Embed Size (px)
Citation preview
Enabling Grids for E-sciencE
www.eu-egee.org
EGEE-II INFSO-RI-031688
A New Milestone for Grid Interoperation: SRM-SRB Interface Development
WeiLong, UengAcademia SINICA Grid [email protected]
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
2
Outline
• Overview• Architecture• Progress• Future Plan• Conclusions
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
3
Overview
SRM
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
Why SRM?
• SRM is an unique interface for accessing diffident backend storages for diffident middleware.
• Easy to develop applications to adapt different backend storages.
• Provide space and file management on the storage system.
• SRM is the web service interface and the implementation usually depends on the backend storage technology.
4
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
5
Why SRM?
• Storage Elements (SE) can use different type of technologies– CASTOR, dCache, DPM, BeStMan,...,etc.– DRM (Disk Resource Manager)/TRM (Tape Resource
Manager) /HRM (Hierarchical Resource Manager)
• Grid middleware needs to access files with an uniform interface– Manage storage resources– Not a file transfer protocol
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
6
SRM features
• Provides space management• Provides an uniform access interface• Manages DRM/Tape/HRM• Does not transfer files itself.• Manage the life time of file
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
7
What is SRB ?
• Developed by San Diego Supercomputer Center• A distributed file management system (Data Grid), based on a
client-server architecture.• Allows users to access files seamlessly across a distributed
environment, based upon their attributes rather than just their names or physical locations.
• It replicates, syncs, archives, and connects heterogeneous resources in a logical and abstracted manner.
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
8
SRB Physical Structure
Oracle Client SRB ServerOracle RDBMS
SRB Vault@ location B
SRB Vault@ location B
SRB Vault@ location D
SRB Server
Storage Space
Storage Driver
SRB Server
Storage Space
Storage Driver
SRB Server
Storage Space
Storage Driver
User@
location X
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
10
SRM-SRB Architecture Overview
Core
SRB+DSI
Auxiliary Filecatalog
Gridftp/management API
SRM API
File transfer (gridftp)
Web Service
Data server management
Users/applications
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
7
Information in Auxiliary Filecatalog
• AMGA server, it stores partial filecatalog, resource and SRB host information…– Users Information– Resources Information– Files Information– Space Metadata– Resource States– …
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
13
Architecture Overview (cont.)
Web Service
Core
Data server management
Non MES+DSI
MES Server(GSI enabled)
SRB storage space
Non MESNon MES
Auxiliary filecatalog(AMGA)
Put a file (SURL)
SURL->PathDefault space is available? If yes, some files need to be uploaded
Non MES(+DSI)
Non MES(+DSI)
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
14
Architecture Overview (cont.)
Web Service
Core
Data server management
Non MES+DSI
MES Server(GSI enabled)
SRB storage space
Non MES(+DSI)
Non MES(+DSI)
Auxiliary Filecatalog(AMGA)
Return TURL
Transfer to TURLYes, default space is ready and this file has been set to pinned.
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
15
Architecture Overview (cont.)
Web Service
Core
Data server management
Storage space
MES Server(+GSI enabled)
Storage space
Non MES(+DSI)
Non MES(+DSI)
Auxiliary Filecatalog(AMGA)
Upl
oad
a fil
e(gr
idftp
)
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
16
Architecture Overview (cont.)
Web Service
Core
Data server management
Storage space
MES Server(GSI enabled)
Storage space
Non MES(+DSI)Non MES
(+DSI)
Auxiliary Filecatalog(AMGA)
put file done
SURL -> Path
Set this file has been uploaded and unpinned
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
17
Architecture Overview (cont.)
Web Service
Core
Data server management
Storage space
MES Server(GSI enabled)
Storage space
Non MES(+DSI)Non MES
(+DSI)
Auxiliary filecatalog(AMGA)
The Status of PutDone
Return the Status of PutDone
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
18
Support Flexible File/Space Types
• SRM system has a caching mechanism and has to take care of SRM issues like file lifetime, space management,…,etc.– Volatile space– Durable space– Permanent space
• In our implementation– Use AMGA as auxiliary catalog and record all space
usage, space type, and some file metadata inside.
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
19
Checking Disk Status
• How to get the disk usage of the space?– Need to know the free and used space on SRB server– SRB does not provide the mechanism to monitor resource
usage– We need to know the usage
Space management
• In our implementation– InfoServer:
Deployed on non-MCAT enabled SRB server– SRBInfoServer:
Deployed on MCAT-enabled SRB server
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
20
Checking Disk Status(Cont.)
Web Service
Core
Data server management
Storage space
MES Server(GSI enabled)
Storage space
Non MES(+DSI)Non MES
(+DSI)
• Auxiliary Filecatalog• (AMGA)
SRBInfoServer
InfoServerInfoServer
Update all resource information
Query SRBInfoServer
Query InfoServerQuery InfoServer
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
21
Checking Disk Status(Cont.)
Web Service
Core
Data server management
Storage space
MES Server(GSI enabled)
Storage space
Non MES(+DSI)Non MES
(+DSI)
• Auxiliary Filecatalog• (AMGA)
SRBInfoServer
InfoServerInfoServer
Update status of each resource
Resource info
Logical Resource info
Resource infoResource info
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
Progress
• Space Management Functions– srmReserveSpace– srmReleaseSpace– srmUpdateSpace– srmGetSpaceMetaData– srmChangeSpaceForFiles– srmGetSpaceTokens
• Permission Functions– srmSetPermission– srmCheckPermission– srmGetPermission
• Directory Functions. – srmMkdir– srmRmdir– srmRm– srmLs– srmMv
• Data Transfer Functions– srmPrepareToGet– srmBringOnline– srmPrepareToPut– srmCopy– srmStatusOfCopyRequest– srmReleaseFiles– srmPutDone– srmAbortRequest – srmSuspendRequest– srmResumeRequest – srmGetRequestSummary – srmGetRequestTokens
• Discovery Functions– srmGetTransferProtocols – srmPing
23
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
24
Future Plan
• Internal space management functions– Use a thread to recycle expired space
• More async operation– Space functions
srmReserveSpace, srmStatusOfReserveSpaceRequest srmUpdateSpace, srmStatusOfUpdateSpaceRequest srmChangeSpaceForFiles, srmStatusOfChangeSpaceForFilesRequest
– Transfer functions srmBringOnline, srmStatusOfBringOnlineRequest srmPrepareToGet, srmStatusOfGetRequest srmPrepareToPut, srmStatusOfPutRequest srmAbortRequest, srmAbortFiles,srmResumeRequest, srmSuspendRequest,
srmGetRequestSummary, srmGetRequestToken• Implement IRODS-DSI• Transfer to SRM-IRODS Interface
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
Conclusions
• ASGC (Academia Sinica Grid Computing) launches the SRM-SRB Interface Software Version 1.0
• Implement the SRM functions for SRB based on SRM v2.2
• Built on a GridFTP server, an AMGA Server developed for SRB, to make an SRB look like a Classic Storage Element.
• Allowed the gLite tools to transfer files between SRMs and SRBs
• Next, the interoperation between SRB and DPM, dCache, Castor, etc. would be realized.
25
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
Further Information
• For more information on SRM-SRB Interface, please see – http://www2.twgrid.org/SRM-SRB/– email: [email protected].
• SRM Working Group: http://sdm.lbl.gov/srm-wg/• SRB: http://srb.diceresearch.org• EGEE: http://www.eu-egee.org/• ASGC: http://www.twgrid.org
25
EGEE-II INFSO-RI-031688
Enabling Grids for E-sciencE
26
Thank you
