You are on page 1of 24

Landslide modelling

with LCG tools


Michael Wallace - Research Assistant, Particle Physics,
University of Bristol
Overview

Landslides and landslide simulation

Landslide data management issues

How WLCG tools and resources can help

Overview of proposed architecture


Impact of landslides
Economic impact
!"!#
!"#
#
#!
%&'()* +(,()- ./0,*12
!"#$ &'( $" )*+&#),&(# *+& -""&#
.//0 1 .//2 3) 4,+" (5(+$
31-( 45'66'1& %+78

Developing countries
more vulnerable

Less resources for


disaster recovery

Accumulated risk due


to poorly planned
development schemes
!"!#
!"#
#
#!
%&'()* +(,()- ./0,*12
!"#$ &'( $" )*+&#),&(# *+& -""&#
.//0 1 .//2 3) 4,+" (5(+$
31-( 45 16 789:
Economic impact

Developing countries
more vulnerable

Less resources for


disaster recovery

Accumulated risk due


to poorly planned
development schemes
Disaster planning

Landslide risk of a slope for a given storm

Size and runout of a landslide

Economic impact of a landslide


Things that are useful to know:
courtesy of Dr Liz Holcombe
Landslide modelling

Bristol University has world leading


landslide modellers

chasm -> Landslide risk, hazard, size

questa -> Economic impact

Used by the World Bank for planning


development schemes in developing
countries
Data management

Analysing different scenarios across 100s of


slopes results in many simulations and a
long overall runtime

10s of TB generated per year for a typical


usage scenario

No facilities for managing this data


Current simulation
workow
!" $%&'(
)*'+",-
.'/-0 /12'1//*+
3/,2*(45'&(- +&'/1&/+
$,&(- 2,6/*17/1"
8'+(+"/* 4-(11/*+
9'6'- /12'1//*+
:14%" 0("(
;(-'0("/0 '14%" 0("(
!'7%-(<,1 */+%-"+
9,+"+=>/1/?"+
Current simulation
workow
!" $%&'(
)*'+",-
.'/-0 /12'1//*+
3/,2*(45'&(- +&'/1&/+
$,&(- 2,6/*17/1"
8'+(+"/* 4-(11/*+
9'6'- /12'1//*+
:14%" 0("(
;(-'0("/0 '14%" 0("(
!'7%-(<,1 */+%-"+
9,+"+=>/1/?"+
Improved simulation
workow
!"#$% '"()*+,)+- ./(/% )+'/+))*0 1/0$0-)* 2%$++)*0

345678 9/+$*:
;)"'*$2</# =$-$
6%"2) =)>+/?"+0
6/,@%$?"+ /+2@-0
6/,@%$?"+ "@-2@-0
A"9 ,$+$')*
!B. #",2@?+' '*/=
1CD8.
A"90
E@-2@-0
;)"'*$2</#$%
0#/)+#)0
D)%)$0)
6/,@%$?"+ =$-$ 6/,@%$?"+ =$-$ 6/,@%$?"+ =$-$
6/,@%$?"+
*)F@)0-0
Improved simulation
workow
!"#$% '"()*+,)+- ./(/% )+'/+))*0 1/0$0-)* 2%$++)*0

345678 9/+$*:
;)"'*$2</# =$-$
6%"2) =)>+/?"+0
6/,@%$?"+ /+2@-0
6/,@%$?"+ "@-2@-0
A"9 ,$+$')*
!B. #",2@?+' '*/=
1CD8.
A"90
E@-2@-0
;)"'*$2</#$%
0#/)+#)0
D)%)$0)
6/,@%$?"+ =$-$ 6/,@%$?"+ =$-$ 6/,@%$?"+ =$-$
6/,@%$?"+
*)F@)0-0
D)2%/#$?"+
What is CouchDB?

Top-level Apache project

Document-oriented database

Data stored as JSON documents, indexed


by UID

Views allow indexing by any document eld


Why CouchDB?

Multi-master replication out of the box

Geo-indexing out of the box

Runs on phones, laptops, servers & clusters

Provides application logic and can serve


client-side code

Browser + CouchDB = Web application


Why not CouchDB?

Greater storage requirements than a


typical RDBMS for most data

Uncertainty over write speeds - need to


test how many writes per second we can
scale to
Proposed architecture
8lgCouch
!ob senunel
!ob submluer
8everse proxy
[ob
CllenL Couchu8 lnsLances
CLher. ulrac uummy
Crld SysLem
ouLpuL
Web Cul
CouchApp
Lask/[ob sLaLe
_users
accounung

slmulauon daLa
- lnpuL
- ouLpuL
- mlsc (e.g. lmages)
Lask requesLs
CPASM/CuesLa
- meLadaLa
(releases)
- execuLable
as _auachmenL
SLochasuc emulaLor
1ask senunel
8reak down Lask
requesL lnLo muluple
[ob denluons
L
a
s
k

[
o
b
s

ulspaLches lndlvldual
[obs Lo grld sysLem
1ask Lracker
s
L
a
L
e

u
p
d
a
L
e

Task request

Species software release and input les

Species variable parameters:

Stochastic parameters

Investment scenarios

Rainfall scenarios
Job denition

Species software release and input les

Specic parameters generated by task


sentinel

For stochastic jobs:

Seed for generation of stochastic


parameters
What is a job?

We submit a job wrapper

~400 lines of python

Downloads job denition

Downloads executable and inputs

Runs executable

Parses and uploads output les


Executables

Geographical sciences tag a release in SVN

We automatically check out the release,


build it and commit it to a new releases DB

Regression tests are run

Release is replicated to the production DB


Where are we now?

First release pending

Only handles simple tasks (1:1 task/job)

Jobs run on local server

Minimal GUI - submit tasks, view raw


results

VO for landslide modelling set up


What next?

Scalability testing

Read/write throughput?

How many concurrent jobs?

Task splitting plugins

Job submission to DIRAC

Jobs running on WLCG under landslides VO

Develop GUIs
In conclusion...

We are applying WLGC tools, resources


and concepts to landslide modelling

This allows us to develop an integrated


system to provide better informed disaster
planning and reduced economic impact

Planning to have a product deployed with


engineers in the Organisation of Eastern
Caribbean States by 2012
One more thing...

Please support our VO!

vo.landslides.mossaic.org

You might also like