This page contains a Flash digital edition of a book.
is larger, but still sitting in the same physical space. We approached DDN to integrate storage and servers in one solution,’ he said. According to DDN’s Laura Shepard, the


company’s SFA products ‘are the most dense product in the industry so you have higher capacity for your footprint.’ At Triumf, she said, they had been developing ‘in-storage processing – the ability to expose volumes on the storage controller itself and embed data- intensive applications on the storage controller’. Tis means that data-intensive applications can be co-resident with their data so there is less latency associated with the round trip: ‘Tis is very significant when you are having primarily a data transaction, as we are seeing here at Triumf. From a convergence standpoint, you are eliminating components such as switches, ports, and licensing – which reduces the acquisition cost as well as the management cost.’ By eliminating external components, the system also achieves a smaller footprint, she said.


The Atlas detector is the largest of its type in the world at 46 X 25 X 25 metres; (the people in the diagram are to scale). It produces 1PB of raw data per second, but triggering and filtering systems can reduce this to more than 100MB per second


➤ at a premium at Triumf, and in Vancouver in general, so it was critical to go with the highest density possible.’ DDN offers some of the highest density storage, which was one of the reasons that Triumf opted for the company as its supplier. All the Tier 1 centres have to be up and


running when the experiment is collecting data because the buffer space at Cern is not infinite, it is only a few days. So it is important that that data is farmed out to the Tier 1 centres as soon as possible. Triumf has an end to end connection with Cern through the LHC’s own optical private network, a multi-Gigabit per second link distinct from the normal research network used by the universities. Te providers are Géant in Europe and the Canadian Network for the Advancement of Research, Industry and Education (Canarie) in Canada.


Not compute but storage capacity is critical Raw data goes to tape directly. But the role of a Tier 1 centre is not just archival. It has to ship the right data to the Tier 2 centres around the world. It is in these centres that the analysis and physics are done. Triumf holds the data for analysis on disk. Tis data, Tafirout pointed out, ‘is critical; it’s urgent. Tis is where DDN comes in. We need really high-performance disk storage to handle the derived data that users need access to, around the clock, from around the world. Tis performance is critical.’


24 SCIENTIFIC COMPUTING WORLD Triumf itself is not a Top500 site; indeed,


none of the Tier 1 centres are in the Top500. ‘It is not the compute capacity, but the storage capacity with the highest performance that is critical,’ Tafirout said. And, he continued, as a Tier 1 centre, it’s constant evolution. ‘We have to replace old technology, and keep up with what is available.’ Te technology dating from 2009, however, was a bulky storage system. Triumf bought it because it had a good price performance and the organisation had enough space in the server room. ‘Now the data centre


Innovating storage at Cern At Cern too, change is the order of the day. Cern built most of the soſtware infrastructure itself to support its data storage operations. ‘For the physics data, we use two systems: Castor and the other is much newer, called EOS,’ Lamanna said. ‘So you inject files and they go to tape in Castor. Te users see an infinite file system, in the sense that we have some disks, but it is just a cache and the users do not know about the details of our infrastructure. We catalogue the file and then it flows to tape and eventually is removed from the disk.’ If it is still on the disk and the user wants it back, then there is essentially no latency. If the file is on tape – and in the long run, as Lamanna explained, the majority is on


In July 2012 Atlas and CMS announced they had independently found the Higgs boson @scwmagazine l www.scientific-computing.com


CERN


CERN


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40