Speaker
Description
How can we understand a storage system when we cannot see what is happening inside it? Is the highest IOPS number always the best
result—and what does IOPS actually mean?
In this hands-on workshop, participants will explore a storage system as a black box. They will run simple experiments, change factors such
as request size and the number of simultaneous operations, and observe how these changes affect throughput and response time. Along the
way, they will learn how to interpret benchmark results, ask useful questions, and avoid common mistakes.
We will then connect these observations to Ceph, an open-source distributed storage system. Through practical examples, participants will
discover how Ceph stores data across multiple machines and how replication, networking, and coordination influence performance.
The workshop is designed for students curious about Linux, cloud computing, storage, or distributed systems. No previous experience with
Ceph or storage benchmarking is required. Bringing a laptop is optional, but participants who want to follow the hands-on exercises are
encouraged to bring one with a Linux environment or Linux virtual machine. Participants without a laptop can still follow the
demonstrations and discussions.
Any other info we should know?
Participants will learn how to:
- Design simple experiments to evaluate an unfamiliar storage system.
- Understand IOPS, throughput, latency, and their relationships.
- Identify misleading results and common benchmarking mistakes.
- Observe how workload changes affect storage performance.
- Understand the fundamentals of Ceph and distributed storage.
- Connect benchmark results to factors such as replication, networking, and coordination.
- Use an evidence-driven approach to investigate other computer systems.
Participants with a Linux environment can follow the exercises hands-on, while others can learn through the demonstrations and group
discussions. No previous Ceph or benchmarking experience is required.
Special accommodations
Please carry a laptop with linux env or a linux VM.
Session author's bio
Utkarsh Bhatt is a Software Engineer in the Ceph Engineering team (a member of the Openstack Engineering family) at Canonical. His team is responsible for producing the packages, charms, snaps, rocks and everything in between for the Canonical Ceph storage solutions. He graduated in 2020, and joined Canonical in May, 2022 after working for around 2 years in the Data Team at Huawei Technologies, Where he worked on DR for the flagship DoradoNAS storage solution.
| In Person Attendance | In-person |
|---|---|
| Please confirm that there are included headshots of all speakers in their profiles | Yes |
| Agree to Privacy Policy and Notice | I agree |
| Level of Difficulty | Intermediate |