A Perfect X: The rise of Macs in life sciences
Forget, for a moment, the Apple faithful's fanatical love for the Mac -- declared with numbing monotony during the reporting for this story. (It's truly amazing.) The real question is: Can Apple's growing presence in the life sciences transform it into more than an afterthought in mainstream computing and perhaps a dominant player in all of science?
To that question, no one yet has an adequate answer. However, there is no questioning the strengthening rise of Apple in life science. Apple stalwarts have been lurking in the shadows for more than a decade and watching with derision as the world went Windows. They suffered lasting estrangement from IT departments and learned to fend for themselves while Apple's self-inflicted market missteps threatened them with extinction in the mid-1990s.
Now they're fighting back with some needed help from Apple.
In the past five years, the introduction of Mac OS X, the G5 processor, the rackmounted Xserve line, and recently Xserve RAID, Xsan, and Xgrid have fanned new flames in Apple hearts. A small army of 1990s graduate students who tackled all things 'omics by writing and using open-source applications -- many on the Mac -- are now moving into lead positions in academia and industry, where they are busily writing grants and drafting budgets that make room for Apple.
"I've been waiting a long time for this, since they first started talking about Copland (Mac OS 8 upgrade) around 1994," says Michael Barmada, assistant professor in the Department of Human Genetics at University of Pittsburgh's Graduate School of Public Health. Just four months ago, he powered up a 121-node Xserve cluster that was originally going to be a 200-node Linux system -- before he changed his mind.
Copland, of course, was a minor flop. But the return of Steve Jobs to the Apple helm in September 1997 was followed with his announcement in 1998 of the Mac OS X roadmap. Mac OS X Server, a version for developers, was introduced in 1999, followed quickly by a desktop version, Darwin 0.1, in 2000. It was essentially the same OS that Jobs had developed at NeXT Computer, built on bulletproof Unix. Since then have come Cheetah, (release 10.0), Puma (10.1), Jaguar (10.2), and Panther (10.3). And Tiger (release 10.4) is expected to pounce on the scene soon this year.
X Marked the Spot
X made all the difference, Barmada says: "We have a lot of software, and it's all home-grown stuff. No one in the field is really a professional programmer, and so it's code that's cobbled together and very poorly annotated and, you know, is an amalgam of Pascal, Fortran, and C and C++. You name it, and we've probably got it. I think there's actually some old Cobol code floating around somewhere. So the fact that I could bring that all over to the Mac relatively easily was very helpful."
There were other advantages, too. It is easy to use both office productivity tools and life science applications on the Apple platform. (Cutting and pasting from Word to informatics applications is mostly a breeze.) Security is excellent -- in part because fewer hackers have targeted the Mac, but mainly because of its security-rich Unix underpinning. Mac OS X rarely crashes. And ease of management is a huge plus, Mac OS X users say, not least because IT departments remain mostly neutral to hostile to the Mac.
Suddenly, life science is Apple's fastest-growing business market.
"It wasn't so much that we went after that community, [but] I wouldn't call it serendipity," says Philip Schiller, Apple senior vice president of worldwide product marketing, who has a B.S. in biology from Boston College. "We were very specific about what Mac OS X should be. We knew that we needed to build it on the reliability and open-source nature and performance of Unix, and, not coincidently, the scientific community saw the value as well."
The analyst community has also taken note. "It's the dual nature. I can be in a Unix environment and pretty easily port my Linux open-source applications without rebooting. That's an attractive convenience," says Michael Swenson, research manager at Life Science Insights. "The G5 processor has been designed to improve floating-point performance, which the G4 wasn't. Also, when you think about small clusters, one issue is ease of use and, clearly, the other issue is access to the right applications. Apple has a pretty good story to tell on both."
Don't be misled. Big hurdles remain. Apples' worldwide market share has dwindled to around 2 percent (3-plus percent in the United States), and Mac unit sales are fairly flat. Macs are still more expensive than PCs on the desktop. Technical issues persist -- it is a computer, after all. The rush to "free" Linux remains in full force. In-house and third-party Mac technical support is relatively scarce. Power users and developers grumble that Apple provides few details about its long-term technology plans. Moreover, Apple has made slower progress in convincing laboratory instrument makers to write drivers and interfaces for the Mac.
Curse of Redmond
Then there's Microsoft Windows. It's everywhere, and supported by virtually everyone (developers, IT departments, retailers, etc.). Only the graphics arts community has stubbornly resisted Window's onslaught, staunchly remaining the single biggest market segment for the Mac.
"I think we are still an underdog in many places," Schiller agrees, but that is clearly changing. "Our goal isn't to be the world's biggest supercomputer company -- that's not who we are. But we do see a market for taking these Xserves and providing department-level clusters so people can now have access to high-performance computing and cluster computing that is easy to set up and easy to use in any department."
Indeed, Apple clusters are sprouting up all over academia, though market share numbers for that segment are difficult to come by (Swenson declined to estimate for this article, as did Apple). And Schiller's comments notwithstanding, Virginia Tech's massive 1,000-plus-node, 12.2-teraFLOPS Apple cluster has grabbed most of the recent headlines by virtue of its spot among the top 10 supercomputers in the world. But there are many more modest installations.
Idaho State University is a good example. Molecular evolution researcher Michael Thomas arrived in the summer of 2003 to take an assistant professorship, and promptly built a 10-node Xserve cluster serving 30 faculty and student users.
"Before that, I was working at a bioinformatics center (Medical College of Wisconsin, Milwaukee) that had a huge cluster of Linux machines but also Compaq and Sun clusters -- really high-end stuff -- and most of the users on those were power users so people who developed their own programs and their own tools and were pretty much the ones who had access to it. Apple was a drop in the bucket there," Thomas says.
Searching for something more broadly accessible, Thomas read about the iNquiry package (cluster configuration and informatics toolset for Xserve). "It wasn't for sale yet, [so] I called BioTeam (iNquiry developers), and talking with them is what clinched it."
As it turns out, Thomas -- or, perhaps more accurately, his bioinformatics programmer, Luobin Yang -- added fuel to Apple's spreading fire. iNquiry is intended to make it easy to configure and administer an Apple cluster and also has about 200 open-source informatics applications. It was not ready for prime time, Thomas says, and he and Yang helped work out its kinks.
"I worry a lot, especially when he's on the phone chatting with iNquiry and the BioTeam and telling them the latest, greatest thing he's put in there. My nightmare is that one day he'll be moving to Cambridge," Thomas says, adding quickly that Yang's wife is a nursing student and gets free tuition while he works there. "So I've got at least three more years." iNquiry is now bundled in Apple Workgroup Cluster for Bioinformatics.
Ease of management was critical to Thomas' Xserve choice. He receives little or no support from central IT. "I'll get tenure or not (in roughly four years) based on my science, not the cluster. I didn't want to be tied up administering it, and I wanted Luobin to be writing programs, not administering it." It took six weeks to gather all the equipment, but just a day or so to get the cluster up and running -- not counting iNquiry troubleshooting. Thomas' research on alternative splicing in genome evolution requires multiple searches and homolog alignments between species.
Last summer, members of the University of Idaho informatics group visited the Idaho State cluster, liked what they saw, and are now setting up their own 250-node Xserve cluster. "They actually didn't like iNquiry and have been griping about it quite a lot. But they really liked Xserve, and it fits in really well with the other stuff they're doing," Thomas says.
Thomas has growing plans of his own. He is remodeling the current IT space to accommodate more machines, and plans to swap older G4s for G5 systems. Also, whenever one of his colleagues plans to submit a grant, he writes a little section for them to add in, supporting the purchase of more nodes where needed. The microscopy center has also inquired about adding a couple of nodes for its use.
It seems a little ironic that Apple's resurgence is reminiscent of the PC revolution. Departments in corporate America, unable to get what they wanted from central IT organizations, had sufficient budgets to go out and buy PCs themselves and use them for office productivity tools -- primarily word processing, spreadsheets, and later for presentations. The PC spread outward from departmental desktops.
