2. Stranded Power
40 to 80%
Low Power
Utilization
- 20 to 60%
Peaks & buffers Redundancies &
Backups
Power infrastructure – Expensive & Low utilization
Static power budgets CapEx: $6 - $12 Mill/ MW
OpEx: $0.5 - $1 Mill/ MW/ Yr
Revenue: $1 – 3 Mill/ MW/ Yr
3. ⌁ Intelligent battery, power relays
⌁ Implement software policy
VPS provides Software Defined Power® (SDP)
S/W Power control plane H/W hooks to S/W+
⌁ Collect, aggregate power use data
⌁ Optimize power distribution with
predictive analytics
Like S/W defined compute - Pool, dynamically optimize and allocate resources
4. ⌁ Generate 20-50% additional revenue
⌁ Defer $10-15M/MW CapEx, Avoid $1M/MW/Yr OpEx
⌁ Resell SDP to customer Customer optimizes distribution Recurring revenue
How does SDP benefit your data center?
5. FACILITY
ENDOFROWATFACILITY
ORWHITESPACE
WHITE SPACE (IT GEAR)
VPS Solution: Non-Disruptive, Seamless Deploy
Intelligent Batteries
3RD PARTY
H/W
• Deploy in brownfield data centers – with no change or disruption to existing power infrastructure
• Deploy in greenfield data centers – with fewer power components and redundancies
7. With ICE
Batteries Peak Shave > Unlock Power reserved for peaks
Without ICE
Provisioning for Peak > Low Power Utilization
UPS Container with 4 Racks + HVAC
ICE Software collects Power Data across your Data Center
1.Sets Control Limits of Utility power
2. Battery provides balance power
3. Under a 5% State of Charge Discharge Limit configured
VPS ICE S/W performing peak shaving
Battery shaves peaks Utility serves base line load customer can stack more
IT Racks or workload in existing power footprint
8. • Software configurable redundancy at rack level
• Maintains High Availability for Mission critical racks (or servers)
2N
Dual Corded Rack
10 kW
10 kW 10 kW
50% power is locked in redundancy
Traditional 2N environment
Mission Critical
Rack
(99.984%)
10 kW
10 kW 10 kW
100% power utilization with dynamic redundancy
2N environment with ICE Dynamic
Redundancy
Less Critical Rack
(99.95%)
10 kW
• Say Green Line Fails. ICE
ensures power on Yellow line
to higher priority rack
• ICE will switch off power to
the yellow on 1N rack in 4
milli seconds as the absolute
last resort. It measures real-
time power and takes no
action
• if total draw total is less
than 10KW
• If battery is available to
ridethrough
• If power draw can be
reduced thru lowered
CPU frequency by ICE
policy based on actual
power deficit in real time
• But after performing safe shut
downs, alerts, Vmotion, back
ups and other relevant actions
ICE Switch ICE Switch
VPS ICE S/W performing Dynamic redundancy
9. Dynamic Policy-based Orchestration
Predictive Provisioning
Power Variability Management
Power Availability Control
Resiliency-aware RSD Provisioning
Hardware
and
Software
Control
Systems
Mechanisms
Orchestrating
HW and SW
Control
Systems
Customer
Value
Proposition
Platform
Value-added
Capabilities
DCMServers
POD
Manager
Data Center
Mgt
VPS SDP exposed as API for easy integration data center systems
10. Customer Benefits: Economics
USE CASE Existing Facility
Current 2MW Data Center has no Power Available, Needs to add 1 MW
WITHOUT ICE: Build 1MW of Additional Facility WITH ICE: ICE Increases Power within existing footprint
$10 M Additional Facility CapEx
$1.0 M Additional Maintenance OpEx per year
Saves Facility CapEx spend $6 M (net of $4M VPS Cost)
Saves $0.75M / Year in OpEx (net of $0.25 M in VPS Cost)
GRID
GENERATOR
X-
FORMER
DISTRIBUTION
COOLING
SYSTEM
UPS
IT
IT
IT
IT
ICE
ICE
SG
4 Year Savings = $9.0 M/MW Average Per Server saving
$1,750
(@200 W servers, 5000 servers/MW $9,000,000/5,000=
$1,750)
11. Customer Benefits: Economics
USE CASE Existing Facility
Current 2MW Data Center has no Power Available, Need to add 1 MW for new customer
WITHOUT ICE: Build 1MW of Additional Facility WITH ICE: ICE Increases Power within existing footprint
$10 M Additional Facility CapEx
$1.0 M Additional Maintenance OpEx per year
Saves Facility CapEx spend $6 M (net of $4M VPS Cost)
Saves $0.75M / Year in OpEx (net of $0.25 M in VPS Cost)
GRID
GENERATOR
X-
FORMER
DISTRIBUTION
COOLING
SYSTEM
UPS
IT
IT
IT
IT
ICE
ICE
SG
4 Year TCO deferred = $9.0 M/MW
+ Additional revenue in 4 years = $4 to 8 Million
12. Customer Benefits: Economics
USE CASE
New Facility
Build 1 MW
WITHOUT ICE: New Facility WITH ICE: ICE Increases Power within existing footprint
$10 M Facility CapEx
$1.0 M Maintenance OpEx per year
Saves Facility CapEx* spend ($3.4 M, net of VPS Cost)
Saves $0.2 M / Year in OpEx
GRID
GENERATOR
DISTRIBUTION
COOLING
SYSTEM
IT
ICE
ICESG
* Build 0.6 MW instead of 1MW at a cost of $11 M per MW= $6.6 M
For the 0.6 MW footprint – ICE $3.0 M; other Power & Cooling $3.6 M
13. Thank You
Virtual Power Systems, Inc
4699 Old Ironsides Drive, Suite 100
Santa Clara, CA
www.virtualpowersystems.com
14. The VPS ICE Solution Software Power Control Plane
Pool Power Sources
Optimize Power Distribution
Automatically Allocate Power Budgets
ICE Hardware Battery & Power Conversion
ICE Switches & Sensors
3rd Party Power Hardware IT
Management Systems
Learn Power Distribution
Remote Monitoring & Automated Control
ICE CLOUD &
REMOTE
SERVICES
ICE
APPLICATIONS
ICE OPERATING
SYSTEM
POWER
HARDWARE
SENSORS & CONTROLS
IT SYSTEMS
ICE Reference Designs Licensee Manufactured
Connect, Communicate, Control, Persist Data and
Provide API Interface for Applications and Devices
15. ICE Use cases
1. Added capacity at near Tier 3 availability
2. Tier 2 data centers offering higher availability at rack or row/room levels
• Greenfield next generation data centers
• Brownfield expansion to new whitespace
3. Policy implementation Executing optimized power distribution decisions
4. Added capacity in paired data centers at flexible levels of availability
5. Data center consolidationTapping into unused redundant power
16. Use Case 1: Added capacity at near Tier 3 availability
Before ICE: Max
utilization 45%
After ICE: Max
utilization 90%
17. Use Case 2: Tier 2 data
center offering higher
availability at
Room/Rack/Rack PDU
outlet
(Brownfield)
DC Source augmenting AC powerTraditional Tier 2 power distribution
Design features:
• Augmented DC power supply
• Decentralized UPS
• Decentralized smart Switch for
redundancy, source favoring, policy
executions
• Software managed hardware
• Scales with the racks / rows
(rightsized expansion)
NOTE: This design assumes specialized
rack design. Concept can be
implemented for standard racks, with
minor re-design.
18. DC Source augmenting AC powerTraditional Tier 2 power distribution
Tier 3 SLA whitespace
added in a Tier 2 data
center, with DC power
augmentation
Use Case 2: Tier 2 data
center offering higher
availability at
Room/Rack/Rack PDU
outlet
(Greenfield)
19. DC1
2N
50% load
DC2
2N
50% load
TODAY
Paired Data Centers
Normal Operations
DC1
2N
0% load
DC2
2N
50% load
Paired Data Centers
Failure Scenario
X
DC1
2N
50%
load
WithICE
Paired Data Centers
Normal Operations
Paired Data Centers
Failure Scenario with Dynamic Redundancy Only in double failure,
paired Data Centers will
lose 1N load
1N
50%
load
DC2
2N
50%
load
1N
50%
load
DC1
2N
50%
load
1N
50%
load
DC2
2N
50%
load
1N
50%
load
+ +
X
DC2
2N
50%
load
1N
50%
loadX
Utilization can be 100% (doubled)
with Dynamic Redundancy
1N Availability is higher in paired
Data Centers, due to pairing
Use case 4: Added capacity in paired data centers at flexible levels of availability
20. TODAYWithICE
Use case 5: Data center consolidation Tapping unused redundant power
DC
2N
50% load
Power locked in for
extended period of
time
2N
50% load Migration
or expansion
to a new
whitespace
Loc1 Loc2
DC
1N
50% load
2N
50% load Migration
or expansion
to a new
whitespace
Loc1 Loc2
Pre migration: Uses
redundant power in Loc2
DC
2N
50% load
1N
50% load Migration
or expansion
to a new
whitespace
Loc1 Loc2
Post migration: Uses
redundant power in Loc1
End of Row Peak Shaving and Dynamic Redundancy is depicted in the Single Line Diagram Form. The blue box shows peak shaving equipment with stored energy in the form of batteries or other power sources can be used to manage the load of ether the A or B power system. When redundant capacity is being used in normal operations and a system transfers from 2N to 1N the peak shaving equipment selects the remaining source and shares the load to maintain UPS loads at normal levels while the necessary amount of redundant load is shed using various techniques. The red star is the additional load with a slightly lower availability that in the case of 1N operation is reduced to manage the load on the UPS for long term stability.
