Archive for the ‘Scientific Software’ Category

September 22nd, 2019

In many introductions to numpy, one gets taught about np.ones, np.zeros and np.empty. The accepted wisdom is that np.empty will be faster than np.ones because it doesn’t have to waste time doing all that initialisation. A quick test in a Jupyter notebook shows that this seems to be true!

import numpy as np

N = 200
zero_time = %timeit -o some_zeros = np.zeros((N,N))
empty_time = %timeit -o empty_matrix = np.empty((N,N))
print('np.empty is {0} times faster than np.zeros'.format(zero_time.average/empty_time.average))


8.34 µs ± 202 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
436 ns ± 10.6 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
np.empty is 19.140587654682545 times faster than np.zeros

20 times faster may well be useful in production when using really big matrices. Might even be worth the risk of dealing with uninitialised variables even though they are scary!

However…..on my machine (Windows 10, Microsoft Surface Book 2 with 16Gb RAM), we see the following behaviour with a larger matrix size (1000 x 1000).

import numpy as np

N = 1000
zero_time = %timeit -o some_zeros = np.zeros((N,N))
empty_time = %timeit -o empty_matrix = np.empty((N,N))
print('np.empty is {0} times faster than np.zeros'.format(zero_time.average/empty_time.average))


113 µs ± 2.97 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)
112 µs ± 1.01 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)
np.empty is 1.0094651980894993 times faster than np.zeros

The speed-up has vanished! A subsequent discussion on twitter suggests that this is probably because the operating system is zeroing all of the memory for security reasons when using numpy.empty on large arrays.

November 20th, 2018

In a recent blog post, Daniel Lemire explicitly demonstrated that vectorising random number generators using SIMD instructions could give useful speed-ups.  This reminded me of the one of the first times I played with the Julia language where I learned that Julia’s random number generator used a SIMD-accelerated implementation of Mersenne Twister called dSFMT to generate random numbers much faster than MATLAB’s Mersenne Twister implementation.

Just recently, I learned that MATLAB now has its own SIMD accelerated version of Mersenne Twister which can be activated like so:

seed=1;
rng(seed,'simdTwister')

This new Mersenne Twister implementation gives different random variates to the original implementation (which I demonstrated is the same as Numpy’s implementation in an older post) as you might expect

>> rng(1,'Twister')
>> rand(1,5)
ans =
0.4170 0.7203 0.0001 0.3023 0.1468
>> rng(1,'simdTwister')
>> rand(1,5)
ans =
0.1194 0.9124 0.5032 0.8713 0.5324

So it’s clearly a different algorithm and, on CPUs that support the relevant instructions, it’s about twice as fast!  Using my very unscientific test code:

format compact
number_of_randoms = 10000000
disp('Timing standard twister')
rng(1,'Twister')
tic;rand(1,number_of_randoms);toc
disp('Timing SIMD twister')
rng(1,'simdTwister')
tic;rand(1,number_of_randoms);toc

gives the following results for a typical run on my Dell XPS 15 9560 which supports AVX instructions

number_of_randoms =
    10000000
Timing standard twister
Elapsed time is 0.101307 seconds.
Timing SIMD twister
Elapsed time is 0.057441 seconds

The MATLAB documentation does not tell us which algorithm their implementation is based on but it seems to be different from Julia’s. In Julia, if we set the seed to 1 as we did for MATLAB and ask for 5 random numbers, we get something different from MATLAB:

julia> using Random
julia> Random.seed!(1);
julia> rand(1,5)
1×5 Array{Float64,2}:
 0.236033  0.346517  0.312707  0.00790928  0.48861

The performance of MATLAB’s new generator is on-par with Julia’s although I’ll repeat that these timing tests are far far from rigorous.

julia> Random.seed!(1);

julia> @time rand(1,10000000);
  0.052981 seconds (6 allocations: 76.294 MiB, 29.40% gc time)
August 24th, 2018

Audiences can be brutal

I still have nightmares about the first talk I ever gave as a PhD student. I was not a strong presenter, my grasp of the subject matter was still very tenuous and I was as nervous as hell. Six months or so into my studentship, I was to give a survey of the field I was studying to a bunch of very experienced researchers.  I spent weeks preparing…practicing…honing my slides…hoping that it would all be good enough.

The audience was not kind to me! Even though it was only a small group of around 12 people, they were brutal! I felt like they leaped upon every mistake I made, relished in pointing out every misunderstanding I had and all-round gave me a very hard time.  I had nothing like the robustness I have now and very nearly quit my PhD the very next day. I can only thank my office mates and enough beer to kill a pony for collectively talking me out of quitting.

I remember stopping three quarters of the way through saying ‘That’s all I want to say on the subject’ only for one of the senior members of the audience to point out that ‘You have not talked about all the topics you promised’.  He made me go back to the slide that said something like ‘Things I will talk about’ or ‘Agenda’ or whatever else I called the stupid thing and say ‘Look….you’ve not mentioned points X,Y and Z’ [1].

Everyone agreed and so my torture continued for another 15 minutes or so.

Practice makes you tougher

Since that horrible day, I have given hundreds of talks to audiences that range in size from 5 up to 300+ and this amount of practice has very much changed how I view these events.  I always enjoy them…always!  Even when they go badly!

In the worst case scenario, the most that can happen is that I get given a mildly bad time for an hour or so of my life but I know I’ll get over it. I’ve gotten over it before. No big deal! Academic presentations on topics such as research computing rarely lead to life threatening outcomes.

But what if it was recorded?!

Anyone who has worked with me for an appreciable amount of time will know of my pathological fear of having one of my talks recorded. Point a camera at me and the confident, experienced speaker vanishes and is replaced by someone much closer to the terrified PhD student of my youth.

I struggle to put into words what I’m so afraid of but I wonder if it ultimately comes down to the fact that if that PhD talk had been recorded and put online, I would never have been able to get away from it. My humiliation would be there for all to see…forever.

JuliaCon 2018 and Rise of the Research Software Engineer

When the organizers of JuliaCon 2018 invited me to be a keynote speaker on the topic of Research Software Engineering, my answer was an enthusiastic ‘Yes’. As soon as I learned that they would be live streaming and recording all talks, however, my enthusiasm was greatly dampened.

‘Would you mind if my talk wasn’t live streamed and recorded’ I asked them.  ‘Sure, no problem’ was the answer….

Problem averted. No need to face my fears this week!

A fellow delegate of the conference pointed out to me that my talk would be the only one that wouldn’t be on the live stream. That would look weird and not in a good way.

‘Can I just be live streamed but not recorded’ I asked the organisers.  ‘Sure, no problem’ [2] was the reply….

Later on the technician told me that I could have it recorded but it would be instantly hidden from the world until I had watched it and agreed it wasn’t too terrible.  Maybe this would be a nice first step in my record-a-talk-a-phobia therapy he suggested.

So…on I went and it turned out not to be as terrible as I had imagined it might be.  So we published it. I learned that I say ‘err’ and ‘um’ a lot [3] which I find a little embarrassing but perhaps now that I know I have that problem, it’s something I can work on.

Rise of the Research Software Engineer

Anyway, here’s the video of the talk. It’s about some of the history of The Research Software Engineering movement and how I worked with some awesome people at The University of Sheffield to create a RSE group. If you are the computer-person in your research group who likes software more than papers, you may be one of us. Come join the tribe!

Slide deck at mikecroucher.github.io/juliacon2018/

Feel free to talk to me on twitter about it: @walkingrandomly

Thanks to the infinitely patient and wonderful organisers of JuliaCon 2018 for the opportunity to beat one of my long standing fears.

Footnotes

[1] Pro-Tip: Never do one of these ‘Agenda’ slides…give yourself leeway to alter the course of your presentation midway through depending on how well it is going.

[2] So patient! Such a lovely team!

[3] Like A LOT! My mum watched the video and said ‘No idea what you were talking about but OMG can you cut out the ummms and ahhs’

May 23rd, 2017

I’m working on optimising some R code written by a researcher at University of Sheffield and its very much a war of attrition! There’s no easily optimisable hotspot and there’s no obvious way to leverage parallelism. Progress is being made by steadily identifying places here and there where we can do a little better. 10% here and 20% there can eventually add up to something worth shouting about.

One such micro-optimisation we discovered involved multiplying two matrices together where one of them needed to be transposed. Here’s a minimal example.

#Set random seed for reproducibility
set.seed(3)

# Generate two random n by n matrices
n = 10
a = matrix(runif(n*n,0,1),n,n)
b = matrix(runif(n*n,0,1),n,n)

# Multiply the matrix a by the transpose of b
c = a %*% t(b)

When the speed of linear algebra computations are an issue in R, it makes sense to use a version that is linked to a fast implementation of BLAS and LAPACK and we are already doing that on our HPC system.

Here, I am using version 3.3.3 of Microsoft R Open which links to Intel’s MKL (an implementation of BLAS and LAPACK) on a Windows laptop.

In R, there is another way to do the computation c = a %*% t(b)  — we can make use of the tcrossprod function (There is also a crossprod function for when you want to do t(a) %*% b)

 c_new = tcrossprod(a,b)

Let’s check for equality

c_new == c
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
[1,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[2,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[3,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[4,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[5,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[6,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[7,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[8,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[9,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[10,] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE

Sometimes, when comparing the two methods you may find that some of those entries are FALSE which may worry you!
If that happens, computing the difference between the two results should convince you that all is OK and that the differences are just because of numerical noise. This happens sometimes when dealing with floating point arithmetic (For example, see http://www.walkingrandomly.com/?p=5380).

Let’s time the two methods using the microbenchmark package.

install.packages('microbenchmark')
library(microbenchmark)

We time just the matrix multiplication part of the code above:

microbenchmark(
original = a %*% t(b),
tcrossprod = tcrossprod(a,b)
)


Unit: nanoseconds
expr min lq mean median uq max neval
original 2918 3283 3491.312 3283 3647 18599 1000
tcrossprod 365 730 756.278 730 730 10576 1000

We are only saving microseconds here but that’s more than a factor of 4 speed-up in this small matrix case. If that computation is being performed a lot in a tight loop (and for our real application, it was), it can add up to quite a difference.

As the matrices get bigger, the speed-benefit in percentage terms gets lower but tcrossprod always seems to be the faster method. For example, here are the results for 1000 x 1000 matrices

#Set random seed for reproducibility
set.seed(3)

# Generate two random n by n matrices
n = 1000
a = matrix(runif(n*n,0,1),n,n)
b = matrix(runif(n*n,0,1),n,n)

microbenchmark(
original = a %*% t(b),
tcrossprod = tcrossprod(a,b)
)

Unit: milliseconds
expr min lq mean median uq max neval
original 18.93015 26.65027 31.55521 29.17599 31.90593 71.95318 100
tcrossprod 13.27372 18.76386 24.12531 21.68015 23.71739 61.65373 100

The cost of not using an optimised version of BLAS and LAPACK

While writing this blog post, I accidentally used the CRAN version of R.  The recently released version 3.4. Unlike Microsoft R Open, this is not linked to the Intel MKL and so matrix multiplication is rather slower.

For our original 10 x 10 matrix example we have:

library(microbenchmark)
#Set random seed for reproducibility
set.seed(3)

# Generate two random n by n matrices
n = 10
a = matrix(runif(n*n,0,1),n,n)
b = matrix(runif(n*n,0,1),n,n)

microbenchmark(
original = a %*% t(b),
tcrossprod = tcrossprod(a,b)
)

Unit: microseconds
       expr   min    lq    mean median     uq    max neval
   original 3.647 3.648 4.22727  4.012 4.1945 22.611   100
 tcrossprod 1.094 1.459 1.52494  1.459 1.4600  3.282   100

Everything is a little slower as you might expect and the conclusion of this article — tcrossprod(a,b) is faster than a %*% t(b) — seems to still be valid.

However, when we move to 1000 x 1000 matrices, this changes

library(microbenchmark)
#Set random seed for reproducibility
set.seed(3)

# Generate two random n by n matrices
n = 1000
a = matrix(runif(n*n,0,1),n,n)
b = matrix(runif(n*n,0,1),n,n)

microbenchmark(
original = a %*% t(b),
tcrossprod = tcrossprod(a,b)
)

Unit: milliseconds
       expr      min       lq     mean   median       uq       max neval
   original 546.6008 587.1680 634.7154 602.6745 658.2387  957.5995   100
 tcrossprod 560.4784 614.9787 658.3069 634.7664 685.8005 1013.2289   100

As expected, both results are much slower than when using the Intel MKL-lined version of R (~600 milliseconds vs ~31 milliseconds) — nothing new there.  More disappointingly, however, is that now tcrossprod is slightly slower than explicitly taking the transpose.

As such, this particular micro-optimisation might not be as effective as we might like for all versions of R.

March 29th, 2017

UK to launch 6 major HPC centres

Tomorrow, I’ll be attending the launch event for the UK’s new HPC centres and have been asked to deliver a short talk as part of the program. As someone who paddles in the shallow-end of the HPC pool I find this both flattering and more than a little terrifying. What can little-ole-me say to the national HPC glitterati that might be useful?

This blog post is an attempt at gathering my thoughts together for that talk.

The technology gap in research computing

Broadly speaking, my role in academia is to hang out with researchers, compute providers (cloud and HPC) and software vendors in an attempt to be vaguely useful in the area of research software. I’m a Research Software Engineer with a focus on Long Tail Science: The large number of very small research groups who do a huge amount of modern research.

Time and again, what I see can be summarized in this quote by Greg Wilsongwilson

This is very true in the world of High Performance Computing.

Geek Top Gear

I love technology and I love HPC in particular. I love to geek out on Flops, Ghz, SIMD instructions, GPUs, FPGAs…..all that stuff. I help support The University of Sheffield’s local HPC service and worked in Research IT at The University of Manchester for around a decade before moving to Sheffield.

I’ve given and seen many a HPC-related talk in my time and have certainly been guilty of delivering what I now refer to as the ‘Geek Top Gear’ speech.  For maximum effect, you need to do it in a Jeremy Clarkson voice and, if you’re feeling really macho, kiss your bicep at the point where you tell the audience how many Petaflops your system can do in Linpack.

*Begin Jeremy Clarkson Impression*

Our new HPC system has got 100,000 of the latest Intel Kaby Lake cores...which is a lot!

Usually running at 2.6Ghz, these cores can turbo-boost to 3.2Ghz for those moments when we need that extra boost of power. Obviously, being Kaby Lake, these cores have all the instruction extensions you’d expect with AVX2, FMA, SSE, ABM and many many other TLAs for all your SIMD needs. Of course every HPC system needs accelerators…..and we have the best of them: Xeon Phis with 68 cores each and NVIDIA GPUs with thousands of tiny little cores will handle every massively parallel job you can throw at them….Easily. We connect these many many cores together with high-speed interconnect fashioned from threads of pure unicorn hair and cool the whole thing with the tears of virigin nerds.

YEEEEEES! Our new HPC system is the best one since the last one and, achieving over a Gajillion Petaflops in the Linpack test (kiss bicep), it will change your life forever.

more_power

Any questions?

Audience member 1: What’s a core?
Audience member 2: Why does it run my R script slower than my laptop?
Audience member 3: Do you have Excel installed on it?

There is a huge gap between what many HPC providers like to focus on and what the typical long-tail researcher wants or needs. I propose that the best bridge for this gap is the Research Software Engineer (RSE).

Research Software Engineer as Alpine guide

In my fellowship proposal, I compared the role of a Research Software Engineer to that of an alpine guide:

Technological development in software is more like a cliff-face than a ladder – there are many routes to the top, to a solution. Further, the cliff face is dynamic – constantly and quickly changing as new technologies emerge and decline. Determining which technologies to deploy and how best to deploy them is in itself a specialist domain, with many features of traditional research.

Researchers need empowerment and training to give them confidence with the available equipment and the challenges they face. This role, akin to that of an Alpine guide, involves support, guidance, and load carrying. When optimally performed it results in a researcher who knows what challenges they can attack alone, and where they need appropriate support. Guides can help decide whether to exploit well-trodden paths or explore new possibilities as they navigate through this dynamic environment.

At Sheffield, we have built a pool of these Research Software Engineers to provide exactly this kind of support and it’s working extremely well so far. Not only are we helping individual research groups but we are also using our experiences in the field to help shape the University HPC environment in collaboration with the IT department.

Supercomputing: Irrelevant to many?

“Never bring an anecdote to a data-fight” so the saying goes and all I have from my own experiences are a bucket load of anecdotes, case studies and cursory log-mining experiments that indicate that even those who DO use HPC are not doing so effectively. Fortunately, others have stepped up to the plate and we have survey and interview data on how researchers are using compute resources.

How Do Scientists Develop and Use Scientific Software? is a report on a 2009 survey of 1972 researchers from around the world. They found that “79.9% of the scientists never use scientific software on a supercomputer

When I first learned of this number, I found it faintly depressing. This technology that I love so much and for which University IT departments dedicate special days to seems to be pretty much irrelevant to the majority of researchers. Could it be that even in an era of big data, machine learning and research software engineering that most people only need a laptop?

Only ever needing a laptop certainly doesn’t fit with what I’ve seen while working in the trenches. Almost every researcher I’ve met who does computational research wishes it was faster or that they had more memory to allow them to do larger problems. Speed is the easiest thing to sell to researchers in the world of RSE. They come for faster execution and leave with a side-order of version control, testing and documentation. A combination of software development and migration to even a small HPC system can easily result in 100x or even 1000x speed-ups for many researchers.

In my experience, it’s not that researchers don’t need HPC, it’s that the jump from their laptop-based workflow to one that makes good use of a HPC system is too large for them to bridge without a little help. Providing that help can result in some great partnerships such as the recent one between the Sheffield RSE group and the Sheffield Faculty of Arts and Humanities.

language

Want to know how that partnership started? I compiled an experimental R/Rcpp package that they were struggling with and then took them for coffee and said ‘That code took a while to run. Here’s how we can make it go faster….Now…what exactly are you doing because it looks cool?’ Fast forward a year or so and we are on the cusp of starting a great new project that will include traditional HPC and cloud computing as part of their R-based workflow.

My experiences seem to be reflected in the data. In  their 2011 article, A Survey of the Practice of Computational Science, Prabhu et al interviewed 114 randomly selected researchers from Princeton University. Princeton have a very strong, well supported HPC centre which provides both resources and the expertise to use them. Even at such a well equipped institution, the authors write that  ‘Despite enormous wait times, many scientists run their programs only on desktops’ although they did report much higher HPC usage compared to the Hanny et al survey.

Other salient quotes from the Prabhu interviews include

“only 18% of researchers who optimize code leveraged profiling tools to inform their optimization plans”

“only 7% of researchers leveraged any form of thread based shared memory CPU parallelism”

“Only 11% of researchers utilized loop based parallelism”

“Currently, many researchers fit their scientific models to only a subset of available parameters for faster program runs.”

“Across disciplines, an order of magnitude performance improvement was cited as a requirement for significant changes in research quality”

HPC: There’s plenty of room at the bottom

Potential users of HPC look different to those of 20 years ago. The popularity explosion of languages such as MATLAB, Python and R have democratized programming and the world is awash with inefficient research software. Most of the time, this lack of efficiency is not a problem (see ‘In defense of inefficient scientific code‘) but if a researcher needs to scale up what they are doing, it can become limiting. Researchers might wait for days for the results to come in and limit the scope of their investigations to fit the hardware they have access to — their laptop usually.

The paper of Prabu et al said that an order of magnitude (10x) speed up was cited by researchers as a requirement for significant changes in research quality. For an experienced Research Software Engineer with access to cloud or HPC facilities, a 10x speed-up is usually pretty easy to achieve for this new audience. 100x or even 1000x can be achieved fairly frequently if you employ multiple hardware and software techniques. Compared to squeezing out a few percent more performance from HPC-centric code such as LAPACK or CASTEP, it’s not even all that difficult. I recently sped up one researcher’s MATLAB code by a factor of 800x in a couple of days and I’m a fairly middling developer if I’m brutally honest.

The whole point of High Performance Computing is to accelerate science and right now there is more computational science around than there has ever been before. Furthermore, it’s easier than ever to accelerate! There’s plenty of room at the bottom.

