Storage accounting

The following page contains the results of the Storage Accounting activity in progress at INFN-CNAF. This activity started at the beginning of March 2011 and is being carried out by Andrea Cristofori, Enrico Fattibene and Paolo Veronesi.

Usage Record used

The Usage Record (UR) utilized is generated in a way that allows a per file accounting. The fields that are part of the UR used are described on the paragraph "Architecture".

Architecture

The Accounting System utilized to collect, send, store and publish data is DGAS. Data are collected at a site level, sent and stored to the designed HLR for that particular site (site or multi site HLR). sysDefStorageAccounting is the table that contains the data sent with the UR on the HLR. The first six columns describe the table attributes. The remaining columns contain the following fields: GlueSchemaField correspond to the Glue Schema parameter used; Description describe the meaning of the field; AlwaysNULL tells if the field, in the current implementation, is expected to be always NULL.

Field Type Null Key Default Extra   GlueSchemaField Description AlwaysNULL
ID bigint(20) MUL NULL auto_increment        
RecordIdentity char(64)   PRI         Hash of site, VO, SA, and TimeInstant values  
GlobalFileId char(64) YES   NULL       Unique file identifier (LFC file ID) yes
LocalFileId char(64) YES   NULL       Local path of file yes
GlobalGroup char(64) YES   NULL       Grid site name no
GlobalUsername char(64) YES   NULL       Unique identifier of the user yes
LocalUserId char(64) YES   NULL       Local user ID yes
Charge int(10) YES   NULL       The charge associated with the storage utilization yes
Status char(64)             Status of a request to the storage system yes
Host char(64)           GlueChunkKey : GlueSEUniqueID Storage Element host no
SubmitHost char(64) YES   NULL       The file source host yes
ProjectName char(64) YES   NULL     GlueSAAccessControlBaseRule VO name no
ProjectPartition char(64) YES   NULL     GlueSALocalID Storage Area name no
StorageType char(64) YES   NULL       StoRM, dCache, DPM, etc.  
ProtocolType char(64) YES   NULL       The protocol used in the event yes
Network int(10) YES   NULL       The network utilization for the operation yes
Disk int(10) YES   NULL     GlueSAUsedOnlineSize Used space no
TimeDuration int(10) YES   NULL       Time between 2 measures no
TimeInstant int(10) YES   NULL       Time of the measure no
ServiceLevel char(64) YES   NULL       Persistent, volatile, etc.  

Implementation

URs are, in the current implementation, generated by retrieving accounting data from the Information System. A dedicated script is run once a day from a UI and collects data for all the Italian sites. The URs are then sent to a test HLR.

Data visualization

HLRmon development server (hlrmon-dev.cnaf.infn.it) collects data from the configured HLR (at the moment the HLR used is the dgas-test-vm01.cnaf.infn.it test instance) typically once a day, aggregates them per day and stores the results in its internal DB. The table ( sedata_new) that contains these data is described in the left half of the following table. The remaining columns correspond to: the corresponding field in the HLR table (if applicable) and the description of the field.

Table sedata_new schema

Field Type Null Key Default Extra   HLRTableField Description
ID bigint(20) NO PRI NULL auto_increment      
YYMMGG date NO   0000-00-00     TimeInstant Measurements day
Site varchar(45) YES   NULL     GlobalGroup Grid site name
SEName varchar(45) YES   NULL     Host Storage Element
VOName varchar(45) YES   NULL     ProjectName VO
UsedSpace bigint(20) YES   NULL     Disk Used space
Class varchar(45) YES   NULL     ProjectPartition Storage Area name
last_mod_time timestamp NO   CURRENT_TIMESTAMP        

HLRmon does not store only information related to URs but it also stores, in another table, data related to the total and free space of each Storage Area (SA), the total space and the used space for each Storage Element (SE) taken directly from the Information System. This table ( storage_info_system) is described in the left half of the following table. The remaining columns correspond to: the corresponding field in the Glue Schema (if applicable) and the description of the field.

Table storage_info_system schema

Field Type Null Key Default Extra   GlueSchemaField Description
ID bigint(20) NO PRI NULL auto_increment      
YYMMGG date NO   0000-00-00       Measurements day
Site varchar(45) YES   NULL       Grid site name
SEName varchar(45) YES   NULL     GlueSEUniqueID Storage Element
SETotalSpace varchar(45) YES   NULL     GlueSETotalOnlineSize SE total space
SEUsedSpace bigint(20) YES   NULL     GlueSEUsedOnlineSize SE used space
SATotalSpace bigint(20) YES   NULL     GlueSATotalOnlineSize SA total space
SAFreeSpace bigint(20) YES   NULL     GlueSAFreeOnlineSize SA free space
Class varchar(45) YES   NULL       Storage Area name
last_mod_time timestamp NO   CURRENT_TIMESTAMP        

HLRmon shows data aggregated per site, Storage Element and Storage Area. For each of these aggregation keys, and for the interval of time desired, a set of charts with the temporal trend is produced. Charts are available at:
https://hlrmon-dev.cnaf.infn.it:8443/hlrmon/report/storage.php

The same information can also be accessed in a tabular form, with the possibility to save it in xls format:
https://hlrmon-dev.cnaf.infn.it:8443/hlrmon/report/table_storage.php

A per-VO aggregation is not possible at the moment because a Storage Area can be shared among different VOs. In this case the Glue Schema allows the publication of the space used in the Storage Area, but not of the space used by each VO. For this case is not possible to know the space used by each VO.

-- Main.Enrico Fattibene - 2011-07-07

Edit | Attach | PDF | History: r11 < r10 < r9 < r8 < r7 | Backlinks | Raw View | More topic actions
Topic revision: r11 - 2011-07-15 - EnricoFattibene
 
This site is powered by the TWiki collaboration platformCopyright © 2008-2024 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback