Archive for May 10th, 2006

[CSG Spring 2006] Last Data Center panel of the day

Klara says that Wisconsin has been using VMWare to achieve server virtualization, using VMotion. They’re achieving some economy of power and cooling that way. One comment from the audience is that it’s harder to monitor virtual servers. Shel reports that they’re seeing a lot of uptake on use of VMWare offering, especially for departments that want to have staging and test servers available.

This panel exceeded my attention span, mostly I think due to a long day after a short night’s sleep. So I missed a bunch of no doubt interesting points.

The day ends with some discussion of whether it makes sense for CSG to broker some experiments with multiple institutions working together to help each other do backup and data resumption services.

Technorati Tags: ,

[CSG Spring 2006] Sue Riseling – University of Wisconsin Access Control

Sue is the Chief of Police for the University, and also a Vice Chancellor.

They did a physical security survey of every building on campus – ended up being just under 2500 spaces surveyed. The State of Wisconsin bids every access control system as a separate bid. They had seven different access control systems for 345 buildings.

With new federal regulations they told the state that they wanted to do a single access control system. The first priority was to secure things like select chemicals, nuclear reactor, etc. That’s done, now moving on to next priorities.

Now they have a single system mandated. Using Andover system. Departments have local control over access and hours, but the central control is in the hands of the police.

Each department has to fund the system themselves.

They have fingerprints for each cardholder in high risk areas. Cards are issued by the police. So far they’re not integrated with the HR system to automatically know when people leave.

The web site for this is at http://www.uwpd.wisc.edu/Access%20Cards%20and%20Control.html

Sue goes on to tell a couple of hair-raising stories about mistaken identification of dead kids based on borrowed or forged driver’s licenses used for – she uses that as an intro to talking about authentication and authorization as being a central problem of our time that we need to be thinking about. That, of course, comes as no surprise to this crowd.

Technorati Tags: ,

[CSG Spring 2006] Key lessons from schools that have developed new data centers

Jim Pepin – USC is moving into an old garment factory downtown LA. They’re building on the fourth (top) floor so that there’s no possibility that the data center can be innundated by water from floors above.They’re dedicating about 5,000 sq ft to administrative computing (payroll and personnel). There will be two data prep rooms for building stuff out – one 800 sq ft, one 300 sq ft. This will all be 24 inch raised floor. There will be about 4500 sq ft for high perfomance computing with 7 50 ton air conditioners, with 1.2 megawatts of air and power.

Jim notes that people under 40 aren’t used to having water under the floor in a data center, unlike guys that grew up with mainframes, and that’s creating some debate in the data center design world.

Shel from Berkeley is talking about their new data center. It took them 7 years to justify need, 2 years to get budget approval, 18 months to build base building, 6 months to build the data center, and 1 weekend to move in. The facility is on the periphery of the campus, and researchers indicated that if they couldn’t come to touch their machines that would be an issue – but they never do actually come. It would have been cheaper to build further away.

They built redundant power (2500 KVA).

They’re on the third floor of a building, which required a unique seismic design.

Ten percent of the cost was moving out of the old building. They planned for 75 hours downtime, it ended up being 62 hours for about 400 servers.

$11.7 million for data center and move. base building $23 million. Electrical service was $4 million.

Shel got signing authority for any data center space on campus. That allows them to move more folks into the central data center. There will still be some departmental data centers, but fewer of them.

They went to 18 inch raised floors in most spaces rather than 24 inch. All the cabling is in ladder racks above, tied down.

Each four cabinets have 96 fiber and 96 copper connections. There’s conversation on the back channel about whether that’s enough network capacity in an era of high density blade servers. Walter notes there that “for an IBM 14 blade chassis, we’d want to have something like 37 copper ports per chassis: We’d use switches for uplinks (8 ports) but for iSCSI SAN, we’d want each blade to have a direct connection to the switching backplane (28 ports) plus one management port.”

Lessons – Don’t rely on your campus design tema or architects. 65-80 wattt per sq ft is plenty – if you have a separate ultra high density room. Design for expansion or modular – you will need it. A full load bank for testing is a good thing. Standardize, standardize, standardize. Design for lights out. Use the move opportunity to plan change.

Michigan – Is building a new 10,000 sq ft data center which will also house Internet2. A big chunk of the electrical service is outside the building. They’re building 24 fiber connections to each half-row of racks. Clusters will have switching built into the racks for their own distribution.

east half of room is “server class” space – low density 160 watts per sq ft. Going with Liebert XD power stuff. The high density part of the room will be Liebert XDO planning average of 240 watts per sq ft across the entire space with ability to go to 300 in small areas.

2 mw for equipment, 2 mw for cooling, lighting, etc.

There’s some discussion about whether people are providing UPS to research clusters or not.

Michigan is using flywheels to provide complete equipment power until generators can come up – flywheel can carry load for about 25 seconds, generators are guaranteed to be up in 12-13 seconds. They decided not to go with batteries for the leaking and exploding reasons.

There will be ~ 400 fibers on three paths between campus and the facility.

Took three years of discussion with strategic deans to get them to agree that they would stop building local data centers and move into this one. They’ll be subsidizing about half of the operating costs centrally. Electricity will be paid for by customers – priced to drive behavior. Shel notes that at Berkeley they charge $8 per rack unit per month, all inclusive. Shel is willing to provide the cost model if people are interested in seeing what’s included.

Harvard – They were faced with an eighteen month eviction notice. They were looking for 4-5k sq ft – they ended up with 5k sq ft of computer floor, plus a staging area. Facility designed for central administrative computing, not research computing. Now FAS is bringing in lots of research computing.

They built the room at 55 watts per sq ft. The back channel thinks that sounds awfully low. They have dual 450 Kva UPS, and they’re adding a third.

Technorati Tags: ,

[CSG Spring 2006] Data Center Survey Results

Bill Clebsch is talking about the results of the data center survey of CSG folks.

There was 100% response from the CSG schools.

60% of the data centers were over 20 years old. 36% answered their facilities are 11-20 years old. That means that 96% are over 11 years old!

Intel says it’s not worth investing in data centers over 8 years old. Intel is fond now of buying old supermarkets for data centers because they have high ceilings. They’re also using two foot raised ceilings and cooling to 66 degrees instead of 68.

39% have 10-20k sq ft of usable space; 36% have 5-10k sq ft.

Bill notes that the older faculty, who bring in much of the money, have an “anthrpomorphic” relationship with their computer – want to be able to touch it. Younger faculty don’t care.

There’s a 50-50 split on wehther facilities are centrally funded or have rates for usage. It’s hard to incent researchers to come into the data center – they perceive the space in their own buildings as “free”.

The majority have no required standards in place for data center equipment.

Everybody’s looking for more – floor space, power, etc

50% are planning expansions of 51-75% more floor space.

Most people are expecting 11-20 years of use from new facilities.

Many people are using consultants to help plan – it’s important to find an consultant that understands high performance computing.

Current allocations for research and academic computing are low in data centers (64% allocate less than 25% of existing space), but future needs are thought to be very high (36% plan to allocate 26-50%, 29% plan to allocate 51-75%).

2/3 of respondents have some level of colocation service. 1/3 offer a service where faculty buy into large “condo” clusters.

There’s some discussion of whether there are ways for groups of institutions to collaborate on backing each other up for disaster recovery.

Technorati Tags: ,

[CSG Spring 2006] Data Center Futures – Context Panel

We’re here a the Pyle Center at the University of Wisconsin for the spring CSG meeting.

Bill Clebsch from Stanford is leading the Data Center Workshop.

The first panel, being introduced by Jim Pepin from USC, is setting the context.

There was a meeting two weeks ago with NSF about cyberinfrastructure in a campus level. Twenty years ago we were the people supporting high end folks on campus. Over the last ten years we’ve gotten to a point where we dont change anything quickly. Whole lot of stuff going on on campus around clusters and networking. People in departments want to build clusters – what will that drive in terms of campus infrastructure for housing those.

Previously there were no line items in grants for facilities and operations. The sweet spot used to be 32 node clusters, but now it’s 128 or 256 nodes. If faculty want to get funded they need to be at the cutting edge of research, and the 32 node cluster in a grad student office isn’t that. There’s been a lot of pushback and rethinking at the funding agencies.

Kevin Moroney – You could suffer “death by [NSF] Young Investigator Awards” with 16 node clusters.

Jim notes that science has become team science, which drives more computing.

The back channel has some chatter about water-cooled racks in machine rooms.

Penn State has just invested in some water-cooled racks.

Some discussion about power consumption – people are planning 18-24 kW per rack for high performance computing and 4-6kW for an enterprise computing rack.

Walter points out in the back channel that “APC has a ‘half rack CRAC’ where you have a ‘real’ system in the rack “hut”. he hut is a set of back to back racks which has a roof so that heat is contained in the hut and the in-rack AC units suck in that air and exhaust room temp air. A new CMU research facility that is using the APC system has a writeup at http://www.pdl.cmu.edu/PDL-FTP/News/newsletter05.pdf

Jim and Kevin note that in colo offering if you charge by rack then people will stuff the racks as full as possible, raising the kilowatt per rack average in the machine room.

Jim notes that USC has rented some colo space in a downtown facility for mirroring for disaster recovery.

There’s a bunch of discussion about offsite backup, disaster recovery, and how campuses prioritize what services will be brought back in which order.

Theresa from MIT notes that for key communication services (web, email) it’s good to have geographically distributed load balanced services, so that one location can go down and service is only more heavily loaded.

Patrick from MIT says that funding bodies are starting to hold campuses responsible for recovery of research data, so if the data center is holding research data it’s another thing to worry about.

In a brief discussion on power infrastructure (most campuses here have multiple power routes into the data center) Jim says that USC is always being compared to Disney for what they do – Steve Worona says that’s a Mickey Mouse solution :)

Theresa says that the financial systems can actually stand some outage time, but it’s the communications infrastructure – web and email – that is truly critical to keep running, along with the network. That’s the lesson of 9/11.

About a third of institutions here report that they’ve had a total data center outage within the last three years.

Technorati Tags: ,


subscribe

Pages

Latest tweets

interesting links

What I’m listening to

May 2006
M T W T F S S
« Apr   Jun »
1234567
891011121314
15161718192021
22232425262728
293031  

Follow

Get every new post delivered to your Inbox.