The End of Row ICE Blocks are placed in a panel, which can inject power in a parallel configuration. Parallel configuration means that the entire battery capacity can be used for Peak Shaving. Two Master devices are used, one primary and the other backup in case primary fails. Rest are slave units, used for supplying power instead of control logic.
ICE Switch is placed in each rack, the new 1N racks which use redundant power will require ICE Switch for sure.
End of Row Peak Shaving and Dynamic Redundancy is depicted in the Single Line Diagram Form. The blue box shows peak shaving equipment with stored energy in the form of batteries or other power sources can be used to manage the load of ether the A or B power system. When redundant capacity is being used in normal operations and a system transfers from 2N to 1N the peak shaving equipment selects the remaining source and shares the load to maintain UPS loads at normal levels while the necessary amount of redundant load is shed using various techniques. The red star is the additional load with a slightly lower availability that in the case of 1N operation is reduced to manage the load on the UPS for long term stability.
The End of Row ICE Blocks are placed in a panel, which can inject power in a parallel configuration. Parallel configuration means that the entire battery capacity can be used for Peak Shaving. Two Master devices are used, one primary and the other backup in case primary fails. Rest are slave units, used for supplying power instead of control logic.
ICE Switch is placed in each rack, the new 1N racks which use redundant power will require ICE Switch for sure.
This is an example of a power system that offers several advantages in a data center with a single utility feed available that may only be classified as Tier 2. The utility source is paired with a Grid Tied Generator and parallel AC and DC power distribution system. Each rack contains a UPS system with an AC and DC feed. An expandable rectifier system provides a source of DC power and battery storage systems are located on the end of each row. In case of failure of the Utility Source, the Diesel Generators (DG) come on automatically and the transfer switch aligns power to the DG source. However, in the event of loss a major component such as the step-down transformer or the Switchgear both sources are ineffective because of loss of path for distribution of power. Using DC Rectifiers, available in Modular Capacity of 500 KW or less, AC can be converted to DC and use to power the rack mounted UPS systems, installed in each rack. The system favors AC power during normal operation but in the event of any failure in the AC system the DC system uses the batteries while the Grid-Tie generator starts the load may shift from the batteries to the generators. This system is modular and both the AC UPS system and the DC system expands as the number of racks in the data center increases over time. In addition, the availability of the power in the racks is greater than a what is available in a typical Tier 3 data center topology with a reduced cost over all and a delayed capex cost as the system expands.
Being a brown field deployment, the benefits of Software defined Power and new augmented DC rectifiers can only be availed by new Whitespace deployment.
This is comparison of a typical data center that is not using Software Defined Power capabilities today and uses 50% of the capacity on each side of the power system during normal operations. Then in the event of a transfer from 2N to 1N operation all of the power is transferred to the remaining side of the power system.
With Software Defined Power some data centers are able to use 100% of the capacity of both sides of the power system during normal operations. ICE hardware and software is used as described in the previous slides to manage the redundant power in the event of a transfer from 2N to 1N operation. The redundant power that is above and beyond the 1N capacity limit is managed by peak shaving initially and then a series of load shedding protocols are executed by the ICE software until the system is stable in 1N mode at the full capacity of the remaining power feed. The load shedding protocols are defined by predetermined policies and operational variables in place at the time of the event.
This is a use case where a data center that is operating at or near its design capacity without ICE will have challenges when replacing legacy systems with new systems when the new systems need to be installed an put on-line while the legacy systems continue to run. In this example both systems will be running at the same time until the new system are proven to be reliable and the old systems can be taken off line. When both system are running the data center may not have enough capacity to run both system reliably.
With ICE this type of load migration can be managed with less disruption and with higher availability for critical system in the data center than without ICE. Redundant capacity can be used for the additional capacity head room while the migration takes place. In the event of a hiccup in the new equipment it will be treated with lower priority during the migration until ready to transfer the workloads. As the workloads transfer, the priority of the equipment changes as well. The process becomes a seamless transition where power is managed in such a manner as to maximize availability based on the criticality scoring of each load in real time.
This is an overview of Software Defined Power’s capabilities to converge IT operations with Facility operations to manage capacity in such a manner as to increase the utilization of a facility beyond what is available with management tools previously available.