Salsa-DR Working Group
February 1, 2008
 
**Attending**
Don McLeod, Cornell University (chair)
Bob Dixon, Ohio State University
Tim Crawford, Stanford University
Steve Olshansky, Internet2
Dean Woodbeck, Internet2 (scribe)
 
**Internet2 Member Meeting**
The working group discussed how to approach the DR BoF at the Spring Member Meeting (April 21-23 in Arlington, Virginia). The exact date/time for the session is not yet established.
 
One option would be to bring in a vendor (or vendors) of interest on some specific topic. Last time, there were peer experiences presented, particularly by Duke and Stanford. Or it could be a round-robin of ideas.
 
There is also the opportunity for people from this group to make presentations. If someone has a topic to cover, even just 5-10 minutes worth to describe a project, that would help stimulate conversation. Don is looking for volunteers for this purpose.
 
Don will send a note to the email list soliciting feedback.
 
**Sun Black Box**
There was a general discussion about a Sun project that takes a shipping container (20’ x 8’ x 8’) and turns it into a data center – complete with power, racks of computers, cooling and other necessities – that is self-contained. Sun can drop-ship these anywhere. Stanford’s linear accelerator has one and is buying another; mainly to avoid having to build another building. However, these seem to have a lot of potential for disaster recovery, perhaps with institutions forming a cooperative to purchase one.
 
An institution did a presentation at EDUCAUSE about a similar product, which they used for DR. That school set up a location with utilities and connectivity already available and this device can be backed into a parking space.
 
Don mentioned that this might be one possibility for a presentation at the Member Meeting – having a technical person from Sun discuss the details and requirements.
 
**Staff emergency planning agreement**
Don reported that Cornell’s emergency planning involves both infrastructure and defining staff responsibilities. For example, the university has identified those in key and essential positions and who would need to be on-site during an emergency and who would work from home. The latter includes some requirements for home connectivity; that these people would need at least two means of communication (land line and cell phone, for example). He wondered if others have gone to this extent in defining staff roles.
 
At Ohio State, each computing staff employee is designed to a category (critical, standby, etc), which is noted on an ID card. In the case of an emergency, for example, this card would allow them though police lines.
 
Stanford has identified those who would provide leadership and those who are first responders, but haven’t gotten as specific as Cornell.
 
**Emergency Testing**
Bob Dixon from Ohio State emphasized the importance of testing equipment and protocols prior to an emergency. Ohio State had a power outage, for example, but there had been no testing concerning how long the batteries in the emergency power supplies would last. As it turns out, the batteries didn’t last long. In addition, the emergency generator ran, but a power switch failed. They then hired a company with generators and switches – they came in and that ran for a day or so.
 
Ohio State recently tested their emergency response system. Users received cell phone calls and text messages within 10 minutes. Calling all campus phones took longer – about an hour. Bob will inquire about Ohio State’s service provider and provide that information to Don.
 
Tim reported that Stanford just engaged a company, NTI Group, for emergency notification services. Stanford is loading data and hopes to test in 30 days.
 
Tim also mentioned that it is important to have a number of technologies in place to protect data and ensure service continuity, but that prioritization also becomes an issue. If there is an emergency and you have limited capacity (power or connectivity, for example), do you have a priority scheme to know what you can shut down and bring up later?
 
Don mentioned some difficulty with quantifying the need (and, thus, the budget) for disaster recovery capability. Tim suggested a question to ask is “why are we doing this in the first place?” That should lead to the important reasons for doing this.
 
**Data Center Power Recommendations**
Joe St. Sauver sent a note to the email list concerning a suggested project: developing recommendations for what data centers should have in terms of electrical power. Areas of interest might include provisioning multiple power feeds, run-times and testing for UPS, generator back-up, power transfer switches, EPO (emergency power off) configurations, and others.
 
A discussion ensured concerning what each institution is trying to achieve – where are the biggest pain points? If there are repeating themes, that would provide a list of topics to work from.
 
Tim mentioned that Stanford is just embarking on a new campus for the graduate school of business and he could share how they are approaching things from a DR point of view.
 
**Next Call March 7, 2008, 1:00 p.m. EST**