||SSS v0.1 Token
||A credential also known as 'project ID'. Multiple users may be
associated a single account ID and each user may have access to multiple
accounts. (See Credential )
|ACL (Access Control List)
||Within the scheduling world, an access control list is used and applied
much as it is elsewhere. An ACL defines what credentials are required
to access or utilizes particular objects. The principle objects to
which ACL's are applied are reservations and
ACL's may contain both allow and deny statements, may include wildcards,
and may contain rules based on multiple object types.
||Reservation META1 contains 4 access statements.
Allow jobs owned by user john or bob
Allow jobs with QOS premium
Deny jobs in class debug
Allow jobs with a duration of less than 1 hour
|AS (Access Statement)
one of ALLOW, DENY
OBJECT one of USER, GROUP,
<AS TYPE="ALLOW" OBJECT="USER" NAME="john"/>
<AS TYPE="ALLOW" OBJECT="USER" NAME="bob"/>
<AS TYPE="ALLOW" OBJECT="QOS" NAME="premium"/>
<AS TYPE="DENY" OBJECT="CLASS" NAME="debug"/>
<AS TYPE="ALLOW" OBJECT="DURATION" NAME="1:00:00"/>
(used in reservation and job creation and queries)
||A logical, scalar unit assigned to users on a credential basis, providing
access to a particular quantity of compute resources. Allocations
are consumed by jobs associated with those credentials.
(used in ???)
||(see Queue ) A class is a logical container
object which holds jobs allowing a site to associate various constraints
and defaults to these jobs. Class access can also be tied to individual
nodes defining whether or not a particular node will accept a job associated
with a given class. Class based access to a node is denied unless
explicitly allowed via resource manager configuration. Within Maui,
classes are tied to jobs as a credential.
||job cw.073 is submitted to class batch
node cl02 accepts jobs in class batch
reservation weekend allows access to jobs in class batch
(used in ???)
||A single processing unit. A CPU is a consumable resource.
Nodes typically consist of one or more CPU's. (same as processor
||An attribute associated with jobs and other objects
which determines object identity. In the case of schedulers and resource
credential based policies and limits are often established. At
submit time, jobs are associated with a number of credentials such as user
, group , account , QOS
, and class . These job credentials subject
the job to various polices and grant it various types of access.
In most cases, credentials set both the privileges of the job and the
ID of the actual job executable
|Job cw.24001 possesses the following credentials:
||A quantity of local disk available for use by batch jobs. Disk
is a consumable resource .
||A description of the environment in which the executable is launched.
This environment may include attributes such as the following:
command line args
local user id
local group id
process resource limits
|Job cw.24001 possesses the following execution environment:
||A credential typically directly mapping to
a user's UNIX group ID.
||The fundamental object of resource consumption. A job contains
the following components:
A list of required consumable resources
A list of resource constraints controlling
which resources may be allocated to the job
A list of job constraints controlling
where, when, and how the job should be run
A list of credentials
An execution environment
||A set of conditions which must be fulfilled in order for the job to
start. These conditions are far reaching and may include one or more
of the following:
When the job may run (i.e., after time X, within Y minutes, etc.)
Which resources may be allocated (i.e., Node must possess at least
512MB of RAM, run only in partition or Partition C, or run an HostA and
Starting job relative to an particular event (i.e., start after job
X successfully completes)
|RELEASETIME>='Fri Jun 06, 10:00AM'
||A quantity of physical memory (RAM). Memory is provided by compute
nodes. It is required as a constraint or consumed as a consumable
resource by jobs. Within Maui, memory is tracked and reported in
||Node node001 provides the following resources
Job cw.24004 consumes the following resources per task
||A node is the fundamental object associated with compute resources.
Each node contains the following components
A list of consumable resources
A list of node attributes
||A node attribute is a non-quantitative aspect of a node. Attributes
typically describe the node itself or possibly aspects of various node
resources such as processors or memory. While it is probably not
optimal to aggregate node and resource attributes together in this manner,
it is common practice. Common node attributes include processor
architecture, operating system, or processor speed. Jobs often specify
that resources be allocated from nodes possessing certain node attributes.
||A node feature is a node attribute
which is typically specified locally via some form of configuration file.
Node features are opaque strings associated with the node by the resource
manager which generally only have meaning to the end user or possibly to
the scheduler. Commonly, a node feature will be associated with a
subset of nodes allowing end users to request use of this subset by requiring
that resources be allocated from nodes with this feature present.
In many cases, node features are used to extend the information provided
by the resource manager.
(This may be used to indicate that the node possesses a 950 MHz Pentium
III processor and that the node is owned by the Geology dept)
||A processing unit. A processor is a consumable resource.
Nodes typically consist of one or more processors. (same as CPU)
|Quality of Service (QOS)
||An object which provides special services, resources, etc.
||(see Class )
||An object which reserves a specific collection or resources for a specific
timeframe for use by jobs which meet specific conditions
||Reserve 24 processors and 8 GB of memory from time T1 to time T2 for
use by user X or jobs in the class batch
<RESERVATION STARTTIME="98066503"><AS TYPE="ALLOW" OBJECT="QOS"
NAME="premium"/><RES TYPE=PROC COUNT="4" <CONSTRAINT ATTR="PROCSPEED"
(use: reservation creation, modification, destruction, and reporting)
||All consumable resources are
tracked and managed in a number of ways. A compute node's available
resources is calculated as its configured
resources minus the sum of the resources actually utilized by all job
tasks running on the node.
||Node cl003 is configured with 4 processors and 512 MB of memory.
This node is executing 2 tasks of job clserver.0041 which is utilizing
1 processor and 60 MB of memory each. Additionally, it is also running
1 task of job clserver.0047 which is using 1 processor and 250 MB of memory.
Node cl003's available resources are thus
processors = 4 - (2 * 1 + 1 * 1) = 1
memory = 512 - (2 * 60 + 1 * 250) = 142 MB
||Any object which can be utilized (i.e., consumed and thus made unavailable
to another job) by, or dedicated to a job is considered to be a resource.
Common examples of resources are a node's physical memory or local disk.
As these resources may given to one job and thus unavailable to another,
they are considered to be consumable. Other aspects of a node, such
as its operating system, are not considered to be consumable since its
use by one job does not preclude its use by another.
Note that some node objects, such as a network adapter, may be dedicated
under some operating systems and resource managers and not under others.
On systems where the network adapter cannot be dedicated and the network
usage per job cannot be specified or tracked, network adapters are
not considered to be resources, but rather attributes.
Nodes possess a specific quantity of consumable resources such as real
memory, local disk, or processors. In a resource management system,
the node manager may choose to report only those configured resources available
to batch jobs. For example, a node may possess a 80 GB hard drive
but may have only 20 GB dedicated to batch jobs. Consequently, the
resource manager may report that the node has 20 GB of local disk available
when idle. Jobs may explicitly request a certain quantity of consumable
||A resource constraint imposes a rule on which resources can be be used
to match a resource request. Resource constraints either specify
a required quantity and type of resource or a required node attribute.
All resource constraints must be met by any given node in order for a match
to be established.
||A job may request that a block of resources be dedicated while the
job is executing. In this case, the scheduler is responsible to guaranteeing
that these resources, whether utilized by the job or not, are set aside,
unavailable to other jobs.
||A quantity of virtual memory available for use by batch jobs.
Swap is a consumable resource provided by nodes and consumed by jobs
||An atomic collection of consumable resources.
||The user credential used to provide access to functions and resources.
In local scheduling, global user IDs map directly to local user IDs.
||The user credential under which the job executable will be launched.