What type of Amazon EC2 server for MathWorks Cloud Center

4 ビュー (過去 30 日間)
Salina13
Salina13 2021 年 7 月 9 日
コメント済み: Salina13 2021 年 7 月 10 日
I want to start using parallel computing to speed up my code. I already tried it using my local computer (a Mac 2 cores), with great result. However, I would like to further reduce the running time. Would Matlab cloud center be a good solution for this? And what type of Amazon EC2 server would be good to start with? I have never used this before, so any suggestions are welcome. Moreover, I have a free matlab account provided by my university. If I use Matlab cloud center, do I need to pay for the use of matlab on the EC2 server?
  5 件のコメント
Walter Roberson
Walter Roberson 2021 年 7 月 9 日
Do you have a sample data set that can be used to time your code? And can you make your code available? If the answer to these is yes, then bundle those together and drop it on Google Drive, and use my Contact information to send me the link; I will run it on my new iMac (10 core, i9) so you would get an idea of how performance improves.
[Hmmm, I guess I should put the rest of the memory in my iMac !]
Salina13
Salina13 2021 年 7 月 10 日
Unfortunately I cannot make my code available to you. But I appreciate the offer. Thank you

サインインしてコメントする。

回答 (1 件)

Raymond Norris
Raymond Norris 2021 年 7 月 9 日
Just to clarify Walter's 2nd comment, he's suggesting a compute cluster. What you're saying is that your university does not have an HPC cluster/data center, correct? If it does, that could be an easier approach.
To use MathWorks Cloud Center, someone will need to start a cluster using an AWS account, which needs to be paid for by someone (you, the university, etc.). The MATLAB Parallel Server license may already be available for you. Best to contact Technical Support (support@mathworks.com) to find out what you are entitled to run and where.
Regarding the best EC2 instance, without knowing anything about your code (how long it takes to run, memory required, CPU/GPU, etc.), it's impossible to make a proper recommendation. My suggestion would be to initially start small, get accustomed to running jobs on AWS and then scale over time once you get the hang of things (again, someone will need to pay for the compute time even if the licenses are already available). Overtime, you'll want to familiarize yourself with the different supported instances, but I would suggest the m5.xlarge to begin with. It's got 2 cores, which helps compare it to the results you got from your Mac. If all goes well, then increase the core count to an appropriate number (i.e. you might get diminishing results with a 32 or 48 core machine).

カテゴリ

Help Center および File ExchangeMathWorks Cloud Center についてさらに検索

製品


リリース

R2020a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by