Hacker News new | past | comments | ask | show | jobs | submit login
Square Kilometre Array prototype 'scope achieves first light (theregister.com)
88 points by defrost 9 months ago | hide | past | favorite | 33 comments



Disclaimer. I work for the SKA.

This article is totally wrong in some many ways, that I'm just dumbfounded. I think this was written by some AI that has no clue about reality.

Allow me to correct just a few points:

1st. China is a latecomer to the project. In fact, while they officially joined, they have very little presence on the build process. There are presently no Chinese teams working along the rest of them on the Observatory.

2nd. SKAMPI is (was) an internal technology to setup an integration ground for the work from the various teams on the SKA Observatory. It was deprecated about 6 months ago, because it was just unmaintainable (the teams were constantly breaking it). I don't understand where they got the name from, because it has absolutely nothing to do with the name. No, the SKA telescopes are not "known as SKAMPI". There was a collaboration known as "SKAMP" though, maybe that was mentioned somewhere, and it all got mixed up.

3rd. The telescopes are located in South Africa and Australia, and the headquarters where we centralise the monitoring and control in the UK.

4th. It's not driven by docker, that's just silly. The build and integration process mostly runs on 2 Kubernetes clusters. The image build process was made in house so that it's mostly automated by GitLab pipelines we created. The teams only have to submit the code, tag a release, and the pipeline builds and publishes all the artefacts for them. The controlling resides at the antennas (in this case for the Mid telescopes) also in a k8s cluster that orchestrates the various services we need.

I'm really appalled at the terrible reporting here.

EDIT: On 2. Seems like this was now adopted. A collab with Max Planck Institute, that was sometimes abbreviated as SKA-MPI, now (as of 24th of January) became SKAMPI (using the same name as the deprecated integration platform).


I'd guess the Reg got SKAMPI from the page they link to on the SKAO's own website, dated 24th Jan. this year:

https://www.skao.int/en/news/512/ska-mid-prototype-dish-crea...

No idea where the the Docker mention comes from - the other SKAO link they reference mentions Kubernetes.


Thank you for that. We were talking about this internally just now.

A team member from South Africa was just talking referring the collaboration with Max Plank Institute here, hence the name (which was SKA-MPI, and just now became SKAMPI). This is a new thing.

Other people are confused as well, since SKAMPI was the name used for the integration platform I mentioned before.


It's a fantastic project - I wish it and you every success!


Thank you. I'm very happy to be a part of it.


The kubecon talk was pretty interesting

https://youtu.be/quW8FbW1fVM?si=n-OdIdwdHUCJNQ10


Oh, you found that, thank you!

That talk gives a good high level overview of the platform we created on the observatory and that is now being used as a basis for the regional centres (where the data will be moved into and made accessible).

It's a nice talk from a more scientifically oriented point of view, but I think this one here describes better the platform we created: https://www.youtube.com/watch?v=iEpa9ltz3Xg


> China is a latecomer to the project.

I remember reading somewhere that the project only became feasible because China came up with way to save on the construction costs of the telescopes (or something like that).


I don't think that's a real issue.

The innovation isn't happening on the Mid Dishes. This is well understood tech that's been around for several decades. There are some interesting advances in the Low antennas, but those are being built mostly by Italian companies.

I think - as general knowledge - that manufacturing in China usually saves in the cost. But that's not the big part of the budget.


> I think - as general knowledge - that manufacturing in China usually saves in the cost. But that's not the big part of the budget.

'General knowledge' is that it only saves significantly in cost if your willing to accept significantly lower quality.

If somehow it was figured out to make 'well understood tech that's been around for several decades' of the same quality, at a significantly lower cost, that's literally more important than probably every other innovation or contribution in the project.


There's some really mind-bending technology out in the Karoo.

I recall a talk from a (now defunct) conference I attended a number of years ago that was given by one of the SKA researchers describing the kinds of problems they were solving with managing the vast amounts data coming out one of their telescopes: https://www.youtube.com/watch?v=OIFhC_q6Jqw


I attended a conference where they talked about this radio telescope:

https://en.wikipedia.org/wiki/Canadian_Hydrogen_Intensity_Ma...

It really is a lot of data that they have to crunch.


I wonder why they use "'scope" instead of telescope throughout the headline and article?


Someone has a pet term they're trying to jam into the lexicon? Tims are a menace.


I think in UK for example, telescopes are called "scopes" by amateur stargazers, at least.


But to someone from Anglo-America it's probably a telescopic sight.


If you put a decent transmitter on each of those 197 dishes that will be in south africa, could it knock out satellites with the combined focused EMI?


The more general version of that question would be: Whats the closest distance it can focus to?


Isn't this an interferometer? As in, compensating for lack of focus with lots of math and some clever assumptions? I don't think you can operate it in reverse the way you can with a single optical/radio system.


Isn't that the entire point behind phased arrays[1]?

[1]: https://en.wikipedia.org/wiki/Phased_array


GPS and some geostationary satellites should be possible, LEO not so much as they'd be below the horizon for at leas half of the array.


Interesting how docker was profiled as important to the project. Certainly there are hundreds or thousands of other just as important technologies involved.


According to the top comment now, which was written two hours after yours, they use two Kubernetes clusters. Details at https://news.ycombinator.com/item?id=39201980


Unless they don't use Docker images in those k8s clusters that's not too bad for a news article I guess.


    And then there's all the back-end compute and storage needed to handle the streams of data all those observatories will produce – a topic The Register will monitor.
https://pawsey.org.au/

https://pawsey.org.au/supercomputing/

https://pawsey.org.au/systems/data-cloud/

https://pawsey.org.au/visualisation/


Does Pawsey also look after the South African SKA data? Seems a bit far for data transfer



> One of the computers will be based in Cape Town.

