Colocation Data Center Availability : Making Maintenance Windows Obsolete Christopher Thames, Director of Critical Facilities
Colocation Data Center Availability: Making Maintenance Windows Obsolete How can you provide critical infrastructure power and cooling systems 24x7 without increasing costs to colocation customers? Presenters will discuss the outsourced environment, which includes live UPS load shifts for maintenance, equipment faults, or UPS commissioning. These improvements allow data centers to be concurrently maintainable and fault tolerant and eliminate the issue of stranded power in the data center.
What systems are colocated in a third party data center? What SLA do they provide for power and cooling uptime? Regardless of SLA, what is your expectation? 100% Uptime Managing Colocation Data Center Availability Expectations
More on Colocation Data Center Expectations Power Redundancy Power Density Cooling System Redundancy SLAs and Contracts Connectivity Customer Experience
Data Center Power Reliability Utility Build a datacenter assuming no utility Generator (Backup) Power Lessons from Hurricane Sandy: -Fuel resupply contracts -Diesel generator oil resupply -Automated? -Starters, Batteries Load Testing and maintenance (later…)
Uninterruptible Power Supply (UPS) System Designs – Capacity or “N” UPS system – N+1: Isolated or parallel redundant “Catcher UPS” or multiple UPS on single output bus – System plus system (2N) redundant High capital expense, lower system efficiency – Distributed redundant (N+∞) Reliability increases as system scales up Battery Backup – Flywheel vs. lead acid (dry and wet cell) – New technologies Data Center Power Reliability
UPS Innovation on the Market Today Double versus single conversion “Efficiency” Modes Multi-module versus single-module systems Internal bypasses Plug-and-play UPS maintenance …… What else? Improve the distribution system redundancy.
UPS Innovation on the “Topology” Level Pod-based power distribution: constraint on customer growth “Stranded” power: increased cost to provider (and customer) Low UPS power utilization: decreased efficiency 2N 50/50 distribution: increased risk to customer in case of failure Better UPS topology design, better cost, better availability
Not Just Improve Availability, But Get Rid of Maintenance Windows Think outside the box Leverage traditional, yet resilient designs Mix in innovative improvements Communicate value to customers: 100% availability, even during maintenance Example?
Building A Delivery System to Maximize Maintenance Flexibility – Eliminating Maintenance Windows
UPS Switch Point US Patent No. US B2 “A scalable power distribution system for a data center and methods for scaling a power distribution system…”
UPS Swith Point UDP AXFRMUSPUPS A UDP BXFRMUSPUPS B A/B Initial phase installs two UPSs to support the critical load
UPS Switch Point UDP AXFRMUSPUPS A UDP BXFRMUSPUPS B UPS C A/B As we scale, we add UPS C and need to redistribute load
UPS Switch Point UDP AXFRMUSPUPS A UDP BXFRMUSP UPS B Bypass UPS C Bypass A/B Place UPS B & C in Bypass so they output same source
UPS Switch Point UDP AXFRMUSPUPS A UDP CXFRMUSP UPS B Bypass UPS C Bypass A/C Disconnect UPS B
UPS Switch Point UDP AXFRMUSPUPS A UDP CXFRMUSPUPS B UPS C A/C Put UPS B & C back into online mode, load is now A/C
A/C UPS Switch Point UDP AXFRMUSPUPS A UDP CXFRMUSPUPS B UPS C The critical load was NEVER offline
UPS – External Maintenance Bypass with Alternate Power Source US Patent No. US B2 Redundant isolation and bypass of critical power equipment 100% uptime SLA, even during maintenance windows
UPS Z UPS – External Maintenance Bypass with Alternate Power Source UDP AXFRMUSP UDP BXFRMUSPUPS B A/B Batt Charging UPS AUIPUOP UPS Input Panel LOAD BANK UPS Output Panel
UPS – External Maintenance Bypass with Alternate Power Source UPS AUIPUOP UPS Z LOAD BANK Utility Initial condition: UPS Z offline, UPS A on Utility
UPS – External Maintenance Bypass with Alternate Power Source UPS AUIPUOP UPS Z LOAD BANK Utility Startup UPS Z
UPS – External Maintenance Bypass with Alternate Power Source UPS AUIPUOP UPS Z LOAD BANK Utility Disconnect Utility, connect UPS Z to UIP for A
UPS – External Maintenance Bypass with Alternate Power Source UPS A Bypass UIPUOP UPS Z LOAD BANK Utility Disconnect Utility, connect UPS Z to UIP for A
UPS – External Maintenance Bypass with Alternate Power Source UPS A Bypass UIPUOP UPS Z LOAD BANK Utility Connect UPS Z to UPS Output Panel
UPS – External Maintenance Bypass with Alternate Power Source UPS A Bypass UIPUOP UPS Z LOAD BANK Utility Disconnect UPS A from UOP
UPS – External Maintenance Bypass with Alternate Power Source UPS A Bypass UIPUOP UPS Z LOAD BANK Utility Disconnect UPS Z from UIP
UPS – External Maintenance Bypass with Alternate Power Source UPS AUIPUOP UPS Z LOAD BANK Utility Return utility to UPS A UIP
UPS – External Maintenance Bypass with Alternate Power Source UPS AUIPUOP UPS Z LOAD BANK Service, Load bank or test UPS A 100% independent from load Utility
UPS AUIP UOP UPS Z LOAD BANK Utility Critical load is UNAFFECTED by testing or maintenance
What Is Your Acceptable Level of Risk for Downtime? Reliability 100% Uptime Unreliable Very low BUDGET Healthy, IT-focused High risk RISK TOLERANCE Very risk averse Low-importance APPLICATION TYPE Prod’n/Network/DR Slow, incremental GROWTH/SCALE Fast, hyperscale “N” Capacity System Bargain Cloud Fault Tolerant & Concurrently Maintainable Fault Tolerant Concurrently Maintainable N+1 UPS Systems Single Points Of Failure Fault Tolerant & Concurrently Maintainable with a Utility Outage
Look for a Colocation Provider with No Maintenance Windows Your expectation = 100% Uptime – Make it your standard – don’t lower standards to meet colocation providers expectations (five 9’s) Choose a provider that can get you there – Design system with robust maintenance program in mind (external UPS bypass with alternate power source) – 2N+2 design? N+2 design? Tier 3? Tier 4? No maintenance windows – No exclusions to SLA!
Thank you Christopher Thames, Director of Critical Facilities Annie George, Product Marketing