Forum:Macosx For Bioinformatics?
11
1
Entering edit mode
11.0 years ago

I am planning computing needs for our group for the 2-3 years to come and came across a difficult decision when having to decide which kind of computer to choose.

I am thinking of a system with ~100Go of RAM, Xeon class processors or equivalent (up to 16 threads). The thing should be able to boot under both Windows and Linux. I could go for a Lenovo ThinkStation or something of the kind, but I am evaluating the potential of getting a Mac Pro.

EDIT: Depending on how we do things, I may end up buying two computers. One computer would be used by lab members to perform routine analyses that they cannot run on their laptops and for which the standard desktops that we have available can't do the job because of too low RAM or processing power. As a result, the easiness of the user interface has some importance. The other would be used by me for my daily work.

I'm not thinking about high end servers, rather something to serve as middle-ground between student laptops and weak lab desktops on one hand and servers and super-computers on the other end.

DISCLAIMER: I am a Linux lover and am very sceptical about Macs. On the other side, I am also very open to seeing what advantages there are to Macs.

Here is where you come in. I would really like to have your opinion about the following:

  1. Are you using a Mac as a main computer for bioinformatic work?
  2. What positive points do you see about using a mac for bioinformatics?
  3. What negative points do you see about using a mac for bioinformatics?
  4. Ease of installing software (talking about bioinfo software here)?
  5. Freedom to hack quick solutions for custom problems?
  6. Whatever opinion you want to share about this question
  7. Why should I (or not) think of getting Macs for our lab?

My goal is definitely NOT to turn this into a flame war. It would be nice if you could keep it about facts.

(Please also go to What Operating System Are You Using For Bioinformatic Work? and vote to say what operating system you are using for bioinformatics)

linux • 20k views
ADD COMMENT
0
Entering edit mode

I think some people are confused about what you want an answer for.

Are you asking whether or not Macs are a good buy for "middle tier" workstations that users will have on their desktops? Or are you asking if a mac is a good choice for a (single) compute server that will service the needs of the people in your lab when they don't want (can't) get access to a larger (presumably linux) compute cluster?

I think you are asking for recommendations on the latter, but it seems like some answers are addressing the former.

ADD REPLY
0
Entering edit mode

Hi Steve, the question could have been more precise and I added an EDIT portion to clarify regarding your comment. At the moment of writing the question, I was considering one or two computers that would fill these needs 1) it would be used by lab members so that they are more independent in their analyses 2) I would use that computer daily for my work.

ADD REPLY
0
Entering edit mode

One thing which I don't think anyone has commented on yet is the higher cost of a Mac and future OSX updates. Would you want to exchange the costs of more compute power for a more slick environment?

ADD REPLY
1
Entering edit mode

For what it's worth, it looks as if future OSX updates will be free just since the new OS (10.9) is a free update from many previous versions (I believe as far back as 10.6). That having been said, in recent years, upgrades have only cost ~ $20, which is not onerous by any stretch of the imagination.

ADD REPLY
0
Entering edit mode

See my reply to Steve's comment. In case 1, maybe I would consider a slicker environment for the many potential users (most of which have Mac laptops). For case 2, where I am the sole user, I would have a tendency to go for a more powerful penguin rather than a comparatively weaker apple for the same price.

ADD REPLY
0
Entering edit mode

If it is a multi-user system the interface shouldn't matter. It is probably best to set it up so that users remote connect to their accounts and run jobs in the same way they would a server or computer cluster, but probably without the job scheduler. Although that may be a good idea just for management depending on the number of users. Otherwise they should be trained in making sure there are no other big analyses running that theirs would interfere with, to talk to other lab members before running jobs that will run for a very long time, etc.

ADD REPLY
12
Entering edit mode
11.0 years ago
JC 13k

I used Mac for 5 years (IT only allowed Windows or Mac for researchers, I hate windows), it was the first time I was using Mac as my main OS. I found really easy to work everyday with it, but my Mac was used only to code, write papers and do som basic R plots/analysis. The real job was performed in several Linux clusters with high scalability, sometimes I prepare and tested my pipelines in the Mac, but never run a full work (occasionally it had a hard drive fail twice, TimeMachine was my salvation). If you are thinking in buying a high-end server for bioinformatics, you have no choice: Linux.

ADD COMMENT
1
Entering edit mode

I have to agree with this. I was a linux and windows user until 2 years agoand only used my macbook for travel and home entertainment. But, now I have switched over in the office and am very happy with my mac pro for office productivity, analysis and interpretation work (IGV, browsers, etc) R coding, some light software/database development, web development. I also use it for simple tests of many informatics tools and for developing educational materials. Its great because it is a jack of all trades. But, for any kind of actual data processing (if you have NGS data) you will have to go onto a linux server. Installing bioinformatics software seems to have gotten dramatically better on mac in recent years. Basic tools like samtools, tuxedo suite, etc install without problems. But, you will still run into the occasional mac specific issues.

ADD REPLY
9
Entering edit mode
11.0 years ago
lh3 33k

I do not have a desktop. I do most things, including entertainment, daily work and coding of course, on my mac laptop. I think mac achieves a good balance between a sophisticated unix interface and ease of use. However, as you are using Mac as a workstation but not its user interface, you will not benefit from Mac. I would recommend Linux instead. Firstly, rpm/deb is still more convenient and more robust, IMO, than macports and homebrew. Secondly, a few tools are harder to compile and install on Mac. If someone have not ported such tools to Mac, you will have a hard time. Thirdly, I more like the development tools on Linux, most importantly valgrind and linux perf. For these reasons, I still have a linux in VM on my mac laptop.

ADD COMMENT
7
Entering edit mode
11.0 years ago

Are you using a Mac as a main computer for bioinformatic work?
For running tasks, I connect to a server which is not a mac. For local use where most jobs are not extremely intensive, I use a linux laptop. However, I can easily use my personal macbook pro and actually prefer it.

What positive points do you see about using a mac for bioinformatics?
Most linux packages should work on a mac. OSX is actually a pretty nice OS. There are software on osx that isn't on linux or hard to install on linux (Adobe illustrator for example).

What negative points do you see about using a mac for bioinformatics?
Installing packages can be a pain if you have never done it.

Ease of installing software (talking about bioinfo software here)?
Central repositories for installing stuff seems to be very transient. I think right now the popular option is using Homebrew. But it seems like new installation software comes up every couple of years.

Freedom to hack quick solutions for custom problems?
You can still run your python/perl/whatever scripts with no problem. So depends on what you mean by hacking I guess.

Whatever opinion you want to share about this question
I would not use a mac for your middle - heavy server needs. But I would recommend a mac as a portable laptop for light work loads.

Why should I (or not) think of getting Macs for our lab?
Don't get it as the server for your lab. I think it's only really good as a personal laptop. The Mac Pro might turn out great, but I would hold off on it and wait for more user testing.

ADD COMMENT
5
Entering edit mode
11.0 years ago
Carlos Borroto ★ 2.1k

I regularly use my Macbook Pro for running quick one-off tasks, as others mentioned for any serious work I still connect to a linux box.

4 . Ease of installing software (talking about bioinfo software here)?

Try Homebrew, it gives you easy access to a huge number of software. It now has a special section for scientific software. You gain access to this section by:

brew tap homebrew/science

For a complete list of software under this section see this.

Installing software then becomes as simple as:

brew install samtools

Finally contributing new "formulas" is very easy and well documented. Give it a try if the software you are interested is not there.

ADD COMMENT
4
Entering edit mode
11.0 years ago

Are you using a Mac as a main computer for bioinformatic work?

I use a few Macs for script and software development. I use the lab's Linux cluster for launching larger analyses.

What positive points do you see about using a mac for bioinformatics?

Unlike most handbuilt Linux desktop options, the hardware and software work well together, so I don't constantly need to be a sys admin like I did with Linux. Wireless networking just works. Plugging in a laptop to do presentations just works. Power management settings just work. The OS X operating system is less fragile to kernel and hardware changes and is easier to fix on my own, when such rare problems arise. OS X can run useful (if still commercial) visualization and productivity tools with easy access to a UNIX terminal, which Linux still cannot do (generally speaking). Many native GUI and command-line tools are available for doing source code control on OS X, so I don't lose any abilities there.

What negative points do you see about using a mac for bioinformatics?

A lot of open source gets distributed as RPM packages or similar, while package systems for OS X are a bit less friendly, in that regard. Some open source GUI apps are X11 based, which not only makes them look ugly but introduces a whole separate set of key shortcuts to learn, just for Linux-y apps. This also makes integration difficult — copying and pasting between native OS X apps is a nice feature, but is difficult between native apps and X11-ized apps.

If you're running a computational cluster, then Linux is a winner here, for the simple fact that the stripped-down hardware you need for that application is cheaper per node. You don't need a Mac Pro's super-duper graphics adapters, for instance, unless you are running custom analyses on fancy GPUs.

On the other hand, now that Thunderbolt shows up as an Ethernet adapter in Mavericks, I'd be interested to see if it gets used as a very fast interconnect for cluster work (a cheaper alternative to Infiniband, for instance).

Ease of installing software (talking about bioinfo software here)?

Depends on the software. Most of it can be compiled easily but you'll need to install Xcode, at a minimum, which used to be a PITA to download (used to need an Apple Developer Center account and navigate Apple's ADC site maze) but it is now pretty easy (OS X App Store + Apple ID + one or two clicks). Just about everything else that you can't compile on your own is something you can manage with macports or homebrew.

On the subject of macports and homebrew, pick one or the other, not both. They will step on each others toes as far as managing dependencies. I prefer macports, but I suspect it is largely a matter of how much you like (or dislike) Ruby, if you want to try homebrew.

Freedom to hack quick solutions for custom problems?

Decently modern versions of Ruby, Python, Perl etc. come pre-installed. So long as your sys admin has given your account admin rights on your Mac, you can compile and install whatever else you need through Apple's Xcode/clang, GNU's gcc and macports or homebrew.

Whatever opinion you want to share about this question

As a political matter, I love Linux's openness and customizability, but as a daily-use operating system, its interface just isn't very well engineered (just one person's opinion) and I suspect that no amount of catch-up work to imitate commercial operating systems will make that aspect of it as easy and reliable to use as OS X. Tools which other staff use for writing and managing papers (Word, Endnote, etc.) just aren't available, and open source and online "equivalents" do not quite make the grade, IMO. Managing document citations in a shared Google Doc is simply a non-starter, for instance. Of course, I'm less amenable to making compromises because I'm a pretty lazy human being. If your time is free and you love the OSS movement and you're willing to put in the work, you can do a lot of these things in Linux.

Why should I (or not) think of getting Macs for our lab?

I suspect it will reduce your administration costs, at the up-front expense of slightly more expensive hardware and the cost of commercial software. You'd need to balance that against the productivity gains from having your staff be able to collaborate together and share data more effectively. Compared with Linux, this is probably a plus mark in favor of OS X and Windows, generally speaking. On the other hand, you're going to spend a bit more up front. Six of one, half dozen of another. You'll need to decide what your priorities are with this investment, I suspect.

ADD COMMENT
0
Entering edit mode

"You'll need to decide what your priorities are with this investment, I suspect." I guess that's spot on. Thanks for the perspectives.

ADD REPLY
4
Entering edit mode
11.0 years ago
Adrian ▴ 700

It needn't be a dichotomy. Unless you are doing serious interactive graphics, video editing, etc, the Mac Pro, while it does look awesome, is probably overkill. Get a linux server with as much RAM and disk as you can, and get a few 27-inch iMacs with SSDs (and upgrade them to 32 gigs of RAM yourself.)

ADD COMMENT
3
Entering edit mode
11.0 years ago
amcrisan ▴ 370
  1. Yes I do, and I can use it for some heavier lifting as I have 32Gs of RAM and a new iMAC

  2. I like Pages or Word; I do use LaTex but not everyone I work with does, so when I need to use a simpler text editor I prefer Pages or Word to LibreOffice. Also, you can do non-analysis type stuff much easier on a Mac. It's got decent customer support.

  3. You will pay more to get less in terms of hardware . I don't like the inactive memory, it used to ballon out of control until I figure out how to manage it. Also, maybe because of how I use it, I find the OS takes up a fair bit of the memory and compute power because of all it's little optimizations it tries to do, and sometimes I find this annoying for parallelization. The fact that Macs use bash_profile as opposed to bashrc will cause some minor annoyances until you fix it.

  4. I have found it to be pretty easy. I needed to install both XCode and homebrew, however, and once homebrew was installed it was easy to get any other libraries I needed to install software. Also, if someone is new to linux environments, MACs allow for nice baby steps.

  5. I think so..

  6. I would get MACs if you have a very diverse lab (people with different skill sets), if you want the computers to do more than just analysis, and if you have money. They're pretty user friendly, the Linux backend allows you do to what you would normally do on a Linux system (with some differences) if you're an advance user and it will hold your hand if you're not. But you will pay more for less, and in terms on analysis I find there isn't anything I can do on a Mac that I can't do on a Linux machine. I also find Ubuntu, while not quite the same as a OS 10.9, to be pretty good at hand holding too.

ADD COMMENT
3
Entering edit mode
11.0 years ago

I am someone who used OSX for my main work laptop since 2009, while getting other work done in parallel on Linux systems in the lab. A little less than a year ago I was faced with the cost of upgrading my dated Mac and buying a new one, or going entirely with Linux. Looking at the price gap between the closed environment of Mac vs Linux, I made the decision to go with a Thinkpad into which I shoved extra RAM and put on my favorite Linux distro. There have been some definite pluses and minuses, and I do not think there is one easy answer here.

On the plus side, all of the development issues that I had working in the Mac environment have gone away. Although Perl, Python, Ruby, and R are all found on Mac, I never was able to install catalyst, some Ruby gems wouldn't install properly (I know this sounds vague...but boy, catalyst was a time sink getting it to run on a Mac...as in, it never worked...), and bioconductor, I gave up trying to install it on my mac in 2010 or so... (although others here may have a different experience...partly is my laziness about upgrading my version of OSX). But now on Linux I can do both tool development and data analysis on the same laptop, and I am quite happy with that.

For years on my Mac I used open-office/Libreoffice, and zotero for reference management for writing scientific papers, so that hasn't changed. You do need to plan well when contributing to papers with non-open source software users, but I just convert whatever to .doc before sending it to collaborators. I've been very happy though with this system, and I don't think my collaborators can tell the difference. There is no Photoshop that will run on Linux, and the gimp is...just...different when it comes to making publication-ready images. So that part is a bit rocky and a learning curve, but not too bad.

Now for the downsides since I made the switch -- Java apps can be a pain with Linux. For example, just today I spent too much time trying to get compatibility with a common web-based screen-sharing video conferencing tool...and ended up jumping over to my old Mac to get the job done. Adobe reader just won't work on my distro either. Also, wifi connectivity is a real downside, as in, this laptop with this Linux distro just cannot find public wifi reliably. I took the Linux laptop to a recent conference, and was not able to connect anywhere, while my Mac would have done the job nicely. That is a big downside, although I never have that problem when I stick to work or home. These are the user interface downsides Alex Reynolds points out too. Plugging that Linux laptop in to a vga projector to give a talk is always an exercise in anxiety...will it find the display? In OSX these things do just work...although you pay more for it.

Yet, the bottom line is whenever I want to comfortably get any work done at the command-line or ssh into anything, or run something in R, I'm very happy I have the Linux laptop, since all of THAT just works, and it wasn't reliably my experience on a Mac. Why not find a way to use both?

ADD COMMENT
1
Entering edit mode

I'm having a hard time parsing what you say about bioc and macs -- are you saying "Bioconductor: just forget about trying to get it to work on a Mac"? If so, I'm not sure why you would say that: I've never had any problems installing/using BioC on my mac ever since I started using Bioconductor "way back" in 2008.

ADD REPLY
2
Entering edit mode

Which is why I followed that statement immediately with "although others here may have a different experience..." Why was it such a nuisance installing BioC on my mac in 2010? No idea, today I just recall a nest of error messages... works fine on every Linux distro I've run since, though. Modified my post...BioC is awesome, after all... just for me mac hasn't been a reliable platform to run everything I run on Linux, that's my message.

ADD REPLY
3
Entering edit mode
11.0 years ago

I think I'm echoing what Dan Gaston, JC, Heng, and perhaps others, are saying here, but to be very blunt:

If you have the money to buy two machines, I'd say you should buy yourself a nice macbook (pro or air) for your own work and buy an even nicer "server/HPC grade" linux machine for the lab.

I've been doing all of my work since forever on macs and have relied on big-iron linux machines for things that need a lot of wall time (short read alignment, variant calling, etc). Like JC said, prepping your pipeline code on the mac and getting it to run on your big-iron-linux machine is rather straightforward.

ADD COMMENT
1
Entering edit mode

Isn't prepping your pipeline code on the mac and getting it to run on your big-iron-linux machine prone to get you into subtle or not so subtle bugs because of different versions of the basic utilities (awk, sed, sort...)?

ADD REPLY
2
Entering edit mode

I haven't really run into these problems, but then again I don't really use sed or awk much -- most all of my text munging is done in another scripting language (either Python or R).

Like Alex suggested, GNU versions of these userland tools are also available to you by either compiling them yourself, or (likely preferably) through a package manager.

