Presentation is loading. Please wait.

Presentation is loading. Please wait.

Author: Laurence Field (CERN)

Similar presentations


Presentation on theme: "Author: Laurence Field (CERN)"— Presentation transcript:

1 Author: Laurence Field (CERN)
Monitoring Overview Author: Laurence Field (CERN)

2 Introduction There are many Monitoring Systems around
Most are end to end They do not interoperate This was highlighted in the LCG/OSG interoperations activity Focus here more on LCG but applicable for other grids What do we need to have seamless monitoring across grids? Install one solution everywhere? Which one? Lets take a look at them all and look for similarities This may lead us to a generic solution What are we trying to achieve?

3 VO/Grid Specific Middleware
Hourglass Model VO/Grid Specific Middleware Security File Transfer Job Submission Service Discovery Monitoring Site Specific Systems

4 Terminology Information, Monitoring and Accounting.
All essentially the same Different properties Information (Service Discovery) Usually static Latest query (pull model) No history Monitoring Usually dynamic Continuous query (push model) Historically information available (history query) Accounting History No data loss It’s all Data! “Data >> Information >> Knowledge” K. Peach CHEP 2004.

5 Globus MDS GIIS Site GIIS Cache Node GRIS Provider

6 GridIce GridIce Site GRIS Provider DB Node Lemon Lemon Server Sensor

7 R-GMA R-GMA Site Producer Servlet Cache Node R-GMA API Sensor

8 Apel R-GMA Site Producer Servlet Cron Node DB Script

9 Grid Cat Grid Cat GridCat Site GRAM SQL Scripts Cron

10 MonaLisa MonaLisa Site MonaLisa Service DB Node API Module

11 Abstract View All systems have a similar architecture Sensors
Obtain the raw data Process this ready to be inserted into the system Site Level Cache/Database Site Level Transport Moves data from the node with the sensor to the site cache External transport. Whatever happens once the data leaves the site System wide schema Central repository for visualisation

12 Interoperability Matrix
Sensors Local Transport Site Cache External Schema Repository MDS Information Providers LDAP Memory LDAP DB R-GMA Various HTTP SQL MySQL Apel Custom Scripts GridICE Lemon Lemmon Server Postgresql Grid Cat Cron SQL Lite GRAM GridCat MonaLisa Monitoring Module ??? SQL DBs Java Objects

13 General System SQL DB at site stores all the data
Need a common schema to define what data should be there Common interface to extract information GSI enabled queries for pull Register and notification system for push Sensors Must produce data in agreement with schema Common sensors would reduce effort and errors But common sensors are not compulsory Site level transport Completely hidden Sites own system could be used Could be easily interfaced to common sensors

14 External Query Interface
General System External Query API External Query Interface Site Boundary SQL Node Insert Interface Insert API Sensor

15 Current LCG System R-GMA Grid ICE MON GIN R-GMA Producer MySQL
GridICE GIIS APEL Cron Lemon Server SE CE Lemon Lemon GIN GIN GIP GIP APEL Script GRIS GridFTP Mon BDII GRIS BDII

16 Possible LCG System R-GMA Grid ICE BDII MON BDII R-GMA Producer MySQL
Lemon Server SE CE Lemon Lemon R-GMA API R-GMA API Sensor Sensor

17 Possible LCG System MON Interface SQL XXX Server SE CE XXX API XXX API
Sensor Sensor

18 Summary Monitoring interoperability between grids is difficult
Even within a single grid there is a problem! Don’t need another system but a consolidation of current systems Common answer, just deploy one implementation Is this the right answer, if so which one do we choose? Most systems seem to be similar Data from a sensor is stored in a sites cache (SQL DB) Focus on the cache and the metrics we need Sensors follow the schema Generic sensors would save effort. Agree on the interface. Both for pull, a simple query. And push, register and notify Leverage experience and effort from all systems Leave freedom to chose implementations


Download ppt "Author: Laurence Field (CERN)"

Similar presentations


Ads by Google