Canada Foundation for Innovation
Consultation on proposed strategies for genomics and high performance computing (new)
This section is for applying institutions and members of the research community who are preparing applications to the CFI. It contains information on program mechanisms, competition deadlines, and on the process for reviewing applications.
The CFI is now preparing the next two competitions under the Innovation Fund, which will be held in 2000 and 2001. As part of this process, the CFI is seeking advice from institutions, groups, and individuals on the strategies for high performance computing and genomics that were suggested by the two relevant Task Forces.
Institutions, associations, and individuals are invited to submit their comments as soon as possible, and no later than August 15, 1999. Simply send a message to firstname.lastname@example.org for comments on the genomics strategy, and to email@example.com for comments on the HPC strategy or a fax to 613-943-0923.
Following the first phase of the Institutional Innovation Fund competition in September 1998, the CFI saw the need to promote further rationalization and integration of proposals in certain areas, including genomics and high performance computing (HPC). In order to maximize the return on its investment and to ensure consistency in the assessment of proposals, the CFI formed Task Forces to look at the infrastructure needs in these specific research areas and to recommend strategies for the first and subsequent competitions.
PROPOSED STRATEGY FOR GENOMICS
The Genome Committee established by the Canadian Foundation for Innovation (CFI) offers the following observations and recommendations regarding a strategy for funding genomics-related infrastructure.
- The Committee recommends that CFI support infrastructure projects which would:
- go beyond acquiring and applying current genomics technologies by aiming to be truly innovative in developing new technology (often this will happen by creating new dimensions of disciplinary interactions e.g. by involving MITAC in Canadian bioinformatics), expanding the range of application of existing or new technologies, or establishing imaginative linkages to accelerate the application and commercialization of genomics in Canada
- contribute to distinctly Canadian capabilities in genomics
- centralize the provision of core facilities within institutions, regions or research networks with the objective of meeting the needs of larger, even national groupings of researchers for access to well managed, client-oriented facilities using forefront methods and equipment.
- Specifically with regard to infrastructure for megasequencing (i.e. more than one million base pairs per annum), the Committee recommends that :
The Committee also considers that Canada's megasequencing needs in universities, hospitals and other research institutions are such that dependence on the private sector to meet all, or even a large proportion of those needs would not now be appropriate.
- new services should be centralized within a university or other research setting, and possibly within a region
- funding of facilities should be conditional on the provision of a credible operational plan which includes forecasts of the needs of the institution (in terms of technical capability and throughput), as well as the needs of other clients; which describes and estimates the costs of the essential operational elements required to ensure cost-effective quality and timely delivery of results; and which includes plans for acquiring, training and retaining key technical support personnel.
- For bioinformatics, the Committee recommends that CFI-funded genomics centres be encouraged to network nationally by affiliating with the Canadian Bioinformatics Resource established by the National Research Council for the purposes of: accessing high speed connectivity; ensuring server compatibility; accessing high performance computing capabilities; accessing specialized resources (e.g. genome annotation, protein folding and sequence assembly); sharing newly-developed software; and encouraging collaborative bioinformatics R & D with appropriate linkages to superior mathematical and computational expertise.
- Looking to future competitions, the Committee recommends that CFI encourage institutions to seek funding support for genomics-related infrastructure that would address Canada's needs for :
PROPOSED STRATEGY FOR HIGH PERFORMANCE COMPUTING
- the application of genomics to a broad range of economic sectors (including agriculture, aquaculture, environment, forestry, health and microbial biotechnology) and be linked to government laboratories, if appropriate
- more forward-looking, innovative and multi-disciplinary approaches to developing new genomics-related technologies or applications, including: lower cost approaches to large scale DNA sequencing ; going beyond existing paradigms in proteomics; massively parallel computing for bioinformatics; creating distributed facilities for DNA array generation, hybridization and scanning; and extending functional genomics to studies of metabolic processes.
As a result of its April 26 to 28 meeting, and of its review of eleven proposals to CFI for HPC funding, the Task Force has the following observations and policy suggestions to make:
Observations on this past competition:
- A good number of the applications request computer installations which do not truly fit the current definition of high performance computing. This is probably the result of two factors: a) there was not a specific call by CFI for high performance computing applications and b) Canadian universities lag far behind what is viewed as the "high end computing" facilities of leading universities in the United States and the proposals represent (in most cases) a reasonable and manageable first step towards achieving that level of computing power.
- The degree of innovation contained in the proposals is quite varied and tends to relate to the level of current facilities.
