Excited for the kick-off of Toreador: TrustwOrthy model-awaRE Analytics Data platfORm. Alpha@UNITO participates within CINI (National Interuniversity Consortium for Informatics). A 6.2M€ EU H2020 RIA ICT-16-2015 big data.
Marco Aldinucci is now director of the University of Torino research unit in the
National Interuniversity Consortium for Informatics (CINI).
Paolo Inaudi, MSc student from University of Torino, just completed a first version of a novel, high-performance libfabric provider for A3Cube In-memory “Ronniee express” network. Opensource under LGPLv3 on GitHUB
A sub-μs latency on pingpong up to 512 bytes. This means just a tiny overhead of the whole protocol over the bare metal latency. An excellent result.
High-Performance cluster - Nodes: Intel i7 cluster -------------------------------------------------- > ./fi_msg_rma_pingpong -p 1000 FORTISSIMO1 name size iters total time μs bandwidth latency/2 2_lat 2 100k 390k 0.18s 2.28 0.88 4_lat 4 100k 781k 0.17s 4.62 0.87 8_lat 8 100k 1.5m 0.18s 9.10 0.88 16_lat 16 100k 3m 0.17s 18.45 0.87 32_lat 32 100k 6.1m 0.17s 36.78 0.87 64_lat 64 100k 12m 0.17s 76.06 0.84 128_lat 128 100k 24m 0.17s 151.70 0.84 256_lat 256 100k 48m 0.17s 299.30 0.86 512_lat 512 100k 97m 0.19s 539.02 0.95 1k_lat 1024 10k 19m 0.02s 898.99 1.14 2k_lat 2048 10k 39m 0.03s 1357.87 1.51 4k_lat 4096 10k 78m 0.04s 1855.91 2.21 8k_lat 8192 10k 156m 0.07s 2245.00 3.65 16k_lat 16384 10k 312m 0.13s 2528.30 6.48 32k_lat 32768 10k 625m 0.24s 2696.31 12.15 64k_lat 65536 1k 125m 0.05s 2793.58 23.46
Low power cluster - Nodes: Intel(R) Atom(TM) CPU C2750 @ 2.40GHz ------------------------------------------------------------------ > ./fi_msg_rma_pingpong -p 1000 paradigm1 name size iters total time μs bandwidth latency/2 2_lat 2 100k 390k 0.29s 1.36 1.47 4_lat 4 100k 781k 0.29s 2.72 1.47 8_lat 8 100k 1.5m 0.30s 5.42 1.48 16_lat 16 100k 3m 0.30s 10.83 1.48 32_lat 32 100k 6.1m 0.30s 21.30 1.50 64_lat 64 100k 12m 0.29s 43.83 1.46 128_lat 128 100k 24m 0.26s 98.82 1.30 256_lat 256 100k 48m 0.27s 191.91 1.33 512_lat 512 100k 97m 0.30s 339.60 1.51 1k_lat 1024 10k 19m 0.04s 548.24 1.87 2k_lat 2048 10k 39m 0.05s 790.99 2.59 4k_lat 4096 10k 78m 0.08s 1022.20 4.01 8k_lat 8192 10k 156m 0.14s 1196.03 6.85 16k_lat 16384 10k 312m 0.25s 1310.76 12.50 32k_lat 32768 10k 625m 0.48s 1376.25 23.81 64k_lat 65536 1k 125m 0.09s 1412.90 46.38
Il Dipartimento di Informatica protagonista del progetto europeo “Paraphrase”, per programmare i computer del futuro
Per tutti i dettagli: www.paraphrase-ict.eu
Maurizio Drocco, a PhD student from the Alpha@UNITO parallel computing group is the recipient of a IBM Ph.D. Scholarship award 2015. Few awards of this kind are assigned every year.
Congratulations to Maurizio.
Dear Professor Aldinucci:
It is my pleasure to inform you that your nomination of Maurizio Drocco has been approved to receive an IBM Ph.D. Scholarship. This award is highly competitive and recognizes your student as well as the quality of your institution. In order to nurture the student relationship the IBM Ph.D. Scholarship program has asked Fabrizio Petrini to act as Maurizio’s mentor and principal IBM relationship contact for this award. The award check will be for the amount of $20,000 USD and will be made payable to the university for Maurizio to use as the student wishes for educational purposes – such as, but not limited to tuition, conference travel, or stipend. Again, IBM is delighted to provide your student an IBM Ph.D. Scholarship Award. Sincerely,
Dr. Lilian Wu
IBM PhD Fellowship Program
Maurizio Drocco and Claudia Misale have been selected for a quite competitive internship at IBM T.J. Watson laboratory in New York (6 months May-Oct 2015). Maurizio is going to work on parallel algorithms fro graph analytics within Fabrizio Petrini group. Claudia is going to work on low-latency interconnects within Carlos H.A. Costas group.
Two MSc students from parallel computing group “alpha” have been awarded with best thesis of the year price
Maurizio Drocco from MSc in Computer Science and Alessandro Ludovico Veltri from the MSc in Networking from the Interfaculty School of Strategic Studies, respectively, have been awarded with “University of Torino best thesis of the year 2012/13 price”. Only one student per master class is awarded every year.
Maurizio Drocco graduated with the thesis on “Parallel stochastic simulators in systems biology: the evolution of the species”, Alessandro Ludovico Veltri graduated with the thesis on “Analysis, setup and assessment of cloud platforms: addressing the defense cloud computing”.
Both of them graduated with “summa cum laude and honour”, for the exceptional quality of their curricula and their thesis .
Best students at UNITO 2012/13 (in Italian)
Programmazione Parallela: evoluzione e nuove sfide
di Marco Danelutto, Università di Pisa
Le recenti evoluzioni dei componenti di calcolo hanno messo a disposizione processori sempre più veloci e potenti che pongono nuove sfide dal punto di vista degli strumenti di programmazione. I classici strumenti utilizzati fino ad ora, infatti, richiedono un notevole sforzo di programmazione quando si vogliano sfruttare appieno tutte le caratteristiche dei nuovi componenti. In questo lavoro discutiamo alcuni recenti sviluppi nei modelli di programmazione parallela strutturata che permettono di sviluppare rapidamente applicazioni molto efficienti in grado di utilizzare tutte le nuove caratteristiche disponibili.
The focus of the RePhrase project is on producing new software engineering tools, techniques and methodologies for developing data-intensive applications in C++, targeting heterogeneous multicore/manycore systems that combine CPUs and GPUs into a coherent parallel platform. Data-intensive applications are one of the most important and commonly encountered classes of industrial application. Such applications are often potentially highly parallel and are a clear match to emerging heterogeneous parallel architectures. However, exploiting this potential effectively can be difficult: it is even harder to obtain good performance for parallel data-intensive applications than for compute- intensive applications, since many additional issues related to data management need to be taken into account. These including structuring the data to make it efficient to access and to process, placement/migration/replication of the data to allow fast parallel access, ensuring data consistency etc. The RePhrase project tackles these issues directly.
Temporary web-site: http://rephrase.weebly.com
Poster at Artemis Co-Summit 2015, Berlin, Germany