Looking at the opportunity to write a quick proposal for equipment to be used in a new DH lab (space for 6, max 10 workstations) and thinking about generic needs of future colleagues and graduate student researchers. The quandry -- we lack information about specific project needs which might emerge. Any recommendations on the foundation or baseline for such a lab?
Tech Foundation for a DH Center Lab
(5 posts) (3 voices)-
Posted 1 year ago Permalink
-
Without a greater sense of the types of projects, it'll be a little hard to make decisions. If there's an emphasis on media production, this media lab at UMW might be a good model, or at least give some example ideas.
Other software that might be fairly common for DH work:
* <oXygen/> XML editor
* R statistics package
* MALLET topic modeling
* Some kind of code editor
* Some FTP applicationWhich of those would actually be used depends a lot on specific types of projects. No matter what choices you make, though, you'll be implicitly guiding what kinds of DH work you are able to facilitate. Those would probably cover a good range of interests, though.
Posted 1 year ago Permalink -
Arno Bosse
MemberIf the lab is to focus on DH as text analysis then you might want to reconsider buying 6-10 "workstations" (in the sense of high-end PCs) at all. Instead, purchase much cheaper Mini PCs. These will be more than up to the task. Then again, you may not want to deal with creating individual user accounts on these machines or else, cleaning-up the user-directories or working with your IT group to mount shared home directories. I certainly wouldn't.. So If you're really going to be focusing on text anyway, why not do something more radical and purchase a series of ultra cheap Raspberry Pi2s instead., attach these to some big, cheap monitors and give each student a fast 16GB SD card at the start of the class. A learning moment, if you will. The new Raspberry Pi 2 has quad-core CPU, 1GB RAM, runs Ubuntu and even comes bundled with Mathematica v10. It'll be more than capable enough for R, MALLET and the like.
Then, since DH is obviously not just about texts, take some of all the money you've saved and buy just one or two well-equipped PCs or iMacs. Two should be more than enough since you've already got 10 or more MiniPCs or Raspberry Pi computers making up the bulk of the lab. And perhaps, I would not even buy these two workstations, I would lease them on a 3 year replacement cycle. Otherwise, come 3-4-5 years from now, someone is going to have to go back to the Dean or whomever and ask for another one-off capital investment. Instead, if your institution is also kicking in some funding, return (!) some of this capital funding you have now and ask for a raise in your operating budget instead (something most grants will not be able to provide). Alternatively, hold onto some of the money you saved by not purchasing workstations and plan/collaborate with other departments on purchasing various kinds of input devices for use in an experimental setting, such as a portable audio recorder, a basic scanner, a cheap video camera, maybe a portable (GoPro-like) camera or a cheap (even "toy") 3D scanner. The general idea being that students should not be exposed just to the (comparatively) clean world of texts but also to noisier, dirtier, computationally less tractable "real-world" A/D sources.
Elsewhere on DHAnswers someone made a comment that a lot of time was spent in their DH lab thinking about furniture (tables, table layouts etc). I think this is very good advice. A place in your lab (e.g. a long, tall table) where students can plug-in devices, place their books or tools, carry over a Raspberry Pi, store artifacts, and build and hack at things together in a group would be great, don't you think? Heck, I'd even put it in the center of the room.
Patrick made a very good point in his comment above, "No matter what choices you make, though, you'll be implicitly guiding what kinds of DH work you are able to facilitate." He's right. Perhaps, as a practical response to that, plan and spend 75% (?) of your budget on what you know you want to do with the students. And leave 25% for cheap HW investments in experiments, risks and whatever might conceivably stick to the wall.. and mitigate the risk & difficulty of deciding what this might be by drawing in other units such as special collections, theatre, media studies, art history, studio art, linguistics, local/campus museums/archives etc.
Posted 1 year ago Permalink -
+ several million for Arno's idea of leaving 25% for experiments and risks. Not only does that reflect the reality of doing DH, and being able to respond to student and faculty needs as they emerge quickly. If the administration is willing to build that in, that's a great signal of their real understanding of and commitment to what a DH lab should be. If you can get that, you'll have a great DH lab!
Posted 1 year ago Permalink -
Arno Bosse
MemberSomething else occurred to me. It's just a small thing but might have a impact down the road depending on your needs. Let's say you're also interested in machine learning, particularly image recognition, where training a model can be computationally taxing. I've noticed that while many toolkits (e.g. Caffe) offer (or even depend) on GPU based optimizations they invariably only support CUDA, which means in effect an nVidia graphics card. However most current Macs and also many older models and of course regular PCs too are sold with video cards from AMD/ATI. So then you'd be stuck or at minimum would need to pony up for another card(s). Consequently, probably a good case could be made for getting one reasonably fast, headless, vanilla PC with lots of RAM and dual video cards to use as a computation server. For example, as mentioned, for machine learning experiments, but also as a place to run MATLAB or Mathematica etc.
Update: A few days later I came across this very useful report on exactly this topic: https://timdettmers.wordpress.com/2014/08/14/which-gpu-for-deep-learning/
Posted 1 year ago Permalink
Reply
You must log in to post.