- The Task Force sees both pros and cons regarding the option of leasing rather than purchasing the equipment. However, leasing tends to tie the institution to a single technology provider to a greater extent and purchasing is the preferred option if low-cost, up-front funds are available.
- Some applicants appear to have achieved better pricing from equipment suppliers than others have. While more careful analyses may conclude that this is not the case, such analyses should assess carefully the charges included by the suppliers for "training sessions" and maintenance contracts. These can be vehicles for recovering a portion of the discounts given on the equipment itself.
- With CFI and other funding in hand, the successful applicants can probably improve the pricing of the equipment being proposed. In any event, given the passage of time, and the rapid pace of change in this technology, the applicants should explore carefully the new options for providing the desired computing power. Furthermore, while it is understandable to want to purchase an overall package of equipment from the supplier willing to enter into a "partnership" with the university, the Task Force noted a number of instances where it felt that auxiliary equipment would be available from other sources at much less cost.
- The "educational" and other discounts offered by the equipment suppliers are not too meaningful unless the base list prices can be compared in absolute terms with those offered elsewhere, especially in the United States where some sites are viewed as being of 'strategic' for marketing purposes. Being able to make a true comparison is of particular importance to potential outside supporters who wish to be able to differentiate between "soft" and "hard" support by the vendors. There were some cases where Task Force members felt that the prices (in $ US) were higher than those possible in the United States. However, this observation should not detract from the fact that, in most cases, the discounts negotiated by the applicants have been quite good.
- Collective procurement at the 64 to 128 CPU levels might improve the discount levels, but it should be noted that an agent charged with collective procurement often does not have the same level of commitment for the best deal as does an individual from the host institution.
- With few exceptions the applicants were ill prepared to defend the proposed equipment by identifying the tasks that necessitated that level of computing power. This lack of a systematic assessment of need was surprising and resulted in the Task Force spending an inordinate amount of time trying to extract the true computing needs of the researchers. As noted below, a specific call for HPC proposals with a well-designed application form would probably correct this situation for future competitions.
- Finally, the applications reviewed by the Task Force served to document quite graphically the extent to which Canadian universities have fallen behind in terms of computing power. As previously noted none of the applications requested true state-of-the-art high performance computing facilities and answers to the Task Force's questions suggest that few of the researchers have had first-hand experience with such facilities. Accordingly, there is an urgent need to start a catch-up process and the provision of the equipment recommended for funding will help be of great assistance in that regard.
- The Task Force is of the opinion that the C3.ca organization can play a leading role in bringing university computing capabilities up to satisfactory levels. It is imperative that facilities such as those requested be shared within the academic (and business) community and C3.ca appears to be the group that can best manage that process. It is therefore recommended that:
- C3.ca be charged with developing and implementing a sharing process for all CFI-supported computing facilities. It should be the clearinghouse or allocations agency for major blocks of shared computer time (CFI to define "major") and it should be the agency to record and report the amount of sharing actually realized.
- Twenty percent of time on all CFI-funded installations should be committed for use by external Canadian users with CFI allocating major blocks of time on the basis of recommendations from an allocations committee and access priorities established by that committee. C3.ca membership should be a condition of access to such major blocks.
- No "user charges" should be permitted for shared time.
- C3.ca should strive to have common accounting (time, CPU use, etc.) adopted by the participating institutions.
- The Board of Directors of C3.ca should include one or two experts drawn from the global HPC community as well as one person from a small Canadian university.
- C3.ca, as the allocations and reporting agency, should develop an effective communications program regarding the effectiveness and the level of sharing being realized.
- If CFI is going to be a major player in the process of introducing HPC computing power to Canadian universities, then the next round of applications should be in response to a competition designed specifically for high performance computing. The call for proposals would elicit the type of information which, in this instance, the Task Force had to extract (with mixed success) during the course of the interviews. For example, the invitation would require applicants to document the research efforts that demand the level of computing power being requested, the experience of the researchers on such equipment as well as experience in developing code for that equipment. In addition, and again contrary to the current experience, hardware and operating costs would have to be closely linked and justified.
- The Task Force is of the opinion that it is critically important that CFI begin planning for a call for proposals to develop a truly world class "high end" computing facility as single national centre which would be maintained at the state of the art level. (Canada does not require, nor could it support, a number of facilities at that level or even approaching that level.) The equipment recommended by this report would, over the next two to three years, produce both the experience and the research challenges required to utilize effectively a national "high end" computing centre. Perhaps four of the applicants to the current competition should be capable of submitting excellent proposals within that time period.
- There were few cases where the current applications for funding noted that the equipment would be used for research into high end computing itself. Research in this field should be encouraged by NSERC (assuming the required computing facilities are available).