FYI, it makes sense for this location as it's still (relatively) close the the SKA telescopes, which are of course in a high, dry, remote and sparsely inhabited area. ( https://en.wikipedia.org/wiki/Sutherland,_South_Africa )

And Cape Town is the nearest large coastal population centre where several sub-sea internet cables connect to the land, so it's closer to the rest of the global network in that regard.


A small correction :). SKA in South Africa is based closer to Carnavron than to Sutherland.

https://www.skao.int/en/arrange-visit

Not that Sutherland is top be ignored, it does host the South African Astronomical Observatory.

PS, it gets damn cold there in the desert, even during summer :).

https://en.wikipedia.org/wiki/South_African_Astronomical_Obs...


History wise the SKA concept was put forward and pressed forward from Australia in the (contrary to wikipedia) 1980s.

Then it got internationally political with grant money coming in and backing from China, Britain, etc.

The story today is:

    The Square Kilometre Array (SKA) is an intergovernmental international radio telescope project being built in Australia (low-frequency) and South Africa (mid-frequency).

    The combining infrastructure, the Square Kilometre Array Observatory (SKAO), and headquarters, are located at the Jodrell Bank Observatory in the United Kingdom.

    The SKA cores are being built in the southern hemisphere, where the view of the Milky Way galaxy is the best and radio interference at its least.

    Conceived in the 1990s, and further developed and designed by the late-2010s, when completed sometime in the 2020s it will have a total collecting area of approximately one square kilometre. 
https://en.wikipedia.org/wiki/Square_Kilometre_Array

https://www.skao.int/en

https://pawsey.org.au/projects/australian-square-kilometre-a...

Pawsey has been up and running for some time now, well in advance of the actual physical SKA element build - modelling and computing for other Murchison radio quiet zone projects (cosmic microwave background, etc.)

It's now getting up and running and more and more fully international - but over the past two decades the bulk of SKA material online has come from Australia, the CSIRO and Pawsey.

It won't be the only local supercomputing centre anymore, with others in S.Africa and elsewhere, but it was the leader for PoC and illustrates the kinds of compute systems that will lie behind the project.


> History wise the SKA concept was put forward and pressed forward from Australia in the (contrary to wikipedia) 1980s.

It's not that simple. As one of The Australian CSIRO's own researchers points out, the basic concept of the SKA was independently developed by a number of researchers and institutions including the Project Cyclops proposal by Barney Oliver and John Billingham in 1971. [0]

> Then it got internationally political with grant money coming in and backing from China, Britain, etc.

It was international from the start, the result of merging a number of different international ideas and proposals.

> Pawsey has been up and running for some time now, well in advance of the actual physical SKA element build - modelling and computing for other Murchison radio quiet zone projects (cosmic microwave background, etc.)

> It's now getting up and running and more and more fully international - but over the past two decades the bulk of SKA material online has come from Australia, the CSIRO and Pawsey.

> It won't be the only local supercomputing centre anymore, with others in S.Africa and elsewhere, but it was the leader for PoC and illustrates the kinds of compute systems that will lie behind the project.

While Pawsey is an interesting and remarkable centre in its own right, and will be an important element of SKA and other radio astronomy, it is not the only processing centre for SKA and it's too Australia-centric to mention only it each time this question is raised.

For instance, the South African array has also been operating for some time, beginning with the KAT-7 and MeerKAT precursor arrays, and South African companies have been at the forefront of developing new mid-frequency high-bandwidth FPGA-based signal processing solutions like ROACH-1, ROACH-2, and SKARAB both for them and the full SKA-Mid array. [1][2][3]

I have already shared information about the Cape Town Science Processing Centre in another comment, but it's worth noting that the different frequencies used means that the data rate requirements are much higher for the South African centre, at around 20 terabits per second versus the 8 terabits per second needed to get SKA-Low data to Pawsey. [4]

SKA is amazing in many ways, not least because of how international it was from the very beginning

[0] https://arxiv.org/pdf/1212.3497.pdf

[1] https://mg.co.za/article/2015-12-03-ska-gives-high-tech-firm...

[2] https://indico.ict.inaf.it/event/2367/contributions/14597/at...

[3] https://www.sarao.ac.za/wp-content/uploads/2016/12/ska_sa_te...

[4] https://www.skao.int/en/explore/big-data


Sure, there was a different big antenna idea of a different size and a different end goal sketched out in the early 1970s.

This antenna project, with a square kilometre target and Lo|Mid frequency ranges was sketched out largely by European and Commonwealth astros at Commonwealth Radio conferences - your link does mention the earlier work by Canadians and parallel|later work by the Dutch to craft small dish arrays to survey the sky (as opposed to the Cyclops SETI goals).

The event that really kicked things along (as I distantly recall) from an idea on the back of envelopes to a project with dot points looking for international backers was the 1988 collapse of the 300 Foot Telescope - whether it was due to load, to vibrations, poor materials, etc. it served to launch several new lines of creating larger telescopes - suspended across natural craters, broken into arrays of smaller dishes, etc.

SKA as a specific project had some back history before it "officially" was launched; before the 2011 legal entity, before the 1998 name adoption, before the 1990 anniversary of VLA's.

My comment was motivated largely by looking back in wonder at just how long it takes to get projects of this scale up and running - I've attended pie in the sky proposals for variations on this since 1982 (same year as my first introduction to the (then) current state of gravity wave detection).

> While Pawsey .. it is not the only processing centre for SKA

I can't see that I made that claim, I did link to Pawsey as an example of other systems that lie behind SKA and another comment linked to the parallel systems being launched in S.Africa.


I actually interviewed for a position on this project xD

But the science knowledge on my side was way too lacking and I didn't want to slow things down.




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: