- cloud providers
- Computation node suppliers
- Parallel tasks on your awful ancient “High Performance” computing cluster that you hate but your campus spent lots of money on and it IS free so uh…
- Local parallel tasks with python
- Scientific VMs
- Scientific containers
I get lost in all the options for parallel computing on the cheap. I summarise for myself here.
There are roadmaps here, e.g. the one by Cloud Native Computing foundation, Landscape. However, for me it exemplifies my precise problems with the industry, in that it mistakes an underexplained information deluge for actionable advice.
So, back to the old-skool: Lets find some specific things that work, implement solutions to the problems I have, and generalise as needed.
Fashion dictates this should be called “cloud” computing, although I’m also interested in using the same methods without a cloud, as such. In fact, I would prefer frictionless switching between such things according to debugging and processing power needs.
My emphasis is strictly on the cloud doing large data analyses and prototyping algorithms. I don’t care about serving web pages or streaming videos or whatever, or “deploying” anything. I’m a researcher, not an engineer.
In particular I mostly want to do embarrassingly parallel computation That is, I run many calculations/simulations with absolutely no shared state and aggregate them in some way at the end. This avoids much of graph computing complexity. UPDATE: Some of my stuff is deep learning now, which is not quite as embarrassingly parallel, but rather excruciatingly parallel.
A local hub for this stuff is C3DIS Conference, the Collaborative Conference on Computational and DAta Intensive Science.
- Is a billion-dollar worth of server lying on the ground tl;dr Amazon is weidly expensive. Why not go, e.g. OVH?
Algorithms, implementations thereof, and providers of parallel processing services are all coupled closely. Nonetheless I’ll try to draw a distinction between the three.
Since I am not a startup trying to do machine-learning on the cheap, but a researcher implementing algorithms, it’s essential what whatever I use can get me access “under the hood” of machine learning; I’m writing my own algorithms. Using frameworks which allow me to employ only someone else’s statistical algorithms is pointless for my job. OTOH, I want to know as little as possible about the mechanics and annoyances of parallelisation and compute cloud configuration and all that nonsense, although that does indeed impinge upon my algorithm design etc.
All the parts of ML algorithm design relate through complicated workflows, and there are a lot of ways to skin this ML cat. I’ll mention various tools that handle various bits of the workflow in a giant jumble.
Plain old HPC clusters, as seen on campus, I discuss at hpc hell.
RONIN (ALLCAPS apparently obligatory) is
an incredibly simplistic web application that allows researchers and scientists to launch complex compute resources within minutes, without the nerding.
It seems to handle provisioning virtual machines in an especially friendly way for ML. It also seems to be frighteningly sparsely documented especially with regard to certain key features for me: How do I design my own machine with my desired data and code to actually do a specific thing? I’m sure there is an answer to this, it is just not anywhere obvious on the project site.
determined promises that
Going from Notebook to Cluster has never been simpler
Seems to be a system for exploratory deep learning training setups.
Determined allows you to focus on the task at hand — training models.
Jump right into a purpose-created environment for DL work — and spend your time getting your models right without having to worry about setup, teardown, and other boilerplate code that can be automated away.
We’ve done this enough to know what you don’t want to spend your time doing.
A built-in training loop abstraction that supports experiment tracking, efficient data loading, fault tolerance, and flexible customization.
High-performance distributed training with no code changes.
Automatic hyperparameter optimization based on cutting-edge research.
tasks and actors use the same Ray API and are used the same way. This unification of parallel tasks and actors has important benefits, both for simplifying the use of Ray and for building powerful applications through composition.
By way of comparison, popular data processing systems such as Apache Hadoop and Apache Spark allow stateless tasks (functions with no side effects) to operate on immutable data. This assumption simplifies the overall system design and makes it easier for applications to reason about correctness.
However, shared mutable state is common in machine learning applications. That state could be the weights of a neural network, the state of a third-party simulator, or a representation of interactions with the physical world. Ray’s actor abstraction provides an intuitive way to define and manage mutable state in a thread-safe way.
What makes this especially powerful is the way that Ray unifies the actor abstraction with the task-parallel abstraction inheriting the benefits of both approaches. Ray uses an underlying dynamic task graph to implement both actors and stateless tasks in the same framework. As a consequence, these two abstractions are completely interoperable.
I have a faint suspicion that even needing to know what tasks and actors are is a red flag for my role as a researcher who is being paid not to care about implementation details.
There is a commerical spinoff called Anyscale that … builds a ray-compliant cloud for you? idk
MLFlow is a platform to streamline machine learning development, including tracking experiments, packaging code into reproducible runs, and sharing and deploying models. MLflow offers a set of lightweight APIs that can be used with any existing machine learning application or library (TensorFlow, PyTorch, XGBoost, etc), wherever you currently run ML code (e.g. in notebooks, standalone applications or the cloud). MLflow’s current components are:
- MLflow Tracking: An API to log parameters, code, and results in machine learning experiments and compare them using an interactive UI.
- MLflow Projects: A code packaging format for reproducible runs using Conda and Docker, so you can share your ML code with others.
- MLflow Models: A model packaging format and tools that let you easily deploy the same model (from any ML library) to batch and real-time scoring on platforms such as Docker, Apache Spark, Azure ML and AWS SageMaker.
- MLflow Model Registry: A centralized model store, set of APIs, and UI, to collaboratively manage the full lifecycle of MLflow Models.
Nextflow is built around the idea that Linux is the lingua franca of data science.
Nextflow enables scalable and reproducible scientific workflows using software containers. It allows the adaptation of pipelines written in the most common scripting languages.
Nextflow allows you to write a computational pipeline by making it simpler to put together many different tasks.
You may reuse your existing scripts and tools and you don’t need to learn a new language or API to start using it.
Nextflow supports Docker and Singularity containers technology.
This, along with the integration of the GitHub code sharing platform, allows you to write self-contained pipelines, manage versions and to rapidly reproduce any former configuration.
Nextflow provides an abstraction layer between your pipeline’s logic and the execution layer, so that it can be executed on multiple platforms without it changing.
It provides out of the box executors for SGE, LSF, SLURM, PBS and HTCondor batch schedulers and for Kubernetes, Amazon AWS and Google Cloud platforms.
From the introduction:
With data now being a primary asset for companies, executing large-scale compute jobs is critical to the business, but problematic from an operational standpoint. Scaling, monitoring, and managing compute clusters becomes a burden on each product team, slowing down iteration and subsequently product innovation. Moreover, these workflows often have complex data dependencies. Without platform abstraction, dependency management becomes untenable and makes collaboration and reuse across teams impossible.
Flyte’s mission is to increase development velocity for machine learning and data processing by abstracting this overhead. We make reliable, scalable, orchestrated compute a solved problem, allowing teams to focus on business logic rather than machines. Furthermore, we enable sharing and reuse across tenants so a problem only needs to be solved once. This is increasingly important as the lines between data and machine learning converge, including the roles of those who work on them.
All Flyte tasks and workflows have strongly typed inputs and outputs. This makes it possible to parameterize your workflows, have rich data lineage, and use cached versions of pre-computed artifacts. If, for example, you’re doing hyperparameter optimization, you can easily invoke different parameters with each run. Additionally, if the run invokes a task that was already computed from a previous execution, Flyte will smartly use the cached output, saving both time and money.
airflow is Airbnb’s hybrid parallel-happy workflow tool:
While you can get up and running with Airflow in just a few commands, the complete architecture has the following components:
The job definitions, in source control.
A rich CLI (command line interface) to test, run, backfill, describe and clear parts of your DAGs.
A web application, to explore your DAGs’ definition, their dependencies, progress, metadata and logs. The web server is packaged with Airflow and is built on top of the Flask Python web framework.
A metadata repository, typically a MySQL or Postgres database that Airflow uses to keep track of task job statuses and other persistent information.
An array of workers, running the jobs task instances in a distributed fashion.
Scheduler processes, that fire up the task instances that are ready to run.
SystemML created by IBM, now run by the Apache Foundation:
SystemML provides declarative large-scale machine learning (ML) that aims at flexible specification of ML algorithms and automatic generation of hybrid runtime plans ranging from single-node, in-memory computations, to distributed computations on Apache Hadoop and Apache Spark.
Tensorflow is the
hot new Google one, coming from the
training of artificial neural networks,
but having sprouted many cloud computation affordances.
See my notes
Listed here as a parallel option because it has some parallel support,
especially on Google’s own infrastructure. C++/Python.
Dataflow/ Beam is google’s job handling as used in google cloud (see below), but they open-sourced this bit. Comparison with spark. The claim has been made that Beam subsumes Spark. It’s not clear how easy it is to make a cluster of these things, but it can also use Apache Spark for processing. Here’s an example using docker. Java only, for now. You can maybe plug it in to Tensorflow? 🤷♂
Turi (formerly Dato (formally Graphlab)) claims to automate much cloud stuff, using their own libraries, which are a little… opaque. (update - recently they opensourced a bunch so perhaps this has changed?) They have similar, but not identical, APIs to python’s scikit-learn. Their (open source) Tensorflow competitor mxnet claims to be the fastest thingy ever times whatever you said plus one.
…an open-source cluster computing framework originally developed in the AMPLab at UC Berkeley. […] By allowing user programs to load data into a cluster’s memory and query it repeatedly, Spark is well suited to machine learning algorithms.
Spark requires a cluster manager and a distributed storage system. […] Spark also supports a pseudo-distributed mode, usually used only for development or testing purposes, where distributed storage is not required and the local file system can be used instead; in the scenario, Spark is running on a single machine with one worker per CPU core.
Spark had over 465 contributors in 2014, making it the most active project in the Apache Software Foundation and among Big Data open source projects.
Sounds lavishly well-endowed with a community. Not an option for me, since the supercomputer I use has its own weird proprietary job management system. Although possibly I could set up temporary multiprocess clusters on our cluster? It does support python, for example.
- Standalone — a simple cluster manager included with Spark that makes it easy to set up a private cluster.
- Apache Mesos — a general cluster manager that can also run Hadoop MapReduce and service applications.
- Hadoop YARN — the resource manager in Hadoop 2.
- basic EC2 launch scripts for free, btw
Interesting application: Communication-Efficient Distributed Dual Coordinate Ascent (CoCoA):
By leveraging the primal-dual structure of these optimization problems, COCOA effectively combines partial results from local computation while avoiding conflict with updates simultaneously computed on other machines. In each round, COCOA employs steps of an arbitrary dual optimization method on the local data on each machine, in parallel. A single update vector is then communicated to the master node.
i.e. cunning optimisation stunts to do efficient distribution of optimisation problems over various machines.
Spark integrates with scikit-learn via spark-sklearn:
This package distributes simple tasks like grid-search cross-validation. It does not distribute individual learning algorithms (unlike Spark MLlib).
i.e. this one is for when your data fits in memory, but the optimisation over hyperparameters needs to be parallelised.
Thunder does time series and image analysis - seems to be a python/spark gig specialising in certain operations but at scale:
Spatial and temporal data is all around us, whether images from satellites or time series from electronic or biological sensors. These kinds of data are also the bread and butter of neuroscience. Almost all raw neural data consists of electrophysiological time series, or time-varying images of fluorescence or resonance.
Thunder is a library for analyzing large spatial and temporal data. Its core components are:
Methods for loading distributed collections of images and time series data
Data structures and methods for working with these data types
Analysis methods for extracting patterns from these data
It is built on top of the Spark distributed computing platform.
Not clear how easy it is to extend to more advanced operations.
dispy is a comprehensive, yet easy to use framework for creating and using compute clusters to execute computations in parallel across multiple processors in a single machine (SMP), among many machines in a cluster, grid or cloud. dispy is well suited for data parallel (SIMD) paradigm where a computation (Python function or standalone program) is evaluated with different (large) datasets independently with no communication among computation tasks (except for computation tasks sending Provisional/Intermediate Results or Transferring Files to the client
dask seems to parallelize certain python tasks well and claims to scale up elastically. It’s purely for python.
The Tessera computational environment is powered by a statistical approach, Divide and Recombine. At the front end, the analyst programs in R. At the back end is a distributed parallel computational environment such as Hadoop. In between are three Tessera packages: datadr, Trelliscope, and RHIPE. These packages enable the data scientist to communicate with the back end with simple R commands.
Thrill is a C++ framework for distributed Big Data batch computations on a cluster of machines. It is currently being designed and developed as a research project at Karlsruhe Institute of Technology and is in early testing.
Some of the main goals for the design are:
To create a high-performance Big Data batch processing framework.
Expose a powerful C++ user interface, that is efficiently tied to the framework’s internals. The interface supports the Map/Reduce paradigm, but also versatile “dataflow graph” style computations like Apache Spark or Apache Flink with host language control flow. […]
Leverage newest C++11 and C++14 features like lambda functions and auto types to make writing user programs easy and convenient.
Enable compilation of binary programs with full compile-time optimization runnable directly on hardware without a virtual machine interpreter. Exploit cache effects due to less indirections than in Java and other languages. Save energy and money by reducing computation overhead.
Due to the zero-overhead concept of C++, enable applications to process small datatypes efficiently with no overhead.
Dataiku is some kind of enterprise-happy exploratory data analytics swiss army knife thing that deploys jobs to other people’s clouds for you. Pricing and actual feature set unclear because someone let the breathless marketing people do word salad infographic fingerpainting all over the website.