Today, the Mac presence on the research desktop varies widely. At the Whitehead Institute, Macs hold a 55-percent to 45-percent edge over PCs, and that has been steady for about a year, says Jay Caprioli, a desktop support team member at Whitehead. Similarly, about half the faculty at Idaho State have Macs on the desktop. At Pitt, Barmada buys Macs "for all of our students and postdocs within the division because they need the Unix utilities."
Once inoculated with the Apple bug, users are seldom cured. Barmada was infected in 1986 while a student at Carnegie Mellon. "I was writing papers and surfing what little Web there was," he says. For Thomas, addiction occurred at Penn State, where much of the molecular evolution software ran on the Mac.
Says Whitehead's Caprioli, "A couple of years ago, one of our faculty switched to Windows for a particular application (voice recognition) and found that the whole platform experience of moving to Windows wasn't worth it, and now he's switched back to the Mac."
Robert Latek, a senior Whitehead bioinformaticist, adds, "From our perspective in the lab, it's definitely more expensive to purchase an Apple system and an Apple laptop than getting some sort of PC system; however, it really pays off in the long run. The machines last a long time, the quality of the equipment and the displays is phenomenal, it makes up for the price, and just its ease of use."
"We've got people out there who are still running on PowerMac G3s going on almost six years now, and as much as we'd like them to upgrade to newer hardware, some of them are pretty happy with what they've got," Caprioli says.
Conquering the Whitehead server room may be more challenging. It's all Linux. "A few years ago, we had everything from SGI, Solaris, Linux, to other flavors of Unix. It just became too difficult to maintain all those different systems," Caprioli says. "We chose at the time what we thought was best, and that was Linux. Also, the enterprise side of our IT department was very comfortable running Linux systems."
Still, Whitehead recently bought a single G5 Xserve and is experimenting with NetBoot for managing its Macs. "Apple's got some great tools on the server level for moving off the network, system deployment off the network, and update installation that's all through NetBoot. We're going to investigate where we can go with other things," Caprioli says.
Although not directly involved, Caprioli reports Whitehead has performed comparison tests on different 32-bit and 64-bit platforms for a new cluster. Apple's costs weren't significantly higher, he says, but installing open-source software on the Xserve posed a challenge. Administrators had a difficult time gathering needed libraries and installing them.
"I think that's something that will improve over time. Right now, it seems to be one of those things that's sort of opposite of the desktop. It is a little bit harder to administer the Xserves for scientific purposes than to manage the desktops," he says, sounding a somewhat contrary note to Idaho State's Thomas.
Cracking Big Pharma is also likely to be difficult for Apple. Cost is one obstacle. Entrenched corporate IT is another. The industry's legendary conservatism is a third hurdle. At Novartis, George Morris, chief operating officer for informatics, says Macs are welcome in the lab but not actively courted. Smaller biopharmas are somewhat more receptive.
Alan Goates, Isis Pharmaceuticals' director of bioinformatics, took a tiny step into the Apple orchard with a single Xserve to support a biosurveillance database project developed for the FBI. The database has about 50 tables, and comprises a few gigabytes of data about organisms, diseases, and host-pathogen relationships. More recently, he's been looking at a small Apple cluster.
"When I got here about seven years ago, this place was strictly a PC-only shop. The IT department wouldn't let you have a Mac, and I had to fight tooth and nail to use one myself," Goates recalls. "That policy fortunately changed as my boss became a vice president and informatics [fell] under his umbrella. He also uses a Mac."
Linux still rules the roost at Isis. Its Linux cluster of about 200 nodes handles all internal R&D (RNAi therapeutics). Apple servers are used only for the collaborative government programs that Goates oversees. Still, this small toehold is a start, and in the past five years, Goates says, Apple desktops have grown from roughly 3 percent to 10 percent, with several executives adopting 17-inch PowerBooks as their main desktop.
Schiller is refreshingly candid. "I don't want to underestimate [the challenge]. The difficulty in trying to turn some of those more established IT departments running large enterprise solutions in pharma and other places is tough. We're not stupidly thinking we can knock them all down easily and quickly. And there is still so much to be done in life sciences -- so many seats to win; so many server farms to win."
Rev up the tractors and let the harvest begin.