Attendees: Martin Schmidt Marilyn Smith M. Frans Kaashoek Claude Canizares Armand Doucette Dan Hastings Michael Howard David Segaloff IT Governance Committee Meeting minutes January 17, 2012, 10:00 11:30 a.m. 4-202 Guests: Mark Silis, Assoc. Director, Operations & Infrastructure, IS&T Scott Thorne, Manager, Data Admin and Warehousing, IS&T Mary Weisse, Assoc. Director, Data Management, IS&T Garry Zacheiss, Manager, Server & System Administration, IS&T AGENDA 1. Introduce Mike Howard 2. Approve December 16th Meeting Minutes [David] 3. Update on Data Roadmap - 15 minutes [Mary Weisse & Scott Thorne] 4. Review & Approve Data Centers (servers, storage, virtualization) Roadmap - 45 minutes [Mark Silis & Garry Zacheiss] 5. Review of Research Computing Progress & Plans 15 minutes [Claude Canizares] 6. Next Meeting Wednesday February 8th 3:30 5pm 4-202 Decisions: Approved the minutes from the December 16 th meeting. Approved the Data Centers Roadmap. Next Steps/Action Items: Develop a recommendation regarding TSM pricing/cost recovery and bring it to the IT Governance Committee for approval. Develop a recommendation regarding the use of Cloud Computing and bring it to the IT Governance Committee for approval. ITGC Minutes 20120117 final.doc Page 1 2/13/2012
Discussion: Marilyn introduced Michael Howard, Vice President for Finance. Mike will replace Israel Ruiz as the Administration Representative on the IT Governance Committee. He is also co- chairing ASPCC with Marilyn and Alison Alden. The IT Governance Committee approved the December 16 th meeting minutes. Update on Data Roadmap Mary Weisse & Scott Thorne o The Data Management Group (DM) has become proficient in the use of the new Cognos reporting tool, which is much more robust than the prior tool, Brio Query. Increased functionality includes drill up and down between reports, total web based solution that allows users to send links to reports, etc. o The DM Team is focused on what data people need to do their business and using that information to generate reports rather than just migrating existing reports to the new tool. o The strategy has been to use focus groups to gather DLC reporting requirements. Focus groups from Finance, Human Resources and EHS have been created. o Mary indicated that the Data Management area will stop building Brio reports for community deployment in March. From that point forward all community wide templates will be built in COGNOS. Data Management will still offer Brio support and help to individuals. She assured the group that validation is being done to ensure the accuracy of new reports being generated within the focus groups by Data Management, DLC users and data providers (who also have a representative in each group). She reminded the group that the data in the Data Warehouse is validated against SAP and other systems of record on a daily basis. o Data Management is working on pain points, most notably, the notification process. Today, members of the MIT community receive lots of emails from the various systems about processes and statuses that need their attention. Most of these notifications do not provide adequate information for them to readily understand what they need to do. Mary s team is working on creating an improved notification process (for issues such as: expiring WBS elements, expiring appointments, etc.) that would be sent to users only when certain conditions are encountered. The notification would include a link to a report, which would provide the user with the appropriate detail. This notification process could allow users to opt- out of some notifications. This process would also allow MIT to combine notifications so that users receive fewer more meaningful notifications. o The team has implemented simulated 24hr availability for Singapore. In response to a question about the availability, Mary indicated that the Data Warehouse itself is never unavailable (except on Saturday from 3:00 3:10 when a snapshot is taken for backup). Data is loaded into the data warehouse at night/early morning (12:00am 7:00am) during this time tables are locked and unavailable to users for a few minutes while data is being updated. Data is stable and available to users from 8:00am 12:00am. During the hours of 7:00am and 8:00am other systems of records are extracting data from the Data Warehouse. Users can use the Data Warehouse during that period but response may be slower than during other times. o In response to a question about how difficult the learning process was for the RAFT/COEUS Awards Module, Mary told the group that her team was just starting ITGC Minutes 20120117 final.doc Page 2 2/13/2012