Computation node suppliers
A place of terror and dismay, a mysterious digital onslaught, into which we all quietly moved.
A fictitious place where dreams are stored. Once believed to be free and nebulous, now colonized and managed by monsters. See ‘Castle in the Air’. […]
[…] other peoples’ computers
via Bryan Alexander’s Devil’s Dictionary of educational computing
See also Julia Carrie Wong and Matthew Cantor’s devil’s dictionary of Silicon Valley
cloud, the (n) — Servers. A way to keep more of your data off your computer and in the hands of big tech, where it can be monetized in ways you don’t understand but may have agreed to when you clicked on the Terms of Service. Usually located in a city or town whose elected officials exchanged tens of millions of dollars in tax breaks for seven full-time security guard jobs.
If you want a GPU this all becomes incredibly tedious. Anyway…
Vast.ai allows everyone else who overinvested in buying GPUs during the bitcoin boom to sell their excess GPU time to make back the cash. Looks fiddly to use but also cheap.
Floydhub do deep-learning-oriented cloud stuff and come with an easy CLI.
Microsoft Azure - haven’t really tried it but presumably Microsoft are good at the computers?
Paperspace is a node supplier specialising in GPU/machine learning ease.
Amazon has stepped up the ease of doing this recently. It’s still over-engineered for people who aren’t building the next instagram or whatever. See my Amazon Cloud notes
Google cloud might interoperate well with a bunch of google products, such as Tensorflow, although it has weirdnesses like relying on esoteric google APIs so hard to prototype offline or with awful internet. See my google cloud notes.
Turi is also in this business, I think? I’ve gotten confused by all their varied ventures and offerings over many renames and pivots. I’m sure they are perfectly lovely.
RunwayML is a platform for creators of all kinds to use machine learning tools in intuitive ways without any coding experience. Find resources here to start creating with RunwayML quickly.
Parallel tasks on your awful ancient “High Performance” computing cluster that you hate but your campus spent lots of money on and it IS free so uh…
See HPC hell.
Local parallel tasks with python
See also the overlapping section on build tools for some other pipelining tool with less concurrency focus.
joblib is a simple python scientific computing library with basis mapreduce and some nice caching that integrate well. Not fancy, but super easy, which is what an academic usually wants, since fancy would imply we have a personnel budget.
>>> from math import sqrt >>> from joblib import Parallel, delayed >>> Parallel(n_jobs=2)(delayed(sqrt)(i ** 2) for i in range(10)) [0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0]
pathos is one general tool here. Looks a little… sedate… in development. Looks more powerful than joblib in principle, but joblib actually ships.
You could also launch spark jobs.
(To work out - should I be listing Docker container images instead? Much hipper, seems less tedious.)
Singularity promises containerized science infrastructure.
Singularity provides a single universal on-ramp from the laptop, to HPC, to cloud.
Users of singularity can build applications on their desktops and run hundreds or thousands of instances—without change—on any public cloud.
- Support for data-intensive workloads—The elegance of Singularity’s architecture bridges the gap between HPC and AI, deep learning/machine learning, and predictive analytics.
- A secure, single-file-based container format—Cryptographic signatures ensure trusted, reproducible, and validated software environments during runtime and at rest.
- Extreme mobility—Use standard file and object copy tools to transport, share, or distribute a Singularity container. Any endpoint with Singularity installed can run the container.
- Compatibility—Designed to support complex architectures and workflows, Singularity is easily adaptable to almost any environment.
- Simplicity—If you can use Linux®, you can use Singularity.
- Security—Singularity blocks privilege escalation inside containers by using an immutable single-file container format that can be cryptographically signed and verified.
- User groups—Join the knowledgeable communities via GitHub, Google Groups, or in the Slack community channel.
- Enterprise-grade features—Leverage SingularityPRO’s Container Library, Remote Builder, and expanded ecosystem of resources. […]
Released in 2016, Singularity is an open source-based container platform designed for scientific and high-performance computing (HPC) environments. Used by more than 25,000 top academic, government, and enterprise users, Singularity is installed on more than 3 million cores and trusted to run over a million jobs each day.
In addition to enabling greater control over the IT environment, Singularity also supports Bring Your Own Environment (BYOE)—where entire Singularity environments can be transported between computational resources (e.g., users’ PCs) with reproducibility.
It is less fancy, but NVIDIA provides GPU images with all their crap pre-downloaded and licences pre-approved.