2. Boyan Krosnov
● Chief of Product
● Cloud architect
● "External technology"
● Previously service providers, networks, packet
processing (SDN)
● Algorithms & data structures
2
3. StorPool
● Fast and efficient software-defined storage system
● Used by MSPs, Cloud service providers, Hosting
companies and in private clouds
● Started in 2011 to solve storage for cloud service
providers
● Clean slate design - scale-out, API-controlled, end-
to-end data integrity, CoW 3
4. StorPool
● Most deployments are with KVM
○ some Xen, some VMWare, some Hyper-V
○ some bare metal / dedicated server
● Deep integrations into OpenNebula, OpenStack,
OnApp, CloudStack, Kubernetes
4
6. Why build a cloud?
For service provider (hosting, public cloud, MSP)
● control & customization
● Price/performance optimization
● Service quality
For web and mobile apps (production private cloud)
● control & customization
● cost optimization
● API/automation
For software developers (test/dev private cloud)
● Self-provisioning of resources (VMs, containers) developers
● API/automation
● Cost optimization and low latency
6
7. Contrast requirements
Best performance I can get at reasonable cost
(private cloud for web/mobile application)
vs.
Good performance and low $/GB for desired
performance level. Market leading, but not too
expensive. (service providers)
vs.
Lowest $/GB (e.g. archival use-case)
20. Hyper-converged: example
Private cloud for production public-facing application (high-end
VDI)
● Independent racks design. Each rack has…
● Top of rack switches, 2x 25GbE to each node
● 38 compute nodes
● Each with 2x 3.84 TB NVMe SSDs
● 88 TB usable space (3-way replication, checksums, CoW,
safety)
Strong case for all of
● SDS
21. When hyper-converged
Hyperconverged:
● Green-field
● KVM virtualization
● Building pods / small availability zones
Stand-alone storage system:
● Retrofit
● Other hypervisors: Xen, VMWare, Hyper-V
● Requirements for independence of infrastructure (I don't
want an issue in my KVM environment to affect my VMWare
environment)
22. StorPool-OpenNebula
StorPool for OpenNebula/KVM
● Uses StorPool native driver in host OS
○ block devices in /dev/storpool/volname are raw disk
images for qemu/KVM
○ data management - e.g. create, delete, clone, resize
snapshot - API calls
● Implements a datastore driver for primary storage
○ https://github.com/opennebula/addon-storpool
OpenNebula/VMWare - iSCSI (scale-out, multipath) and VMFS
23. Set performance expectations
1M IOPS per node
https://storpool.com/blog/the-iops-challenge-is-over-storpool-
holds-the-new-world-record-13-8-mln-iops