Rec. 1: A standing committee be established to review and suggest changes to the Lamont computing environment on an annual basis. Staffed with scientists and technical personnel, this committee should also develop policy and procedures regarding acceptable computing practices, and advise the Network Support Group on its annual budget request. The CPC01 believes that this standing committee should meet in the Fall of each year, so that its recommendations can be considered and (if accepted) implemented by the Director in time for the next Lamont fiscal year. This committee will also be charged with conducting computer users surveys (similar to the one conducted by the CPC01, Appendices A & B) every second year, during the summer preceding the committee's deliberations (next survey due in 2003).
II. For many years the principal source of revenue generation
to recover computing costs at the Observatory has been through network
``subscriptions'' for Sun/Solaris computers (desktop and servers). Although
the approximately $3200/
yr/machine subscription is not unduly onerous compared to sister institution
practices (footnote 2) ,
it is highly inequitable because there are about 120 Sun/Solaris machines
which are charged, whereas there are about 350 Macintoshes and PCs which
are currently not charged. There is little doubt that the great disparity
in cost recovery practices across the various hardware systems has contributed
in a major way to the heterogeneous hardware/operating system environment
found at Lamont today. The CPC01 recommends that:
Rec. 2. Procedures for recovering computing costs at the Observatory be changed to reflect the actual services provided to the different computer platforms at the Observatory. (footnote 3)
III. The unfair subscription structure aside, there are sound technical reasons for PI and group investment in non-Sun/Solaris hardware, particularly PCs running under a windows variant and/or linux. The fact is that PCs are now faster than Sun desktops using standard benchmarks, yet they are 3 to 4 times cheaper than Sun/Solaris machines with the same performance (footnote 4) . In other words, for the same dollar outlay, a user can expect significantly ``faster'' performance from an Intel-based desktop. Secondly, fewer and fewer applications, particularly science applications, are being developed and/or supported for Solaris (and Unix in general). In recent years, the vendors of ArcView GIS, GM-SYS, and ERMapper, for example, have ceased product development for Unix. These are realities of the marketplace, which the CPC01 acknowledges. Thirdly, computing activities at Lamont tend to be sorted into two broad application classes: ``research'', and ``office''. According to the users survey, it is not uncommon for Lamont scientists and technical personnel to use two platforms in the course of the work day, a Sun/Solaris desktop for doing research, and a Mac or PC for office activities (proposals, papers, presentations, etc.) (footnote 5), particularly where the second computer involved is a laptop (footnote 6) . Administrators and other non-technical personnel tend to work exclusively in the Mac or PC domain. The CPC01 recommends that:
Rec. 3. The Network Support Group should be expanded by at least 2 FTE, and responsibilities reallocated among its members to reflect more closely the actual services required now, and in the future, by the various computing ``clients'' at the Observatory. At present, the Network Support Group consist of six FTEs whose responsibilities reckoned on an annual basis are given in Attachment 1. Given the installed hardware, software, and operating system base and its likely evolution, these responsibilities need to be realigned, and new hires made in key areas. Based on the users survey comments, we need to increase the level of support given to Macs, PCs (in general), and clients with PCs running linux, in that order of priority. We probably need to hire a Mac expert familiar with Mac OS X and with generally more technical knowledge than present staff. After that, we probably need a person charged with network, Unix workstation, and PC/Mac security responsibilities. Network security and reliability are among the major concerns voiced in the users survey. Adoption of some or all of the recommendations in this report will require reallocation (to some extent) of priorities and responsibilities among Support Staff members (footnote 7) .
IV. The traditional strength of SUN systems - robust, secure file and other services across a distributed network - can and should continue to serve the Observatory well into the future. However, not everyone needs a desktop computer with large local filesystems and a local operating system. Indeed, many users logon to the fastest SUNs with the largest memory they can find on campus to do their research, the local desktop functioning primarily as a X-windows terminal. Often the remote machine becomes bogged down by many processes being run by such individuals. Fortunately, there are evolving strategies that deal with this situation in a more positive, manageable fashion. The CPC01 recommends that:
Rec. 4. The Observatory consider replacing many of the Sun/Solaris desktops scattered around campus with Sun Ray appliance clusters. A successful experiment implementing the Sun Ray strategy by a research group on campus is described in Attachment 2. The Sun Ray model isolates each cluster of appliances on a dedicated subnet, and consolidates all primary disk and/or tape drives on the server. Management of services becomes more efficient because network personnel rarely have to visit the user's office. Support instead is centralized on to a smaller number of server farms and RAID storage systems distributed in the campus buildings. As performance/cost ratios of servers and disks continue to increase, incremental upgrades become more viable economically when such services are consolidated. In the Sun Ray ``model'', the desktop appliance is analogous to the telephone handset in the client's office, while the server has the role of the telephone system.
V. Ours is an era in which fast and secure communication is paramount. This applies no less to computer networks and associated infrastructure than it does to telecommunications. In fact, we should regard both communication forms as integral parts of information technology in general. However, many users are worried about the reliability and security of Lamont's basic network infrastructure, particularly in view of the unmanaged growth of the present heterogeneous platform/operating system environment. Thus, the CPC01 recommends that:
Rec. 5. Steps be taken to ensure redundancy, reliability, and security of the Observatory's computer network, implementing such changes that will improve user access to network resources. (footnote 8) This recommendation covers everything from basic power (phone systems are robust to power failures, and so should the network), to redundant routing both within Lamont and to the outside world, to security issues (regulation of what may and may not be connected to the network). Many of these issues can be addressed under Rec. 1 and Rec. 3 above; however, they must be solved as matter of urgency. Moving from 9/5 network support to a reliable, secure 24/7 operation will place even greater stress on the Support Group, which must be expanded accordingly (Rec. 3).
VI. Our final recommendations are drawn from the user concerns expressed in the survey. These are:
Rec. 6a. The current Lamont website is a disaster, and needs to
be redone almost completely. (footnote 9)
The Lamont web pages are our most public face to the outside word, a good
avenue for recruitment of graduate students and Post-doctoral scientists,
a resource for everyone working at the Observatory, and the best tool for
explaining what it is we do to the general public. The current web pages
fail in all these tasks.
Rec. 6b. Lamont should foster research into data base management
tools and techniques. (footnote 10)
As implied in the Institution's name, our science depends critically on
the Earth observations we make. We need to take steps necessary to insure
the integrity of our data bases, and user accessibility to these valuable
resources.
Rec. 6c. Training pages for all ``institutionally supported'' applications should made available under the Network Support Group web pages. Many users responding to the survey were surprised by the number of applications that are available on the network, and desired training in a number of them. Although a suite of applications is supported by institutional site licenses, many other software packages originally obtained by individuals and groups can often be made available to other users. As presently configured, the Support Group does not have the resources to develop the material required to train new users. While many commercial applications provide training through on-line tutorials, more progress could be made if ``gurus'' in some of the shareware/freeware applications could be persuaded to contribute to web-based training pages which would then be maintained by the Network Support Group of the future.
Your Comments: Discussion Board
