#DefragCon 2013 Notes: Software is Eating the Datacenter – Anjan Srinivas, Nutanix
Software is Eating the Datacenter – Anjan Srinivas, Nutanix Emerging trends 1. Taiwan – x86 servers going the laptop way, mass produced 2. Consolidation 2.0 – programmable data centers via virtualization 3. Server-side Flash – TBs of solid-state on servers 4. Amazon – utility consumption is dictating how services will be delivered and the user experience (ease of access) Cloud-generation systems converges compute and storage and flatter datacenters that scale out This is accomplished through cheap hardware and software that is built for failure Software-defined data centers are built with hypervisor controllers, no hardware dependency, and nothing factory-defined Hypervisor-resident controllers that are software provisioned, not on bare metal, and dynamically tuned based on DC needs Nothing factory-defined – allows for late-binding of typical hardware services – RAID, etc. Ease-of-use is required to make this all happen Need of consumer-grade frontend, API-driven backend Multi-device, flat UI design Automation via APIs Orchestration between APIs Perhaps going deep with OpenStack, VMware vCAC, Microsoft SCVMM
Software is Eating the Datacenter – Anjan Srinivas, Nutanix
- Emerging trends
- 1. Taiwan – x86 servers going the laptop way, mass produced
- 2. Consolidation 2.0 – programmable data centers via virtualization
- 3. Server-side Flash – TBs of solid-state on servers
- 4. Amazon – utility consumption is dictating how services will be delivered and the user experience (ease of access)
- Cloud-generation systems converges compute and storage and flatter datacenters that scale out
- This is accomplished through cheap hardware and software that is built for failure
- Software-defined data centers are built with hypervisor controllers, no hardware dependency, and nothing factory-defined
- Hypervisor-resident controllers that are software provisioned, not on bare metal, and dynamically tuned based on DC needs
- Nothing factory-defined – allows for late-binding of typical hardware services – RAID, etc.
- Ease-of-use is required to make this all happen
- Need of consumer-grade frontend, API-driven backend
- Multi-device, flat UI design
- Automation via APIs
- Orchestration between APIs
- Perhaps going deep with OpenStack, VMware vCAC, Microsoft SCVMM