Farm batch system and Fermi inter-process communication and synchronization toolkit Page: 4 of 13
This article is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to UNT Digital Library by the UNT Libraries Government Documents Department.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
"Fgr 2:~ FBS..P Desig
FI .e andw wi i
" FARM isaFSdeoIhtrn nea wke ne.. cree ironetfrue
pr c ss s st rt them as+o requste byJ ,rp rst ersau o M a d U ,n tfe M w e
U ie ar an ts a
e Lggr r ogdamo i rspnsbl fr eevn n rr strn erradeetlgifra
- m ti rtl' I -i
Figure 2: FBS Design
n Job Manager (JM) is an FBS process that controls single section running on the farm. LSF
starts JMs as LSF batch processes. JM is responsible for allocating resources on farm nodes
with FLIMD, and communicating with FARMD(s) on nodes allocated for the section by
FLIMD to start user processes, and wait for their completion.
T FARMD is a FBS daemon that runs on each worker node. It creates environment for user
processes, starts them as requested by JM, reports their status to JM and UI, notifies JM when
user process exits.
e Historian is FBS historical database manager. It receives section start/exit statistics and stores
it on disk. UI provides a tool for reading this database and generating reports as requested
by user.
" Logger or log daemon is responsible for receiving and storing error and event log informa-
tion sent by other FBS components. This information is primarily used for FBS debugging
and trouble shooting.
2.2.4 Robustness and Reliability
FBS design makes it highly reliable and robust with respect to failure of individual components.
This is achieved by distributing run-time information among different FBS components and avoid-
ing redundancy of the information. Basic idea is that FLIMD, as the most critical component, can
recover after failure based on information received from JMs. JM is highly reliable component and
most likely reason for its failure is failure of the node where it runs. Since in typical configuration
all JMs and LSF run on the same node, failure of the node inevitably means failure of LSF and the
whole farm, and necessarily leads to re-initialization of the batch system. Unlikely failure of an in-rB& "npQnPmc
Upcoming Pages
Here’s what’s next.
Search Inside
This article can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Article.
Mandrichenko, I.V. Farm batch system and Fermi inter-process communication and synchronization toolkit, article, February 20, 2001; Batavia, Illinois. (https://digital.library.unt.edu/ark:/67531/metadc718290/m1/4/: accessed April 24, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; crediting UNT Libraries Government Documents Department.