to meet with the schools about this and would have a better idea of the difficulty after her meetings. To make RAFT most useful the research community must have award budgets entered into COEUS first. o Data Management offers a Managed Data Services (MDS) to departments, labs, and centers. This is a managed environment for local data. Local data is stored, secured and cared for in the same professional manner as data in other Institute enterprise databases. Data Management then builds views combining their local data with Institute data in the Data Warehouse. This eliminates the need to duplicate central/institute data. The concept is to keep local data local and central data central. They have done this for Abandoned Property (VPF), Institutional Research Honors and Awards and the Office on Minority Education. They are currently working with School of Engineering on their tenure process and Institutional Research on their survey process. o Engineering tenure process School of Engineering is looking to improve their Tenure Review process by maintaining information about Faculty. Data Management is modeling their process and creating an MDS schema, so that they can reference any Faculty data from the Warehouse and maintain other information that isn t maintained centrally. Data Centers (servers, storage, virtualization) Roadmap Garry Zacheiss and Mark Silis o Data Centers Garry walked the committee through the roadmap. In response to a question about why there is a data center in E40, Garry indicated that they would like to eventually move out of E40 and consolidate enterprise computing in OC11 and W92 and Research Computing in MGHPCC and W91, as those sites are more robust and have generators for power backup. o There are currently about 10 racks of equipment in E40 today. As older equipment is replaced they will utilize more blade technology, thus shrinking the size of the footprint and allowing for more consolidation to other sites. IS&T recently added an additional 3,500 sq. ft. to the OC11 (1 Summer St) Data Center. IS&T owned equipment in the data centers are on a 4 year renewal cycle. In addition to the 5 major data center facilities in OC11, W91, E40, W92 and 24, there is a small facility in Baltimore for disaster recovery. The MGHPCC (Massachusetts Green High Performance Computing Center) is scheduled to go live at the beginning of 2013. MGHPCC leadership is currently looking at the business model for the space. o In response to a question about the possibility of using the MGHPCC facility as a backup site, Garry explained that Research Funding Accounting rules require the facility to be used for research purposes. Garry indicated that the Operations and Infrastructure group talks with the folks at the Bates Linear Accelerator Center in Middleton about best practices. IS&T provides network support to this facility. o Enterprise Storage The overarching theme regarding Enterprise Storage is about sustainability. ITGC Minutes 20120117 final.doc Page 3 2/13/2012
Our overall enterprise storage capacity is rapidly increasing, but the budget remains flat. Garry talked about the cost per GB (gigabyte) of storage decreasing dramatically while our storage capacity increases dramatically. We have 3 tiers of SAN storage and applications are assigned to the different levels based on the requirements of the applications. The tiers of SAN storage are: o High- end (EMC DMX platform) - This is our highest performance and most highly available tier; it's currently reserved for MITSIS, SAP, and the Data Warehouse due to its cost. Its costs approximately $18/usable GB. o Mid- tier (EMC Clariion/VNX platform, Fiber Channel storage) - This is our current "bread and butter" storage platform; it occupies a nice midpoint between price and performance ($6/usable GB) and is used by all of our production- virtualized systems and about 25% of our test/development systems. o Low- end (EMC Clariion/VNX platform, SATA storage delivered via IP) - This is our least expensive and lowest performance tier of storage (~$2/usable GB). It's used by the majority of our virtualized test/development systems, and will soon be rolled out to production systems that meet certain criteria (departmental systems, systems providing services replicated across multiple nodes, and web- based systems with no database or a minimal database component). We will hit a major milestone this year as our enterprise storage capacity will increase to 1 PB (petabyte) of spinning disk, which is equal to 1,000 terabytes. The committee was told that 85% of SOSC (Server Operations Service Center) cost is comprised of enterprise storage costs. Garry indicated that the storage platform supporting the Data Warehouse, SAP, and MITSIS systems will reach end of life in FY13. At that time, we will have three options: o Defer renewal by one or more years. o Upgrade "like for like", staying with the EMC DMX (now "VMAX" in later models) platform (high cost). o Upgrade to the mid- tier VNX product line (potential risk). o His group is working with EMC to evaluate the 3 options, develop the cost model, identify the pluses and minuses and develop a recommendation. IS&T is in the process of reviewing the TSM (Tivoli Storage Manager) computer system backup pricing model. We currently have users who won t use our TSM service due to its cost. o Our TSM service pricing is: o There is no charge for basic service, which enable up to 20MB of data to be backed up. o Our standard service, allowing up to 300 GB of data to be backed up is $15/mth. ITGC Minutes 20120117 final.doc Page 4 2/13/2012
