Overcoming the challenge of virtual blindness Colin Richardson on365 Ltd Average IT utilization across all size Data Centers Source: Schneider Electric study Virtualization – why even bother? CPU efficiency Server consolidation Floor footprint Disaster recovery Labor cost Fast server deployment Customers having a “Virtualization First” strategy Source: IDC, Michelle Bailey, 2010: "The Changing IT Landscape: Virtualization, Cloud Computing and New Datacenter Requirements" …however Virtualization is no Silver Bullet either! Typical Virtualization Misunderstandings #1: Virtualization is only the concern of an IT group! WRONG! Your virtualization installation will – like any IT equipment you bring into your DC – impact the physical infrastructure of your DC The challenge is still here! Facility IT • Responsible for critical power and cooling supplies • Responsible for IT operations Applications/Services/Virtualization Compute Power/Storage Space/Network Access Servers/Storage Units/Networking Equipment Critical Power & Cooling Building/Security/Physical Space Business Value DEMAND SUPPLY Virtualization Management DEMAND SUPPLY DEMAND SUPPLY Applications Virtual Machines Hypervisor IT hardware Data Center Infrastructure Building Utility Data Center Infrastructure Management (DCIM) Virtualization impact on PUE and true energy savings Source: Schneider Electric WP118: Virtualization and Cloud Computing: Optimized Power, Cooling, and Management Maximizes Benefits #2: Virtual Machines are free! WRONG! Each VM started is associated with a cost of energy as a minimum Virtualization Sprawl • Starting new virtual machines is so easy, but… – Who controls if a VM are ever used? – Who verifies if power and cooling requirements are still met? • A VM running idle does consume energy! • Each kW in a DC costs £10-15k to build and £1000 per year to run Server power consumption VM impact on idle power 50% Server Idle 1 2 3 4 Number of VMs 5 #3: My VM management SW helps me optimize my energy usage WRONG! Your VM management SW is focused only on balancing the IT resources (CPU, Memory, Storage) Scenario 1: Initial State VM VM VM VM VM VM VM VM VM VM VM Server #2: 200 Wh VM Server #3: 210 Wh Server #4: 200 Wh VM VM VM Server #1: 200 Wh VM Server #5: 200 Wh VM VM Server #6: 190 Wh Total Consumption: 1200 Wh Scenario 2: 1st consolidation Server #1: 230 Wh VM VM VM VM VM VM VM VM VM Server #2: 230 Wh VM VM VM VM VM Server #3: 230 Wh VM VM VM VM Server #4: 200 Wh Server #5: 0 Wh (off – restart 5 min) Server #6: 0 Wh (off – restart 5 min) Total Consumption: 890 Wh Scenario 2: 2nd consolidation VM VM VM VM VM VM Server #1: 240 Wh VM VM VM VM VM VM Server #2: 240 Wh VM VM VM VM VM VM Server #3: 240 Wh Server #4: 140 Wh (idle – real time ready) Server #5: 0 Wh (off – restart 5 min) Server #6: 0 Wh (off – restart 5 min) Total Consumption: 860 Wh Dilemma: The Virtualization Jigsaw Its time to take off the blindfold…! ...and bridge the gap between Facility & IT Case Study: Power Chain Losses 13413 BOE 4254 tons CO2 10730 BOE 3401 tons CO2 8048 BOE 2551 tons CO2 6438 BOE 2041 tons CO2 PUE 5161 BOE 1633 tons CO2 45 45racks racks@ @10kW 8kW Siloed systems are the means to an end! Facility Management IT Management IT Tickets CMDB Server Management VM Management Cooling Control Power Management BMS Data Centre Infrastructure Management (DCIM) Business Value Applications Virtualization Management Virtual Machines Hypervisor IT hardware Data Center Infrastructure Building Utility Data Center Infrastructure Management (DCIM) Making IT Facility-Aware DCIM Power Cooling Space Network … •Server placement •Server health •Power capping •Power dependencies •Cooling dependencies •Event impact •… Virtualization Management CPU Memory Storage VM Allocation … VM Operator Making Facility IT-Aware DCIM Facility Mgr Power Cooling Space Network … Virtualization Management •Server power draw •Server configuration •VM allocation •Idle Servers •… CPU Memory Storage VM Allocation … DCIM becomes the centerpiece for operational efficiency DCIM Facility Management (BMS, Power Management, Cooling control, etc.) Capacity Planning Analytics Simulation Optimization Data Management Reporting Data collection, meters, sensors, etc. IT Systems, VM Management, Etc. Try it yourself… Thank You! Colin Richardson Colin.Richardson@on365.co.uk +44 1509 261100 “According to IDC analysis and buyer perception, Schneider Electric is an IDC MarketScape Leader worldwide.” IDC, Katherine Broderick, Jan 2012: “IDC MarketScape: Worldwide Datacenter Infrastructure Management (DCIM) 2011 Vendor Analysis”