Components of LLVIEW

LLview is a client-server based application. The client LLview is the graphical user interface which display the utilization of a batch system controlled cluster. For LoadLeveler the server part of llview uses the C-API of Loadleveler to extract all relevant data of running and waiting jobs, their sizes and the utilization of the SMP-nodes.


Components and data flow of LLview for IBM Loadleveler

Client llview

The main part of the LLview client is the node display. It shows a small box for each processor of an SMP node. The box color codes the job running on this processor. Furthermore, the node display contains additional elements displaying global information about the node like status, memory usage and cpu load for each SMP node. When moving the mouse pointer over a processor box, the corresponding information is highlighted in the other display elements of LLview. These elements are job list, usage bar, information panel, and a utilization chart. See also: -> Using the LLview client

Server llqxml and other batch system adaptors

For Loadleveler the server part of LLview is a small C-program which uses the data access C-API of LoadLeveler to get the information about the node usage, running and waiting jobs directly. The information is stored in XML format. This program must be executed on a node of the SMP cluster where LoadLeveler is available.
The are two modes in which llqxml can access the data from LoadLeveler. The first is to request data from the LoadLeveler master daemon for whole cluster. The second mode is to request the data from the scheduler daemons of LoadLeveler which are running on each compute nodes. The first mode is faster but gives not all data about the jobs (for example no information about used CPU time). The second mode is slower because llqxml have to connect to each scheduler daemon of the whole machine.

For other batch systems LLview has perl-scripts which extracts the data from the user level query commands of the batch system. For Blue Gene/L there is a perl script which uses the database interface of the system database of Blue Gene/L.

Data Flow

There are three different modes in which LLview can access the data. In every case the data will be generated by the server program llqxml. The result is a XML data file with a simple structure (only a few different XML-tags).
  1. The client part of LLview can access the data directly if the client runs on the same machine or the LLview is configured to use a ssh-connection to the LoadLeveler machine. In this mode LLview executes the llqxml command at every update step.
  2. The usual way is to distribute the data by a WWW server to support clients running on local desktops. In this case, LLview accesses the data from the web server with a user/password authentication method. The perl script getllqxml.pl can be used as a crontab script for regular update of the XML file on the web server. This script is available in the util directory of the LLview distribution.
  3. Additionally, LLview provides a mechanism to record data and replay recent usage statistics. Therefore LLview is able to read a tar file which contains the XML files. LLview can also read flat XML files from a directory. Such tar files can be recorded by a separate perl script getwwwdata.pl (in the util directory) or by LLview itself when using the first access mode (direct access).
The different modes for data access can be selected in the Option panel of LLview (see Managing Configuration Options). The currently used data access mode will be displayed in the status bar of LLview (see Using the LLview client).
last change 04.05.2005 | LLview | Print