Network non-replicable disks in the cloud: why, how, and most importantly — what for?
summary
Overview of a new class of network-attached non-replicated disks offered in K2 Cloud: explains why such disks appear alongside traditional replicated cloud volumes, what technical design and implementation details distinguish them, and practical considerations for when and why to use them.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 1:58Non-replicable cloud disks
- 3:43SDS disk problems
- 5:10Local SSD as alternative
- 6:56Network SSD as compromise
- 7:59Non-replicable disks concept
- 9:25Linux kernel performance
- 10:49What is SPDK
- 12:58SPDK architecture
- 15:45NVMe-oF and SPDK client
- 17:53RDMA and NUMA in SPDK
- 20:22SPDK test results
- 22:50VM performance
- 25:20SPDK inside VM
- 26:26Future: UBLK and VFIO
- 27:07Disk use cases
- 30:23Why non-replicated
- 30:45Data migration issues
- 31:49Windows and VM problems
- 32:51Kernel polling usage
- 34:16Manageability and fault tolerance
- 35:41SPDK controllers state
- 36:45Encryption and security
- 37:06Summary and outlook
- 38:31Q&A
related talks
42mDistributed Storage in the Cloud
12mNelson Nahum | Google Next 2017
27mElastic cloud bursting with Virtual Kubelet and KIP
22mMulti-Purpose Cloud Filesystem
14mA Cloud-native Architecture for Replicated Data Services
21mHybrid Cloud - nothing to worry about! 🌥️
36mResilient Multi-Cloud Strategies: Harnessing Kubernetes, Cluster API, and Cell-Based Architecture
37mCloud Storage 101