LOFAR DEEP FIELDS. CHALLENGES AND CLOUD SOLUTIONS.

LOFAR DEEP FIELDS.

CHALLENGES AND

CLOUD SOLUTIONS.

Jose Sabater Montes

Institute for Astronomy, University of

Edinburgh

Outline

●

LOFAR; Low Frequency Array

●

Surveys KSP: ELAIS-N1

●

Challenges

●

Cloud solutions

●

Summary

LOFAR

●

Low Frequency Array

●

Software defined radio-interferometer

working at low frequencies (30 to 240 MHz)

●

One of the Square Kilometre Array

LOFAR frequencies

●

LBA 30-80 MHz

●

HBA 120-240 MHz

LOFAR

●

Low Frequency Array

●

Software defined radio-interferometer

–

Analogue to digital conversion

–

Digital filtering and processing

LOFAR Surveys KSP

●

Science: AGN, clusters,

star formation, etc.

●

Three tiers:

–

Tier-1 all Northern

sky; ~100 μJy

–

Tier-2 selected fields;

selected areas.

–

Tier-3 a couple of

Blank field surveys, ELAIS-N1

●

Tier-2 and 3 fields with

legacy multi-wavelength

data:

–

XMM-LSS, COSMOS

–

Herschel ATLAS, Groth

strip, Bootes,

GOODS-North, North Ecliptic

Pole, Lockman Hole

Blank field surveys, ELAIS-N1

●

ELAIS-N1:

200 hours

observed so far →

LOFAR aperture synthesis

●

field of view diameter

of ~6 deg at 150 MHz

●

resolution < 5 arcsec

LOFAR imaging

r.m.s.

300 microJy In 10 hours

Calibration on

LOFAR imaging

r.m.s.

300 microJy In 10 hours

Calibration on

Challenges

●

Effect of the ionosphere → new calibration strategies

●

User data calibration

–

10 hours full resolution → ~20 TB

–

Minimum of 2 CPU years to run the calibration

–

Experimental pipeline

●

LOFAR calibration software

–

Difficult to install

Ionosphere

Calibration pipeline example.

Data

Calibration pipeline example.

Data

Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) F re qu en cy

Calibration pipeline example.

Data

F re qu en cy Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n,t_m) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n,t₂) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n,t₁) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀,t₀) Visibilities [u,v,w] (ν₁,t₀) Visibilities [u,v,w] (ν₄,t₀) Visibilities [u,v,w] (ν_n,t₀) Visibilities [u,v,w] (ν₂,t₀) Visibilities [u,v,w] (ν₃,t₀) (...) time

Calibration pipeline example.

Data

F re qu en cy Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n,t_m) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n,t₂) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n,t₁) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀) Visibilities [u,v,w] (ν₁) Visibilities [u,v,w] (ν₄) Visibilities [u,v,w] (ν_n) Visibilities [u,v,w] (ν₂) Visibilities [u,v,w] (ν₃) (...) Visibilities [u,v,w] (ν₀,t₀) Visibilities [u,v,w] (ν₁,t₀) Visibilities [u,v,w] (ν₄,t₀) Visibilities [u,v,w] (ν_n,t₀) Visibilities [u,v,w] (ν₂,t₀) Visibilities [u,v,w] (ν₃,t₀) (...) time

Calibration pipeline example.

Pipeline

Pre-processing Calibration_solutions

Pre-processing Calibration Calibrator data

360 chunks (1 sb)

Main target data

36 chunks (10 sb) Combine data: 9 chunks (40 sb)

~20 iterations

Final calibrated data

Computational solution

needed

●

Parallelizable:

–

Deal with a large amount of data in a

reasonable time.

●

Flexible:

–

Adapt the infrastructure (“hardware”) to

different calibration strategies

–

Deal with quickly changing temperamental

Cloud computing

●

Infrastructure as a

Service (IaaS)

●

Tests on:

–

Ibercloud

–

EGI Federated

Cloud

–

Amazon Web

Services

–

RAL cloud

Ibercloud

●

Based on OpenStack

●

Very easy to use

●

Discontinued and integrated on EGI Federated

Cloud

EGI Federated Cloud

●

Heterogeneous

infrastructure (access

using OCCI)

●

Many resources and

providers

●

Good support

●

Difficult to use:

–

Complex

documentation

–

Site dependent

issues

●

Blocker

: No block

storage implemented

AWS

●

SKA-AWS astrocompute proposal

●

Very easy to use:

–

Good documentation

–

Big set of useful tools (computing, object and

block storage, data transfer, etc.)

●

On-demand and pay as you go - no special

arrangements needed - ideal for a final user

●

It can be expensive

AWS: Data transfer

●

50 TB from the GRID to us-east-1 region:

–

~ 2 months

–

Lot of manual supervision: GRID proxy

renewal, data staging from magnetic tapes,

failed downloads, etc.

●

Consider AWS

Import/Export Snowball

RAL Cloud

●

Based on

OpenNebula (neither

specially difficult nor

easy to use)

●

Good support

●

It works

●

Complex generation

of the VM templates

(human intervention,

prone to errors)

●

Storage?

●

Production run?

Summary

●

LOFAR already producing data. For example,

ELAIS-N1 field - 100 TB to date.

●

New calibration strategies. Useful for SKA.

●

Big software and data managing challenges

associated to a software defined radio-telescope.

●

Cloud infrastructure to calibrate astronomical data:

–

Parallellization – Ability to deal with big data.

–

Flexibility – Quick development and testing of