) Supervisor receive sub- shown in figure 3.
module(SRSM)

The functions of this sub-module are as


followings:
(a) Collects the latest computing cost level
of each execute grid node.
(b) Collects period the current state
information of each execute grid node.
(c) Process new jobs
(d) Process job transfer request from execute
grid nodes.

(2) Super evolution sub-module


(SESM)

The function of this module is as following: The functions of these components are as
(a) Evaluates the suitable transfer grid follows:
nodes.
(1) Receives job transfer request from (1) Supervisor collect resource component
highloading (SCRC):
SCRC collects the information of execute (3) Supervisor request transfer
grid nodes. When information of execute component (SRTC):
grids is inserted, deleted or updated, SCRC SRTC receives job transfer request from all
processes to update execute grid nodes grid nodes. The format of execute job request
(EJR) is as Table 4.
information (EGNI) as Table 1.
(4) Supervisor receive job component
(SRJC):
SRJC receives new job from supervisor.

II.1.2. Supervisor Evolution Sub-Module


(SESM)

SESM processes supervisor evolution


component
(SEC) as Figure 4.

SEC processes the following steps:


(1) From the request cost level of EJR, finds
the same cost level in EGNI and
(2) Supervisor collect current state (2) From the request memory size of EJR,
component find the available memory ((1-Memor Usage
(SCCSC): % of EGNSI)*Memory Size of EGNI) of
above found execute nodes is enough and
SCCSC collects current state information of create available transfer grid nodes (ATGN)
all execute grid nodes and update the as Table 5.
execute grid nodes state information
(EGNSI) as Table 3.

II.1.3. Supervisor Assign Sub-Module


(SASM)
SASM has two components of supervisor
job
scheduling component (SJSC) and evolution sub-module (EESM), execute
supervisor job transfer component (SJTC) process submodule (EPSM) and executive
as Figure 5. send message submodule (ESMSM) as
shown in Figure 6.
(1) Supervisor job scheduling component
(SJSC): (1).Execute receive sub-module (ERSM)
SJSC uses CPU usage in ATGN as the key The functions of this sub-module are as
to sort ATGN in ascending order to creates followings:
order of transfer grid nodes (OTGN) as in
the table 6. (a) Processes new jobs.
(b) Receives the completion of job transfer
from the selected grid node.
(c) Receives job transfer from other grid
node.
(d) Receives the message of doing job
transfer through supervisor’s evaluation.

(2) Supervisor job transfer component (SJTC):

SJTC sends the request of job transfer to the


execute node of the first entry in OTGN. If
returned message is yes then transfer job to
this grid node. If returned message is no, we
find the next entry to process. If we can not (3) Execute process sub-module (EPSM)
find in OTGN, we go to SRSM to process EPSM processes jobs.
again.. (4)Execute send message sub-module
(ESMSM)
II.2. Execute Job Schedule Module
(EJSM) ESMSM processes the following functions.
(a) Sends job transfer request message to
The execute grid node installs execute job supervisor.
schedule (b) Sends receive job transfer message to
module (EJSM). EJSM has four sub- supervisor.
modules, which are execute receive sub- (c) Transfers the job to the selected execute
module (ERSM), execute grid node.
(d) Returns the completion of job to the EESM has component of EEC and the
requested grid node. function of this sub-module is as Figure 8.
(e) Sends execute grid node information to EEC checks the loading of this execute grid
the supervisor when modification. node, if loading is high, sends job transfer
(f) Sends the current state information to request message to supervisor and jumps to
supervisor. ESMSM, else jumps to EPSM.

II.2.1 Execute Receive Sub-Module


(ERSM)

ERSM processes the following components


as Figure 7.
(a) Execute process job component
(EPJC): EPJC Processes new job.

(b) Execute receive job completion


component (ERJCC): ERJCC receives the
completion of job transfer from the selected
execute grid node.
II.2.3 Execute Process Sub-Module
(c) Execute receive job transfer (EPSM)
component (ERJTC): ERJTC receives job EPSM has EPC component to process job as
transfer from other execute grid node. Figure 9.

(d) Execute receive do job transfer


component (ERDJTC): ERDJTC receives
the message to do job transfer to the selected
grid node
through supervisor’s evaluation. It transfers
job to the selected execute grid node.

(a) Execute send transfer job component


(ESTJC): ESTJC sends job transfer
request message to supervisor.
(b) Execute receive job transfer component
(ERJTC): ERJTC sends receive job
II.2.2 Execute Evolution Sub-Module (EESM)
transfer message to supervisor. (2) Sends period the current state
(c) Execute job transfer component (EJTC): information.
EJTC transfers the job to the selected (b) Receive job and message.
execute grid node. (1) Receives local new job.
(d) Execute completion job component (2) Receives job transfer from
(ECJC): ECJC returns the completion of request execute grid node.
job to the requested grid node. (c) Evaluates the loading
(e) Execute send grid information (1) If the load is high, sends to do job
component transfer .request to supervisor.
(ESGIC): ESGIC sends execute grid node (d) Sends message
information to the supervisor when .
modification. IV. Conclusion
(f) Execute current state information At present, the load of all grid nodes may be
component different. Some grid nodes may be heavy
(ECSIC): ECSIC sends the current state and some may be idle. If the jobs in high-
information to supervisor. loading grid nodes can do job transfer to
suitable grid nodes to process. Therefore we
III. Process description should have load-balancing of all grid nodes
in the system .In this paper, we propose
We explain supervisor job schedule module supervisor job schedule module in
and execute job schedule module as follows: supervisor grid node and execute job
schedule module in execute grid nodes. By
III.1. Supervisor Process Description these models, we can make all grid nodes be
The processes of supervisor have the loading balance. When the load of the node
following steps: is high, execute grid node can send request
(a) Collects grid data to supervisor, supervisor selects the most
(1) Collects the latest computing cost suitable node to process the job. Via
level of all grid nodes. processing these models, we can improve
(2) Collects period the current state the grid environment performance
information of all grid nodes. efficiently.
(b) Receive job and message.
(1) Receives local new job.
(2) Receives job transfer request
from request execute grid node.
(c) Evaluates the loading
(1) If the load is high, find the
suitable execute grid node to do job
transfer.
(d) Sends message.

III.2. Execute Process Description

(a) Sends grid information.


(1) Sends the latest computing cost
level.

Potrebbero piacerti anche