MASSIVE STORAGE

The second Students Information Storage Technology Competition is now open for registration, with a total prize pool of 30,000 RMB awaiting winners!

Competition Introduction

Competition Overview

In the age of AI and data awakening, the challenge of information storage has re-emerged, pushing advancements in storage technology to create more efficient ways for humans to handle vast amounts of data.

To foster, select, and recommend top talents in the storage technology field and ignite students' innovative thinking and teamwork spirit, Huazhong University of Science and Technology is hosting the competition, with Huawei Technologies Co., Ltd. as the sponsor.

Eligibility

Full-time undergraduate and graduate students with valid student status are eligible to participate. If any team members are found not to meet the eligibility requirements during the evaluation, the team will be disqualified from competing or winning any awards.

Team Requirements

1. Each team may consist of up to 4 students (including 1 team leader) and may appoint 1 instructor. Each student can only join 1 team, but there is no limit to the number of teams from each university. Cross-university teams are not allowed.

2. Participants can create or join a team via the competition website after successful registration. The team leader is responsible for managing team members.

3. Teams not meeting the team requirements will be disqualified.

Competition Rules

1. Entries should demonstrate innovation or practicality.

2. For the open track, projects can include software or hardware, with a focus on storage technology and application design.

3. The competition is open-ended, allowing participants to work on their projects in their spare time. All submissions must be original and approved by the organizing committee.

4. Projects must be independently designed and developed by the team members. Plagiarism or submitting the same project multiple times is strictly prohibited.

5. Previously commercialized products or those involved in intellectual property disputes are ineligible.

6. Papers are not accepted as submissions for this competition.

Competition Timeline

The competition is divided into two tracks: Challenging Track and Open Track, allowing participants to choose one.

Challenging Track:

Open Track:

*Note: The competition timeline is subject to postponement and other changes, so please check the updates on the official website or messages on the group chat to follow the latest competition schedule.

Competition Topics

The competition is divided into a challenging track and an open track, and participants are free to choose their tracks, with a limit of 1 topic for the same track.

Topic 1: Optimal IO Scheduling for Magnetic and Electric Disks

Magnetic tape media offers a high capacity-to-cost ratio and is widely used for archival and data protection purposes. However, tape drives only support single-threaded read/write operations. Any reversal of tape direction, random seeking, or inefficient access patterns can reduce hardware lifespan and lead to delays of several seconds or even minutes, ultimately affecting tape performance and longevity. In specific read/write-associated scenarios, tape drives can use SSD caching, data arrangement, and other systematic techniques to optimize both performance and lifespan. However, in random read scenarios, tape drives typically rely on IO scheduling techniques to optimally sequence a batch of data access requests, significantly reducing seek delays and wear.

The task of IO scheduling for magnetic and electric disks is a classic Open Loop Traveling Salesman Problem (OTSP) optimization problem.

The key challenge of this problem is to develop an optimal scheduling algorithm that minimizes tape seek delays and wear, while considering variables such as the number of IO requests, IO request times, algorithm complexity, and other constraints. The goal is to create a combined optimization model to achieve optimal IO scheduling.

Topic 2: Storage System Networking in Large-Scale Networks

In large-scale AI training environments within data centers, the high number of compute and storage nodes requires a multi-layer Fat-tree network topology. During AI training, checkpoint operations are frequently performed to ensure quick recovery in the event of failures. These operations involve numerous compute nodes accessing the storage system in parallel, generating substantial storage traffic, primarily consisting of large file transfers.

The main challenge of this problem is to reduce the bandwidth loss caused by traffic collisions during storage access based on the characteristics of the traffic patterns. The goal is to improve the system's effective throughput under such large-scale network conditions.

Other instructions

The official website details the judging criteria, submission requirements, and competition rules.

Challenging Track Open Track

Prizes

The Challenging Track and the Open Track of the Competition will be judged separately, and the Organizing Committee will issue unified award certificates to reward the winning teams, with specific awards as follows.

Awards for the Final of the Challenging Track

Awards for the Final of the Open Track

In addition, a “Storage Cup Award” will be given, with a bonus of 10,000 RMB, along with Huawei’s exclusive opportunities, such as one-on-one mentorship with industry experts, Open Day at Huawei Research Institute, and waived test requirements for recruitment.

 Massive Storage 

Ready to take on the challenge? Head to the competition website and register now for MASSIVE STORAGE The Second Students Information Storage Technology Competition!

Note: Participants scan the QR code of the QQ group posted on the official website after registration, and the notification of important nodes of the competition will be informed in the group at the first time, so please Be sure to join the official competition group chat for updates.