You asked, we will deliver (or explain why not)
You asked, we will deliver (or explain why not)
First of all, thank you for replying to our customer survey! We received over 250 replies, with most of our customer segments well represented, giving very valuable feedback for how to improve our service and software portfolio to meet the needs of today's scientific research.
We are at the moment planning how to develop our services accordingly and picking "low-hanging fruits" (like installing missing software packages) on the go. We would like to put emphasis on some of the points brought up in the survey, and give our responses and suggestions on how to mitigate and overcome the observed shortcomings.
True, not enough capacity
In many responses being short of some particular resource (e.g. large-memory nodes, GPUs, or CPU cores in general) was pointed out. We are painfully aware of this. The current generation of systems was installed in 2012-2014 and thus clearly out of its competitive lifetime. Luckily, and as you have hopefully heard before, we are currently renewing the whole computing and storage ecosystem at CSC under the funding and development program DL2021. The call for tenders is out at the moment, and we should know during the autumn timeframe the exact specs of the upcoming systems. The users will have the first major chunk of new capacity (which will be Taito-like) in their disposal in early 2019. So please try to bear with us and the current situation for a couple of months more. The new systems will be worth waiting for.
You can always ask for software installations
You can see the list of installed software at https://research.csc.fi/software which lists the softwares that are actively maintained in the CSC servers. However, especially in Taito there are many tools and utilities that are installed even though CSC is not actively supporting them.
If you can’t find the software (or version) you want to use in the software list, you can use the module spider command to check if there is a module for the tool you wish to use, e.g.
module spider gatk
Still no luck? Feel free to ask for the installation from CSC Servicedesk (email@example.com). Alternatively, you can try to install the missing tool yourself to Taito. CSC can support you with the installation if facing problems.
Please meet Julia
There were several requests for providing Julia programming environment. Julia is an emerging high-level but high-performance programming language and framework. We are happy to announce that Julia is now available on Taito as a high-performance version, just load the module "julia" and go. Give also a try for the Jupyter notebooks version available at notebooks.csc.fi. The Julia programming environment is supported hereafter in a similar fashion as we develop our R and Python environments.
You can invite us over
There were couple suggestions of CSC staff visiting you local campuses. We are always eager to meet our customers and happy to tell about our services. If you have a group seminar, faculty day or some training that might benefit from a presentation and/or Q&A session of our services, feel free to send us e-mail (firstname.lastname@example.org) and invite us over!
We have updated our material archives
There were lots of wishes for online learning materials, so we gathered them for your convenience. You can search for CSC course materials on other topics from our event archive (hint: the tag cloud helps! Just click “show keywords”). Some topical materials are also gathered in the Materials site.
We are also continuing our webinar series. We record the webinars for later viewing as well. You can find the list of past and upcoming webinars as well as the links to the recordings here: https://research.csc.fi/csc-webinars. Note that you can also suggest new webinar topics using the form there!
Running long jobs
On Taito, the maximum runtime is two weeks (14 days). In cases where longer jobs than this are needed, the first thing to do is to check if the task can be split into several shorter tasks. For instance, even if all the steps of the workflow could be executed as one batch job, however for (too) long jobs you need to run different steps as separate batch jobs. Not trying to do everything at once also helps you in optimizing the resource requests.
In the rare case when the job cannot be split into shorter sub-tasks, then the recommended option is to install your own version of the software to a Virtual Machine running in cPouta. We prefer helping you to set up your own virtual machine for this purpose rather than extend the execution time of your batch job.
Wanted: R and Python courses
Many of you wished for R and/or Python courses, and we are happy to deliver! We have been offering general beginner level courses, as well as some more topical courses for visualisation and data handling, but now we would like to ask you what you really want to achieve with R and/or Python to see whether we can tune our courses to meet those learning goals of yours.
Please, especially if you are planning on joining one of our future courses, take a moment and answer these further questions: https://www.webropolsurveys.com/S/9DCF1D5A391072BD.par
Your voice has been heard. In case of further ideas or comments, please do contact us (directly, via CSC Servicedesk, comment below, when meeting our staff in person,...).
Julkaistu alunperin 29.5.2018.
The author is an application specialist in biosciences, working at customer support and service development at CSC.
maria.lehtivaara (at) csc.fi
Dr Pekka Manninen is a supercomputer geek and a program director in charge of the EuroHPC pre-exascale service LUMI. Twitter:@pekkamanninen