By Sramana Mitra and guest author Shaloo Shalini
SM: If I were to ask you a question, given what you know of your situation at Harvard Medical School (HMS) and the industry in general, what are some of the opportunities that you would point entrepreneurs toward in the context of research done at HMS? We have discussed this to some extent already, but I would like to know more. What kinds of off-the-shelf solutions are required such that your researchers at HMS don’t have to reinvent the wheel? Can you give me some examples of specific areas where you would like solutions?
MA: Sure. A lot of open opportunities may be available in the public–private cloud area, if I might even call this a ‘switch.’ If there could be some interesting innovation to make that switch easier for people who work with private clouds, or help to tie in a private cloud to public clouds. There could be specific portals that serve as a front end to the cloud and allow researchers to do common types of analysis without having to write code or log into a shell prompt. Those are clear opportunities.
Another area where innovation is needed is an area that we now have a good handle on, finally. After years of dealing with petabyte of data, you realize that you need to deal with it differently from, say, with 50 or 100 terabytes.
We still have a big problem in that area. Any time you have a large data set, say one hundred terabytes, and you have a colleague at MIT with whom a researcher at Harvard wants to collaborate – if you look at such potentially sharable data, there is an opportunity to look for smart network solutions. This would be much like they do with bandwidth, traffic, and other things where the network and the solution knows how to move the data closer to where it is being used in the cloud. One such large data set could be shared. If you think of a cloud of storage, say, at Stanford, Harvard, and MIT whereby they all decided to collaborate on a project. Let’s say they donate some storage to a cloud and get reasonably fast connectivity, and a service provider and now an entrepreneur comes in and builds services around this infrastructure and distributes the data into the right places by being able to sense that the data is being heavily accessed at Stanford and hardly being accessed at Harvard.
In this case, the solution could help to keep enough data at Harvard; only that which the researchers commonly use at Harvard will be located there, but the solution helps to shift the weight of the data to Stanford for quick and fast access by Stanford researchers. There has to be lot of thought and development around it. Basically, I am referring to tackling the problem of sharing very large data sets among medical researchers located in different institutes. This is a big open opportunity for entrepreneurs to come up with innovative solutions.
SM: Given the approach to the private cloud you have taken over the past five years, how has your IT organization changed or evolved as a result of cloud adoption?
MA: It is hard to decouple some of that. We made them change the way original IT infrastructure worked at HMS. When I came into the med school, the primary MO of the IT group, with the exception of my boss, John Halamka, CIO, who has been a visionary in helping to reshape the med school with new technology adoption, in essence was to not be very flexible. The practice was to almost treat customers, the researchers, the same as someone working in a bank or hospital where you need to lock them down in terms of control. But researchers usually don’t operate well in that environment. So we made a lot of cultural shifts to improve that. We literally tore up our service. Our old SLA, which was eighty pages, said we will help with MatLab version 9.2 and not 9.4 and we will only work on Windows PC at that time and things like that. At that time there were some Macs around, but everyone used Windows in IT. But 60% of researchers use Macs now. So we made a lot of cultural shifts and said our job is to be flexible. As part of doing that, let’s build a cloud that people can buy into if they want, or they can just use what they have standalone in their respective labs, and let’s prove that cloud model works over time.
The cultural shift to the cloud model at HMS has brought people together to collaborate. Researchers at HMS are not just putting up their own storage arrays under their desks for their own use in small clusters. The cloud model has actually brought people together, and it has saved the school a lot of money because it is much cheaper to take a centralized cloud approach instead of islands of infrastructure for labs to do their own thing individually.
[Note to readers: Companies such as Google, Amazon, and erstwhile grid computing heavyweight Platform Computing and other grid companies such as Cycle Computing are delving into cloud computing technologies and models to address bioinformatics and gene sequencing computing requirements. Unlike technology companies, life sciences and medical research are slow to adopt cloud computing but there are some startups which are looking at newer technologies such as cloud as more efficient and agile form of computing. Luke Timmerman writes here about how gene sequencing has been on a breakneck pace of innovation over the past few years, as instrument makers such as San Diego-based Illumina and Carlsbad, CA–based Life Technologies have lowered the cost of sequencing an entire human genome to as little as $10,000. Upstarts such as Mountain View, CA-based Complete Genomics seek to sequence entire genomes for as little as $5,000, while a rival, Pacific Biosciences, is aiming to sequence genomes in fifteen minutes. Since every human genome has six billion chemical units of DNA, this faster and cheaper form of sequencing is creating enormous datasets that somebody will need to store, analyze, compare, and visualize. Without that capability, it’s just a vast pile of data that doesn’t really lead to valuable new insights for medicine.
Leading microbiologists and computer science expert C Titus Brown is one of the evangelists of cloud computing in genomic research. He is a computer science and microbiology professor at Michigan State University who is teaching students how to use Amazon Web Services to store data for their experiments.]