MDS/NWS API Usage PDF
Abstract
This
document describes Java APIs for MDS and NWS and example
usage.
Introduction
This package provides Java APIs for Globus MDS and NWS service. The
package can be used to accessdifferent attributes associated with grid
resources especially compute resources. The package is developed under
the assumption that MDS2, MDS4 and/or NWS service is properly
configured and running, and someparameters are supplied or available
under default configuration. COG INSTALL PATH/lib/*.jar must be in the
class path along with JavaNWS.jar and gridarm common.jar.
Important Classes
Following classes are important for discovering resources and resource
attributes.
ResourceDiscoverer
This is an abstract class. It provides following methods to access
resource information by using a driver forGIS.
- public public ResourceSpecificationSet discover():
discovers all registered resources along with all foundattributes.
- public ResourceSpecificationSet discover(String name):
discovers attributes of a single resource givenas 'name'.
- public ResourceSpecificationSet discover(String[]
attributes) Discovers all resources and returns onlygiven attributes.
- public abstract ResourceSpecificationSet
discover(ResourceRequest rq) this is abstract method andmust be
implemented by the driver class.
- public static ResourceDiscoverer
getResourceDiscoverer(DiscovererContext context): This method isused to
load appropriate driver for a GIS based on the name provided in the
context.
DiscovererContext
Used to load a particular ResourceDiscoverer. Here are useful methods
to set configuration parameters forresource discoverer.
- public void setName(String name): (required) sets name of
the information service to be used forresource discovering. e.g. mds2,
NWS, mds4
- public void setPort(int port): (optional)i sets port for
information service. e.g 2135 in case mds2
- public void setBaseName(String baseName): (optional) used
to set base dn for Ldap in case of mds2 service.
- public void setAddress(String address): sets host address.
e.g. the name/address of MDS index server, or address NWS NS
host.
- public String getSchemaType(): (optional) sets schema type
possible values in case of MDS2 are glue or default. It is not used for
NWS.
- public void setQueryType(String queryType): sets query
type. eg RSL, XPATH or Basic. Right now only basic is used.
ResourceRequest
This class is used to manipulate a resource query. Constraints can be
set using a setRelation or setter methods for attributes. e.g.
request.setRelation(Constants.CPU AVAULABLE, "4", Relation.MIN) is
equivalent to the RSL relation (cpuAvailable >= 4) Or
request.setRelation(Constants.HOST NAME, "xyz", Relation.EXACT) is
equivalent to request.setRelation(Constants.HOST_NAME, "xyz") or
request.setHostName("xyz") and RSL (HostName=xyz). Use HostRequest
class for compute resource hosts. See class methods and Constants for
possible available relations.
ResourceSpecification
The discover method of resource discoverer returns resource speci
cation set which can be iterated to access/ traverse specifications
returned for all or a set of resources. See class ResourceSpeci cation,
HostSpecification
Grid Information Services
MDS2
Mds2 service can be used for both default or glue schema. The possible
attributes which can be retrieved are listed in Constants class, or
given in HostSpeci cation class as setter/getter methods. The following
parameters should be supplied with DiscovererContext or con gured as
default.
- HOST_NAME: Name or address of MDS index server.
- PORT: port number of index service
- BASE_DN: Base domain name
- SCHEMA_TYPE: Possible values are default | glue. Defaults
to default
Default
configuration The default con guration can be retrieved by using
DiscovererContext.defaultContext() static method. For HOST, PORT and
BASE DN, the method tries to retrieve values from grid-info.conf fon
guration le, which is present in the GLOBUS LOCATION/etc
directory. For it to work, you should pass GLOBUS LOCATION as system
property using -D ag at java command line. The property name
should be org.globus.wsrf.container.webroot as it is being used in GT4
(WSRF) implementation. If you cannot pass this option as command line,
then you can create a symbolic link to GLOBUS LOCATION as 'globus' in
your home directory. If this is also not possible then simply copy
grid-info.conf le in /.globus/etc.
Usage
For example-usage please see the class org.askalon.grms.tools.AskGIS
DiscovererContext ctx = DiscovererContext.defaultContext();
if (dname != null) {
ctx.setName(dname);
}
if (schema != null) {
ctx.setSchemaType(schema);
}
if (host != null) {
ctx.setAddress(host);
}
if (bdn != null) {
ctx.setBaseName(bdn);
}
ResourceDiscoverer rd = ResourceDiscoverer.getResourceDiscoverer(ctx);
ResourceSpecificationSet rsset = rd.discover(request);
Iterator itr = rss.iterator();
while (itr.hasNext()) {
HostSpecification host = (HostSpecification)itr.next();
StringBuffer sb = new StringBuffer();
sb.append("\nHost Specification:\n-------------------\n Contact:");
sb.append("\n\tName: ").append(host.getName());
sb.append("\n\tType: ").append(host.getType());
sb.append("\n\tLRM Type: ").append(host.getLRMType());
sb.append("\n\tAccess Point: ").append(host.getAccessPoint());
sb.append("\n OS: ");
sb.append("\n\tName: ").append(host.getOsName());
sb.append("\n\tVersion: ").append(host.getOsVersion());
sb.append("\n\tRelease: ").append(host.getOsRelease());
sb.append("\n CPU:");
sb.append("\n\tCount: ").append(host.getCpuCount());
sb.append("\n\tSpeed: ").append(host.getCpuSpeed());
sb.append("\n\tAvailable: ").append(host.getCpuAvailable());
sb.append("\n\tCurrent: ").append(host.getCpuCurrent());
sb.append("\n Platform:");
sb.append("\n\tName: ").append(host.getPlatform());
sb.append("\n\tArchitecture: ").append(host.getArchitechture());
sb.append("\n Filesystem:");
sb.append("\n\tCount: ").append(host.getFSCount());
sb.append("\n\tTotal MB: ").append(host.getFSTotal());
sb.append("\n\tFree MB: ").append(host.getFSFree());
Enumeration e = host.deviceNames();
while (e.hasMoreElements()) {
String dname = (String)e.nextElement();
sb.append("\n Device:").append(dname);
sb.append(host.getDevice(dname));
}
System.out.println(sb.toString());
}
NWS
Currently nws service can be used to access cpuAvailable, cpuCurrent of
registered hosts. The information can be queried for all registered
hosts or for a subset of registered host. If the con gured properly
bandwidth and latency values can be retrieved between a registered host
and other target system. Inorder to work, the following parameters
should be supplied or con gured as default;
- NWS_NAME_SERVER (Required) The host name or IP address of
the NWS name server is required. It can be supplied during
initialization of nws driver for resource discoverer as described in
example usage.
- NWS_MEMORY_HOST (Optional) This is optional, parameter. It
represents the nws memory server's name or address.
Default con
guration TBD
usage
The following example describe a simple usage for nws service. It takes
"olperer.dos.uibk.ac.at" as nws nameserver and nds NWS driver by
using ResourceDiscoverer.gerResourceDiscoverer(ctx) method.
iscovererContext ctx = new DiscovererContext();
ctx.setAddress("olperer.dos.uibk.ac.at"); // NWS NS
ResourceDiscoverer nws = ResourceDiscoverer.gerResourceDiscoverer(ctx);
ResourceRequest request = new ResourceRequest();
request.setCpuAvailable(1.5, Relation.MIN);
ResourceSpecificationSet rss = nws.discover(request);
Iterator itr = rss.iterator();
while (itr.hasNex()) {
HostSpecification hs = (HostSpecification)itr.next();
System.out.println(hs.getCpuAvailable());
System.out.println(hs.getCpuCurrent());
}
Back to main Usage
|
