setup desktop computer to run stata remotely?

[Claire Wladis]

I travel a lot and so have always owned a laptop, and run Stata and other statistical software on the laptop. I always buy the most powerful laptop that I can (SSD harddrive, maximum RAM, maximum processor) that isn't to heavy to carry around, but obviously a laptop is a lot more limited in processor and RAM capacity than a desktop. I do a lot of work with rather large datasets. For example, right now I have a job that involves several hundred thousand data points, where I am doing three-way interactions on multi-level logistic models--the job has been running for days. I'm reaching the time where I need to buy a new computer, and I have been torn about purchasing a laptop that is more lightweight, versus one with more powerful specs--but then it occurred to me that I could perhaps purchase a more powerful desktop and set it up so that I can run stata jobs on it remotely in some way. That way I could limit stata jobs on the laptop to simpler ones, and these huge ones I could send remotely to the desktop (and that way I wouldn't have to worry so much about the specs on my laptop). I currently have an MP license with 4 cores. While I don't have an unlimited budget by any means, I do currently have funding allocated for these purchases, and so cost is less important for me than making my workflow more efficient and speedy (up to a point of course).

So I guess my questions are these:

I'm just wondering if anyone else has experience with this, and if they would mind sharing their expertise with me about how well it worked and whether they felt it was worth it?

Are there any basic references (books or websites) that would outline, for a computer-savvy-by-average-person-standards but not-at-all-computer-savvy-by-computer-tech-standards person like me, how I might go about setting this up, and what the limitations might be?

I'd also welcome any other advice that comes to mind, from those of you who are more experienced at this kind of thing than I am!

[William Lisowski]

I think some readers will want to be convinced that the (somewhat complex) setup you envision to run jobs remotely is necessary, vis-a-vis possible gains from a more powerful laptop. Others will want to know what operating system you are running under (I'm guessing some flavor of Windows), since your question depends very much on that information.

Please tell us as much as you can about the characteristics of your current laptop.

[wbuchanan]

Some of the remote access issue comes down to the ISP that you have for your internet access. If the ISP assigns IP addresses dynamically and provides no DNS services that you could use to reference your desktop, it wouldn't be useful/helpful since you wouldn't have a consistent method to connect back to the desktop. If you're still planning to use the laptop for some of your work, it might be worth creating an image with Stata installed that you can use with one of Amazon's servers. Then, when you need to run a job for an extended period of time, you can spin up an instance and let their servers handle that part of the workload while you move on to other things. Unless you were working at home primarily, you'd almost definitely be able to get better quality hardware by purchasing time on someone else's servers vs trying to build your own.

[Richard Williams]

I have Win 7 on my desktop. I use Remote Desktop Connection to access it when away from home. We pay extra for a static ip so i can easily access it. I don't know if this is an optimal setup but it meets my modest needs. If you don't have a static ip I think there may be apps that can tell you what the currently assigned ip is.

[Marcos Almeida]

Hi Claire,

I have no experience in using Stata the way you want. But I have several years of experience in remote access and I hope these comments might be useful to you. In fact, I'm in charge of making several kinds of medical reports, and these reports depend upon two computers located in different institutions. I need to do some travelling, attend congresses and the likes, most of them far away from my hometown. Then, I decided to install a software of remote access, leave the computers ON on a 24/7 basis, and this way I avoided any gaps in the preparation of the reports. No matter the specific software for the reports, the remote access worked wonders.

In short, after some browsing, I found two types of remote access which could fulfill my needs. In the first, we need someone in the main computer so as to allow access from the Web. In the second, provided the computer is turned ON, you may access from wherever you want, be it from your personal notebook (my favorite way), be it from a public PC. It is quite safe, for we need to type to codes, one to log on to the provider of remote access, and one to log on the main computer. Nowadays I'm quite satisfied with Log Me In to perform this task. The sole pitfall I shall remark: time to make a regular report naturally depends on the speed of the Web access but it generally increases around a two to threefold in my case. That said, if you wish to type a command and get the results after a long time,I mean, if you don't need to type, analyze, edit, decide, re-edit, write, check and re-check like me in order to finalize a report, time won't be an issue, because it will basically depend on the speed of the main computer you're already used to work with. In fact, it is like you were in the very same room your computer is located.

Hopefully that helps.