o Our Enterprise level services enables up to 10TB of data to be backed up and the cost is $65/mth. IS&T is also investigating Cloud computing storage providers as a possible alternative for backing up desktop computers. Garry said that he s talked with his counterparts at Stanford and Yale and they have done successful pilots of this technology. One of the possibilities to encourage the community to more widely adopt the TSM backup service would be to stop charging for the service and to provide it to the institute for free. o We currently recover $1M a year in TSM charge, but a majority of that amount is IS&T charging ourselves. o In response to a question about the whether the cost we charge for TSM service correlates to what it costs us for the service, Garry stated that that the total dollar amount we recover from TSM bears a strong correlation to the actual costs we bear (staff, software licensing, and equipment). The manner in which those costs are distributed across the three tiers of service, however, doesn't necessarily correlate to the cost of providing that tier of service, i.e., the "Basic" service is provided gratis, and we subsidize those costs via the rates of the Standard and Enterprise storage tiers. Likewise, the $15/month cost of the Standard service was derived more from the historical service cost ($7.50/month) based on what size of increase the community would accept than it was on a rigorous analysis of the cost of the Standard service vs. Enterprise. o IS&T will develop a specific recommendation regarding TSM pricing/cost recovery and bring it to the IT Governance Committee for approval. Garry was asked if there has ever been a situation where someone not using our service had a problem and could not recover. He said this definitely has happened and they come to us for us to try and help them get their data back. He was also asked if he has investigated using Drop Box as a back up alternative and he said that he was reluctant to use it as a back up solution as it was more of a file sharing system and did not think it would work effectively as a back up alternative. o Virtualization On the weekend of January 5th, Garry s team virtualized the application server tier of MITSIS (where the J2EE and Oracle Forms applications execute); the MITSIS Oracle database remains on a physical, Sun SPARC/Solaris system; its virtualization will coincide with an upgrade to Oracle 11g, which is expected for late summer/early fall. Our Enterprise computing environment is approximately 85% virtualized. Our current VMware agreement needs to be renewed next year and we estimate the renewal cost will be $2.5M. The agreement will be a 3 year agreement covering us for FY13, 14, and 15. o There are currently approximately 2,500 CPUs worth of licenses in use: ITGC Minutes 20120117 final.doc Page 5 2/13/2012
o IS&T: 40% o Lincoln Lab: 50% o Remaining 10%: Sloan, CSail, Facilities, Libraries, Aero/Astro, Urban Studies, Media lab, among others. o Server Platforms We are currently using a blade server configuration in our data centers, which is enabling us to shrink our server footprint. Garry was asked if we were looking at Cloud Computing. o He said IS&T and the OGC are considering a contract with Google for collaboration tools for faculty and students- - at MITCET s request. Google collaboration tools are already being used by a number of faculty members and students across campus. IS&T to develop a recommendation regarding on the use of Cloud Computing and bring it to the IT Governance Committee for approval. The IT Governance Committee approved the Data Centers Roadmap Marilyn wrapped up the meeting: o The next IT Governance committee meeting is scheduled for February 8 th. o She told the committee that IS&T was combining the Customer Support and Computing Experience Roadmaps but that it would not be ready for the 2/8 meeting. o Possible topics for the 2/8 meeting are: Status of the Digital MIT approved projects. Review of Research Computing Progress & Plans. o She asked the members to contact her if there were any items they would like added to the agenda. o Marilyn also advised the committee that we are working on rescheduling the March 7 th IT Governance Committee meeting as several committee members have other commitments on that date. ITGC Minutes 20120117 final.doc Page 6 2/13/2012