Closing the computational gap with people, training and compute power

The UK’s 6 new HPC centers represent the cutting edge of hardware technology. They provide a crucial component of our national hardware infrastructure, will contribute to research in HPC itself and will doubtless be of huge benefit to computational science. Furthermore, all of the funded proposals include significant engagement with the national Research Software Engineering community – the vital bridge between many researchers and HPC.

Co-development of research software with effort from both RSEs and researchers can be an extremely powerful model. Combine this with further collaboration between RSEs and compute providers and we have an environment that I think is both very exciting and capable of helping to close the rich/poor compute divide.

As an RSE who works with both researchers and University-level HPC providers, I ask for 3 things to be considered by these new regional centres.

  • Enjoy your new compute-ferraris. I look forward to seeing how hard you can push them.
  • You will be learning new good practice in how to provide HPC services. Disseminate this to those of us running smaller services.
  • There’s plenty of room at the bottom! Help us to support the new wave of computational researchers.

Thanks to languages such as MATLAB, Python and R, general purpose programming has been fully democratized. I look forward to working with these new centres to help democratize high performance computing.

 

January 12th, 2017

If you are a researcher and are currently writing scripts or developing code then I have a suggestion for you. If you haven’t done it already, get yourself a willing volunteer and send them your code/analysis/simulation/voodoo and ask them to run it on their machine to see what happens. Bonus points are awarded for choosing someone who uses a different operating system from you!

This simple act is one of the things I recommend in my talk Is Your Research Software Correct and it can often help improve both code and workflow.

It quickly exposes patterns that are not good practice. For example, scattered references to ‘/home/walkingrandomly/mydata.dat’ suddenly don’t seem like a great idea when your code buddy is running windows. The ‘minimal tweaking’ required to move your analysis from your machine to theirs starts to feel a lot less minimal as you get to the bottom of the second page of instructions.

Crashy McCrashFace

When I start working with someone new, the first thing I ask them to do is to provide access to their code and simple script called runme or similar that will build and run their code and spit out an answer that we agree is OK. Many projects stumble at this hurdle! Perhaps my compiler is different to theirs and objects to their abuse (or otherwise) of the standards or maybe they’ve forgotten to include vital dependencies or input data.

Email ping-pong ensues as we attempt to get the latest version…zip files with names like PhD_code_ver1b_ForMike_withdata_fixed.zip get thrown about while everyone wonders where Bob is because he totally got it working on Windows back in 2009.

git clone

‘Hey Mike, just clone the git repo and run the test suite. It should be fine because the latest continuous integration run didn’t throw up any issues. The benchmark code and data we’d like you to optimise is in the benchmarks folder along with the timings and results from our most recent tests. Ignore the papers folder, that just reproduces all of the results from our recent papers and links to Zenodo DOIs’

‘…………’

‘Are you OK Mike?’

‘I’m…..fine. Just have something in my eye’

 

January 10th, 2017

I work at The University of Sheffield where I am one of the leaders of the new Research Software Engineering function. One of the things that my group does is help people make use of Sheffield’s High Performance Computing cluster, Iceberg.

Iceberg is a heterogenous system with around 3440 CPU cores and a sprinkling of GPUs. It’s been in use for several years and has been upgraded a few times over that period. It’s a very traditional HPC system that makes use of Linux and a variant of  Sun Grid Engine as the scheduler and had served us well.

iceberg

A while ago, the sysadmin pointed me to a goldmine of a resource — Iceberg’s accounting log. This 15 Gigabyte file contains information on every job submitted since July 2009. That’s more than 7 years of the HPC usage of 3249 users — over 46 million individual jobs.

The file format is very straightforward. There’s one line per job and each line consists of a set of colon separated fields.  The first few fields look like something like this:

long.q:node54.iceberg.shef.ac.uk:el:abc07de:

The username is field 4 and the number of slots used by the job is field 35. On our system, slots correspond to CPU cores. If you want to run a 16 core job, you ask for 16 slots.

With one line of awk, we can determine the maximum number of slots ever requested by each user.

gawk -F: '$35>=slots[$4] {slots[$4]=$35};END{for(n in slots){print n, slots[n]}}' accounting > ./users_max_slots.csv

As a quick check, I grepped the output file for my username and saw that the maximum number of cores I’d ever requested was 20. I ran a 32 core MPI ‘Hello World’ job, reran the line of awk and confirmed that my new maximum was 32 cores.