Best,

Marcos

[Peter Franks]

I use Teamviewer 11 to connect all the computers I use---except it cannot penetrate the Corporate System at work (for which I use the IT approved version of ciscoconnect +remote desktop--- the latter is a little fiddly ---since the deafault settings on the work computer rpeclude oustside connections (thus i have to get IT involved to change these settings)). TeamViewer is a great and easy to use program (and free for non-commercial use)---- its fast and you basically get presented with the DEsktop of the remote computer you are connected to. In other words--- if you set up your monster (RAM/CPU) computer at home with stata installed on it--- you can from any remote site computer access the desktop of the home computer and run stata--- programs and output can be easliy moved around. If your monster computer is inside a corporate firewall--- then you need to find out local rules about remote access--- it is likely possible, but more of a hassle, typically. I do not make use of a static ip--in my setup which has the monster located outside the corporate firewall-- and the whole thing runs flawlessly using TeamViewer. Incidentally, I use a central program/output log system--- which is on google drive-- so it is directly accessed by all the computers I use--- as long as I dont try and run 2 versions of stata at once then everything is nicely synced together. Happy to provide more details if that would help.

[Claire Wladis]

Thank you everyone for your helpful posts. William, I am currently completely on Windows, but have been considering switching to a Mac for my next computer, because I strongly prefer that operating system. But I haven't made up my mind--I don't love that mac laptops can't be upgraded later and I do worry that, even though I have used a mac for a few months recently, there might be differences that I'm not aware of that might bother me after switching. Since I have to lug my laptop around (frequently, but not every day), I prefer something at least somewhat lightweight, which tends to limit options for processors, RAM, etc. My current laptop has an Intel i7 2.7 GHz dual-core processor, 8 GB RAM, and a 512 GB SSD, running Windows 7. It is a bit over 4.5 years old.

wbuchanan, thanks for the information about the ISP issue and the recommendation about amazon's servers. I am going to research this possibility to see if this would make sense. I'm not sure about the ISP setup--ideally I would put the desktop at work, but it would have to be on the wireless network there in order for me to have any control over it, and I don't know anything about those settings--I will have to ask.

Richard, thanks for the recommendation and explanation of your particular setup. I will look into whether Remote Desktop Connection might also work for us.

Marcos, thanks for your detailed answer--this is also very helpful. Yes, I had thought about the limitations of the internet connection, but as you explain, since I would mainly be sending large jobs to the desktop and then just picking up the results once done, it seems that this probably wouldn't be an issue. I will also look into Log Me In.

Peter, thanks for your description of your setup and for the recommendation of Teamviewer--this setup sounds like it would work well for us. I would love to hear more about your central program/output log system, if you don't mind sharing.

Thanks again to everyone for the great detailed advice--I have a lot to look into and to think about--your advice has been really helpful in helping me to figure out how to make this decision!

[wbuchanan]

Claire Wladis if you're going to be setting things up at work, connect with the IT folks about virtual private networks (VPN) and/or virtual environments that they could provide. While I was working in Mississippi, I did a lot of work on a virtual Ubuntu server from my Mac and was able to open up a terminal and connect to the server over ssh any time I wanted, when connected to the network. As a grad student at Brown, access to Stata was managed through a VPN; I could install Stata anywhere I wanted, but in order to use it I had to connect to the VPN so a license key could be allocated. There are likely similar solutions, but if you can get the IT guys to give you some computing resources on their servers it may be an even more powerful solution.

Regarding the Mac vs Windoze thing (see disclaimer below), I wouldn't worry about the upgrade capabilities of the newer MacBook Pro line. One of the reasons behind the design was to increase performance overall. So, instead of having to transfer data through a SCSI connection or something like that, the transfer instead is from one part of the board to the CPU. I'll admit that I was also a bit apprehensive about not having the ability to throw new hardware into the case when it became available, but I've had my new MacBook Pro for a couple of years now and it is an impressive piece of hardware. That said, if you really want power, I'd recommend checking out System76 laptops. They build laptops that run Ubuntu (a Linux distribution with a pretty intuitive user interface and fairly familiar functionality/features) and are able to build laptops that would basically be the equivalent of a server sitting on your lap. For example, their ServalWS line can be configured with 64GB of RAM, up to 5TB of hard disk, and choices of the newest Intel processors. The biggest concern I would have about Mac's moving forward is more about the changes they've made to the operating system with the most recent OS release (OSX El Capitan) that lock the owner/user out of specific directories on the system; if you're not doing much programming it wouldn't really be a problem, but if you've ever created a symlink in /usr/bin, it could cause some challenges (and is one of the reasons that I've not upgraded from Yosemite yet). If you're familiar with the MS platform already it would certainly help reduce any time it would take to get up and running with things. The one thing that has started to bother me even more about MS, is that they force the OS to use an encrypted disk and then proceed to store your secret key on their servers (http://www.theregister.co.uk/2015/12...cryption_keys/). I could see this being potentially problematic if you're working with sensitive data, since any breeches to MS's servers could compromise the data on an individual's machine.

Disclaimer: I have a fairly passionate disliking for MS products so I'll always favor *nix-based systems; figured it might be good to disclose this.