This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>

If you need help, please click here:

Huawei Cloud Storage Shines at CERN– UDS Cloud Storage Project for the European Organization for Nuclear Research

Background

CERN, headquartered in Switzerland, is the world's largest particle physics research organization, boasting over half the scientists in the international particle physics community as its members. It is the cradle of numerous scientific breakthroughs since it was officially founded in 1954. These include the world's first proton-proton collider in 1971, discoveries of neutral currents in 1973, and the W and Z particles in 1983, to name just a few. On July 4, 2012, the laboratory declared that two general-purpose experiments conducted on A Toroidal Large Hadron Collider (LHC) Apparatus (ATLAS) and the Compact Muon Solenoid (CMS) had revealed the presence of a new subatomic particle with a mass of roughly 125 to 126 Giga-electron-volts (GeV) and the expected profile of the long-sought after Higgs boson, which is thought to give mass to all matter. Ironically, this discovery cost Stephen Hawking a $100 bet. Hawking, the preeminent physicist and only authentic celebrity in the hard sciences, and the iconic author of A Brief History of Time, had previously doubted the existence of the Higgs boson.

Key Challenges

The LHC produces data streams of up to 6 GB/s which all need to be stored. These massive data streams are currently forked and stored using more than 140 computing nodes (with 250,000 CPUs and 150 PB of storage). These nodes are distributed throughout the world. In August 2012 alone, CERN transmitted approximately 4 PB of data and 2.6 million files from CERN headquarters to other tier-1 data centers.

As the scale and level of experiments improved, experimental data generated at CERN has shown fast and substantial growth. At present, CERN's local data center has over 25 PB data stored in hard disks, with available hard disk space of less than 40 PB. The entire system architecture is unable to support the large and increasing quantities of experimental data. CERN has an additional 75 PB of data stored in a tape library, which is difficult to query and creates a complicated system of tape management and maintenance.

To address data storage issues, CERN needed a new data storage system that could:

• Store at least 20 PB of new data each year.

• Provide high throughput that can support storage of new data at up to 6 GB per second.

• Offer high reliability to safeguard experimental data that may only occur once, because repeated experiments rarely produce the same information twice.

• Support efficient cross-regional sharing of experimental data to facilitate collaboration among CERN's worldwide research partners.

Solution

CERN openlab's fourth phase heralded the launch of more in-depth cooperative activities, for which the storage system is a critical component. The projects involve automation and controls, databases, networking, and platforms.

As a storage expert, Huawei is committed to developing innovative next-generation cloud solutions. Since joining CERN openlab as a contributor, Huawei has developed a cloud-based Universal Distributed Storage (UDS) system to meet the challenges of exascale data storage (one exabyte = 1,048,576 terabytes).

In addition to this scalability, the Huawei UDS cloud storage system features tremendous reliability and broad compatibility. It also employs dense nodular distribution and is based on energy-efficient ARM architecture. UDS enables storage and sharing of Big Data on its mass object-based storage infrastructure, which integrates object-based storage, a P2P distributed storage engine, and clustering applications. The system also intelligently adjusts the workload at each node to break through previous performance bottlenecks, so system performance grows evenly with each expansion of capacity.

The Huawei UDS cloud storage system features reliability to the tune of 99.999999999 percent (that's eleven nines), while facilitating redundancy across data centers. It also ensures data security throughout the system's lifecycle. It does this by utilizing multi-layer protection at the node, storage, and data center level through multiple copies, erasure codes, and multi-layer encryption verification involving the interface, storage, and transmission processes.

This solution is also supremely cost-effective. Maintenance costs are curtailed through the use of desktop hard drives, intelligent and automatic fault recovery, energy-efficient ARM chips, effective cooling technology, and various other self-maintenance measures that enhance storage availability and reliability. Compatibility with S3 interfaces, NFS/CIFS protocols, and mainstream backup software ensure a comprehensive public cloud solution that is easy to use and meets the storage access requirements of different applications.

As one of the tier-3 data nodes of the LHC system, the Institute of High Energy Physics (IHEP) of China provides data storage, analysis, and interaction services. At the beginning of 2013, the UDS system passed tests conducted by IHEP, which confirms the UDS system's effectiveness in high-energy physics. The UDS system's multi-data center ensures that data produced from different regions is processed in the same resource pool, which facilitates cooperation and collaboration among CERN's global research institutions.

Customer Benefits

In early 2012, Huawei delivered its UDS cloud storage system to CERN, with installation and benchmark performance evaluations completed three months later. Its read/write performance and scalability proved excellent in large-scale data environments. On the whole, these scores were deemed worthy of the very organization that gave birth to the World Wide Web.

CERN is pushing further into the strange boundaries that define the quantum realm, and is now undertaking more in-depth studies to identify new particles in order to explore other enigmas of the universe. Huawei's UDS cloud storage system will have further opportunities to distinguish itself, on a scale that you'd have to be a particle physicist to really grasp.