There are several ways I could have filtered the number of users but I was having awk lessons from David Jones so let’s create a new file containing the users who have only ever requested 1 slot.

gawk -F: '$35>=slots[$4] {slots[$4]=$35};END{for(n in slots){if(slots[n]==1){print n, slots[n]}}}' accounting > users_where_max_is_one_slot.csv

Running wc on these files allows us to determine how many users are in each group

wc users_max_slots.csv 

3250  6498 32706 users_max_slots.csv

One of those users turned out to be a blank line so 3249 usernames have been used on Iceberg over the last 7 years.

wc users_where_max_is_one_slot.csv 
2393  4786 23837 users_where_max_is_one_slot.csv

That is, 2393 of our 3249 users (just over 73%) over the last 7 years have only ever run 1 slot, and therefore 1 core, jobs.

High Performance?

So 73% of all users have only ever submitted single core jobs. This does not necessarily mean that they have not been making use of parallelism. For example, they might have been running job arrays – hundreds or thousands of single core jobs performing parameter sweeps or monte carlo simulations.

Maybe they were running parallel codes but only asked the scheduler for one core. In the early days this would have led to oversubscribed nodes, possibly up to 16 jobs, each trying to run 16 cores.These days, our sysadmin does some voodoo to ensure that jobs can only use the number of cores that have been requested, no matter how many threads their code is spawning. Either way, making this mistake is not great for performance.

Whatever is going on, this figure of 73% is surprising to me!

Thanks to David Jones for the awk lessons although if I’ve made a mistake, it’s all my fault!

Update (11th Jan 2017)

UCL’s Ian Kirker took a look at the usage of their general purpose cluster and found that 71.8% of their users have only ever run 1 core jobs. https://twitter.com/ikirker/status/819133966292807680

 

 

December 12th, 2016

I was in Stockholm last week to give an invited talk at the Workshop on Nordic Big Biomedical Data for Action. I was representing the Software Sustainability Institute and delivered the latest version of my talk Is Your Research Software Correct? screen-shot-2016-12-11-at-12-47-14

It was a great event which introduced me to some nice initiatives going on waaaay up north. Initiatives such as Code Refinery who’s aims align well with those of the UK’s software sustainability Institute. Code refinery was introduced by Radovan Bast — Slide deck at http://cicero.xyz/v2/remark/github/coderefinery/talk-intro/niasc-2016/talk.md/#1

coderefinery

Other talks included the introduction of a scalable, parallel version of BLAST, Big Data Processing for Genomics and Delivering Bioinformatics Software as Virtual Machine images. I also got chance to geek out with some High Performance Computing and Bioinformatics people over interesting Swedish food.

Slides from most of the talks are available at http://www.nordicehealth.se/2016/12/04/workshop-on-nordic-big-biomedical-data-for-action/

October 13th, 2016

I was recently invited to give a talk at the Sheffield R Users Group and decided to give a brief overview of how R relates to other technologies. Subjects included Mathematica’s integration of R, Intel’s compilers, Math Kernel Library and how they can make R faster and a range of Microsoft technologies including R Tools for Visual Studio, Microsoft R Open and the MRAN for reproducibility. I also touched upon the NAG Library, Maple’s code generation for R, GPUs and Spark.

Did I miss anything? If you were to give a similar talk, what might you have included?

September 5th, 2016

One of the great things about being a Research Software Engineer is the diversity of work you can get involved with. I specialise in smaller interventions which means that I can be working with physicists on Monday, engineers on Tuesday, geneticists on Wednesday….you get the idea.

Last month, I got to work with some Ecologists along with Anna Krystalli. We undertook the arduous journey from Sheffield down to Exeter to deliver talks and workshops at a post-conference symposium on reproducibility in science, organised by Malika Ihle and Isabel Winney, at the International Symposium on Behavioural Ecology.

I gave my talk, Is your research software correct?, and also delivered a workshop on using projects and version control using R and RStudio in the Code Cafe style. For the full write up of the day, see the excellent blog post by Anna over at the Mozilla Science Lab blog.

Updates : More resources