History
In this section:
This section presents the evolution of GridUnesp year by year, from its conception to the present day.
The video below summarises the history of GridUnesp since its creation.
2008 - Infrastructure Announcement
UNESP announces the largest high-performance computing infrastructure in Latin America.
Total investment: R$ 8 million
Ministry of Science and Technology: R$ 4.4 million (via Finep/Fundunesp)
UNESP: R$ 3.6 million
Acquisition: Competitive bidding process won by Sun Microsystems of Brazil.
Original hardware specifications:
Processors: Quad Core Intel Xeon 5400
Processing cores: 2,048
Capacity: 33.3 teraflops (trillions of calculations per second)
Composition:
Central cluster: 256 processing nodes on the UNESP campus in São Paulo (Barra Funda)
Six distributed clusters: 16 nodes each, on the campuses of:
Araraquara
Bauru
Botucatu
Ilha Solteira
Rio Claro
São José do Rio Preto
Connectivity:
High-speed connection to the North American Internet2
MetroSampa network linking institutions in the São Paulo metropolitan area
ANSP/RNP/Florida International University network between São Paulo and Miami
FAPESP’s KyaTera network connecting São Paulo to the interior clusters
Benefits for the scientific community:
Remote access for large-scale computational operations
Support for areas requiring processing and analysis of large data volumes
Reduction in research computation time
Performance leadership with low energy consumption
International partnerships (Open Science Grid – USA, European and Asian grids)
2009 - Start of Operations
GridUnesp begins operation, representing a Brazilian milestone in:
High-performance computing
Long-distance optical networks for scientific research
Equivalent capacity at the time:
10,000 personal computers
Storage equivalent to an 18-storey tower of stacked DVDs
Communication speed 5,000 times faster than home broadband
Training:
First course offered at the Center for Scientific Computing of UNESP (NCC/Unesp), covering:
In-person format with online broadcast
Basic operation of GridUnesp
Compilers and development tools
Parallelism and optimisation techniques
2010 - Expansion and Outreach
MasterClass Event:
Reception of high-school students during an international event, including:
Presentation on the GridUnesp infrastructure
Visit to the cluster at the Datacenter
Year-end metrics:
26 registered scientific projects
Areas covered: Biophysics, Biology, Materials Science, Computing, Physics, Geosciences, Chemistry
90 registered members
10 million processing hours
2011 - Accelerated Growth
New call for project registration
12 new scientific projects
41 new users
More than 20 million processing hours
2012 - International Recognition
Higgs Boson Discovery:
Scientists connected to GridUnesp take part in the discovery at CERN.
CERN Visit:
A delegation of six physicists from the European Organization for Nuclear Research visits GridUnesp, related to Brazil’s bid for associate membership.
2013 - Consolidation and New Services
HPC Administrators Workshop:
Support for technical teams of research groups
Training based on GridUnesp’s operational experience
Creation of a professional network for knowledge exchange
Intel Software Conference:
Parallel programming focused on Intel Xeon Phi
Hands-on activities
ANSPGrid CA:
Deployment of the ANSPGrid CA Certification Authority at NCC/Unesp:
Issuance of X.509 grid certificates for researchers
Public-key cryptography for the São Paulo scientific community
Compliance with International Grid Trust Federation (IGTF) guidelines
2014 - Accessibility
Creation of a web portal to facilitate researchers’ access:
Intuitive interface for a complete workflow
Developed with free and open-source code
Accessible via browser
Press visit:
Trainee reporters from the newspaper Folha de S. Paulo visit the infrastructure during the Science and Health Training Programme.
2015 - Expressive Milestones
GridUnesp reaches:
More than 60 registered scientific projects
More than 300 registered users
More than 140 million processing hours
2016 - Upgrade and Scientific Achievements
Gravitational wave discovery:
LIGO (Laser Interferometer Gravitational-wave Observatory) announces the detection of gravitational waves. GridUnesp was:
The second largest contributor to processing, behind only Fermilab
7.5 million hours processed (12.3% of the total in OSG)
Acquisition of new equipment (upgrade):
Funding: R$ 2.2 million from Finep
Supplier: Huawei
New hardware specifications:
Processing:
56 servers (1,568 cores)
2 Intel E5-2680 v4 processors (2.4 GHz) per server
28 cores per server
4 GB per core
40 GB/s network
60 TFlops (theoretical)
Management:
4 servers with 12 cores each
40 GbE network
Storage:
288 TB raw (216 TB effective)
NL-SAS 4 TB disks with Lustre technology
4 control servers
Internal network:
40 GbE interconnecting processing nodes
2 switches with 32 ports of 40 GbE each
Upgrade gains:
Solution for saturated infrastructure
15-fold reduction in energy consumption
New documentation:
Release of the reformulated and updated user documentation.
2017 - Upgrade Completion
Final capacity:
Feature |
Before |
After upgrade |
Processing |
23 TFlops |
77 TFlops |
Storage |
136 TB |
288 TB |
CPU Cores |
2,048 |
3,104 |
2018 - Expansion and Statistics
ANSPGrid CA:
Certificates now issued to research institutes across all of Brazil
Certificate Policy updated (version 2.0)
Statistics:
More than 80 registered scientific projects
500 registered users
170 million processing hours
More than 200 publications using GridUnesp resources
2019 - Science Outreach
Special MasterClass event:
Celebration of the International Day of Women and Girls in Science:
Around 100 girls from public and private schools
Presentation of Physics topics (Standard Model)
Visit to the cluster at the Data Center
2020 - Pandemic and Accelerated Demand
Operations:
Team working in remote mode
Remote maintenance of almost all systems
Demand:
More than 100 registered projects (19 in 2020 alone)
More than 650 registered users (75 in 2020)
Around 13 million processing hours in the year
2021 - Remote Continuity
Maintenance of remote working regime
67 new registered users
More than 11 million processing hours
ANSPGrid CA:
Root certificate renewed for a further 10 years
Certificate Policy updated (version 3.0)
2022 - Return and Consolidation
Return to in-person work
Rare idle periods for the cluster
More than 1,000 daily jobs in the partition (≈50% running simultaneously)
Release of user documentation version 2.0
2023-2024 - Continuous Expansion
New equipment:
New 400 TB storage system with the installation of a more modern and reliable system
GPU nodes to meet demands from machine learning and artificial intelligence
Community:
More than 140 registered projects
More than 850 registered users
Continuous scientific output across multiple areas of knowledge
2025 - New Milestones
Surpasses 1,000 registered users
Intensive GPU use: more than 50,000 jobs in one year
More than 160 scientific applications and corresponding versions installed since 2017
See also
Statistics - Updated usage data
Infrastructure - Current cluster configuration
Publications - Scientific output using GridUnesp
COVID-19 - Impact of the pandemic on operations