As far as package managers go, homebrew (not macports anymore (sorry Alex ;-)) seems to be "the chosen one" these days, especially after you add the science "tap" via brew tap homebrew/science, as suggested by Carlos B.

ADD REPLY
2
Entering edit mode

Probably not. But in case it is a good argument in favour of having test data with known outputs to verify your results are the same regardless of the system they are run on. There also tends to not be much difference between versions of most basic command-line utilities, and you should be keeping your production pipeline identical between your machines for other software anyway.

ADD REPLY
1
Entering edit mode

You can install GNU versions of those core utilities either by compiling source or grabbing a package via macports, etc.

ADD REPLY
0
Entering edit mode

I did not have any problem with version problems of basic utilities, R perl or python.

ADD REPLY
2
Entering edit mode
11.0 years ago
Michael 55k

As a Server

I think it is maybe the hardware that is the limiting factor here, because the MacPro is more like a powerful graphics workstation than a number-crunching - server. I think it is good to use if you want that kind of stylish workstation to put under or on your desk as it is (is it loud?), and do some video editing, 3/4D rendering and protein modeling, and other graphically demanding visualizations, but following the specs I have my doubts about its usefulness as a general a bioinformatics server.

  • "Configurable to 32GB (four 8GB) or 64GB (four 16GB)", that's less than your required 100Gig.
  • Possibly overpowered graphics if you don't need it. If you want to use GPU-based acceleration make sure that the HW and OS are supported by the software you want to use.
  • It doesn't fit in a rack (obviously), but you might have to have a rack full of other equipment (storage, backup, UPS, all other machines).
  • Its form factor doesn't seem to stack well. Your tech department might not be amused over this either, if you want to put it in the common server room.
  • Internal storage is max 1TB. Are you prepared to buy and attach external thunderbolt storage? Or can it be integrated with your existing storage and backup solution, or are you prepared to build that up from scratch (see also point 3)?
  • Is it compatible with the cluster solution you have to use?
  • Is all the software you want to use available or can it be compiled? (That's the primary question for any OS)
  • The price (if you end up installing Linux on it then it will just be a very expensive linux box)

As a Laptop

  • I need stuff that just works, I don't want to waste time and money on figuring out that the power management or network doesn't work with this specific chip in the latest laptop I bought, but windows is not an option.
ADD COMMENT
1
Entering edit mode
11.0 years ago
DG 7.3k

I'll echo what a lot of people seem to be saying and just iterate from my own experience:

Are you using a Mac as a main computer for bioinformatic work?

Sort of. I have two computers at my desk. One is a large workstation (theoretically multi-user) but I am the only bioinformatician on our project, so it is all mine. 50GB of RAM, 8 cores, a few TB of storage. That does all the heavy lifting tasks (mapping of exome sequencing data, variant calling, analyses, etc). But I also have a Macnbook air in front of me that I am using constantly. That's where I do all my writing, most of my coding, my research, and actually look at most of the data. I generate reports intended for bench molecular biologists and clinicians so having access to Microsoft office products is pretty key for ease of use within the team. And I still say those programs are far better than LibreOffice, etc.

What positive points do you see about using a mac for bioinformatics?

It's a UNIX system. It has a few quirks (but so do all unix/linux distros) but you have the command line, you can build software from source, and most people make OSX binaries. While I am very comfortable in a Linux environment the fact that OSX largely "just works" and doesn't require much fiddling and tinkering leaves me more time to actually do research.

What negative points do you see about using a mac for bioinformatics?

For heavy lifting computation the equivalent machines do tend to be far more expensive than a Linux workstation. Sometimes you run in to some problems getting some software compiled and built for the OS, but I find that pretty rare.

Ease of installing software (talking about bioinfo software here)?

Usually just as easy or easier than on Linux. Vastly superior to Windows.

Freedom to hack quick solutions for custom problems?

Again, about the same as Linux, way easier than Windows. Can still do all the same scripting, shell work, etc than I can on any UNIX system.

Why should I (or not) think of getting Macs for our lab?

I wouldn't get it for your workhorse workstation/server needs. A Mac Pro is way more expensive than the equivalent workstation. I wouldn't even go with a name brand solution myself. Find one of your local computer builders who deal with enterprise solution and see what they can do for you. We have done that for our needs in the past and they were cheaper and way better service. On-site service with long warranties for no additional fees, etc. If you can;t find anywhere good then go towards an IBM, Dell, etc. They tend to be easy to deal with as well.

ADD COMMENT

Login before adding your answer.

Traffic: 2427 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6