whitehouse/datacenters

Data Center Optimization Initiative (DCOI) Memo

datacenters's Issues

Clarification on Cost Savings and Avoidance

RE: P.7, "Agencies shall, by the end of fiscal year 2018, reduce government-wide annual costs attributable to physical data centers by at least 25%, relative to the fiscal year 2016 IT Infrastructure Spending data submitted to the Federal IT Dashboard."
Questions:

1. How do the cost savings targets account for increased workload (i.e. growth)?
2. What is the basis of "physical data center spending"? (i.e. how is it defined, how is it derived, and where does it come from?)
3. Is this still tied to an exhibit in the IT budget, or is it now based on the IDC cost savings and avoidance figures, or neither?
4. Will credit for cost savings and avoidance accrued prior to FY2016 be included in the 25%?

Missing Footnote

RE: P.6, "The following optimization metrics are listed in order of priority. Agencies shall achieve and maintain all listed target values by the end of fiscal year 201822"
Question:

1. Footnote 22 does not exist. Can you please provide the footnote for this item?

Define "Essential Enterprise Levels"

The top of page 3 states:

As previously required by the FDCCI, agencies shall continue to principally reduce application, system, and database inventories to essential enterprise levels by...

It is suggested that the term "essential enterprise levels" be defined.

Improving efficiency and data center density w/o air gaps

The DCOI optimization goal (Government-wide Optimization Targets for Tiered Data Centers) includes improving efficiency (server virtualization density), lowering power costs, and improving utilization (improved efficiency). The DCOI optimization of physical data centers would benefit from pursuing opportunity to remove air-gapped virtualization deployments that are often used to achieve multi-tenant separation. With security controls in place to segment and control admin access, greater virtualization densities can be achieved, dramatically reducing costs and reducing the manpower needed to oversee systems.

Use unit cost metrics to measure efficiency

Top of page 2 states:

"The DCOI, as described in this memorandum, requires agencies to develop and report on data center strategies to consolidate inefficient infrastructure, optimize existing facilities, achieve cost savings, and transition to more efficient infrastructure, such as cloud services and inter-agency shared services."

While cloud services can provide greater efficiency, this should not be assumed to always be the case. It is suggested that OMB consider including unit cost related metrics that can be used by an agency to justify their decisions.

Scope for Automated Infrastructure Management

The bottom of page 5 states:

"Agencies shall replace manual collections and reporting of systems, software, and hardware inventory housed within data centers with automated monitoring, inventory, and management tools (e.g., data center infrastructure management) by the end of fiscal year 2018."

It is suggested that OMB clarify whether this applies to all data centers (tiered and untiered).

Title link is 404

Title link points to https://datacenters.cio.gov/datacenters which is a 404.

Performance Metric - Facility Utilization

RE: Table 1 (Facility Utilization), P.7, "[Total Active Rack Count x 30 sq. ft. Total Gross Floor Area]"
And Footnote 26, P.7, “Total gross floor area” is defined as total square footage available for IT equipment."
Questions:

1. The Cost Model Guide defined "total gross floor area" as: "Total sq.ft. usable for data center operations. This is often referred to as white space or raised floor area and does not include auxiliary or mechanical system spaces used. The model automatically adds additional space based on the type and tier of facility being reported. Validation is limited to any value greater than 0.". This definition was interpretted by some agencies as meaning "do not include auxiliary or mechanical system spaces outside of the computer room, such as generator rooms, electrical rooms, etc." Can the intent be clarified (i.e. are agencies supposed to exclude floor space in the computer room that is taken up by HVAC and power equipment, code-required aisleways, etc.?
2. Will agencies be allowed to correct the inventory square footages based on the clarification?
3. Since the Facility Utilization is significantly impacted by the number of server racks versus stand-alone, non-racked IT system equipment (such as tape libraries, storage units, etc), how does the metric compensate for the difference in a server closet having only racks (for example) and an enterprise data center that is populated with a significantly greater amount of non-racked equipment? (Using the same ratio seems counter-intuitive if the objective is to reach optimal utilization of the floor space.)

Clarify definition of a "Closed" data center

Goal 3: Closed Data centers (near the top of page 8) states:

"In all cases, the term “closed” for data centers shall refer exclusively to tiered or non-tiered data centers that: a) no longer consume power; or b) no longer house physical servers (whether in a production, test, stage, development, or any other environment)."

It is suggested that the definition be clarified by adding "for the hosting of IT services" at the end of "no longer consume power". In some instances, data centers are in multipurpose buildings, where the space originally housing the data center may still be in an agency's inventory (consuming some power for non-IT purposes) even if the data center is no longer there.

Self-driving data centers?

The definition of a data center as "rooms with at least one server, providing services" is ridiculously broad. Can you please put in an exclusion for "servers that are installed on mobile platforms." I would hate to get in a situation where OMB expects agencies to catalog the all their self driving cars, vehicle based navigation, communication and sensory systems, and equipment packages that include servers used to service disaster areas,

Gross Floor Area Reduction may be overstated

Towards the bottom 3rd of page 8, there is a sentence that states:

"This target will result in the closure of 52% of the overall Federal data center inventory and a reduction of approximately 31% in the gross floor area occupied by data centers, government-wide."

While the gross floor area (for a closed data center) may no longer be used as a data center, that does not guarantee it is able to be re-provisioned for some other use. In some instances data centers are in Federal buildings in space (i.e., basements) that is not easily used for other purposes.

It is suggested that a unit cost metric be developed that can be used to justify the consolidation (or not) of some of these spaces.

It is also suggested that OMB ensure the Gross Floor Area reduction goal be vetted based on agencies successes over the past 5 years in being able to eliminate gross floor area as they close data centers.

Is closing all non-tiered data centers achievable?

Towards the bottom of page 8 there is a paragraph that states:

"In the long term, all agencies should continually strive to close all non-tiered data centers. Server rooms and closets pose security risks and management challenges and are an inefficient use of resources. As such, although at least 60% of non-tiered data centers are required to be closed before the end of fiscal year 2018, OMB expects that agencies will consider all such facilities as temporary and work to close them."

Closure of all non-tiered "data centers" seems untenable considering remote locations require at least one closet for the purpose of WAN termination and local compute services (i.e., print).

It is suggested that the data center definition be modified so that these locations are NOT considered data centers thereby making the goal more realistic.

Policy effective day missing

Policy

"As of March DD, 2016, FDCCI is superseded by the Data Center Optimization Initiative (DCOI) established in this memorandum."

(https://datacenters.cio.gov/policy/)

Clarify Data Center Applicability

The FITARA section of the 2015 NDAA requires the head of each covered agency, assisted by the CIO of the agency to submit "a comprehensive inventory of the data centers owned, operated, or maintained by or on behalf of the agency;"
On Page 2 of the DCOI guidance its states; All data center infrastructure and services, including contracts for third-party data centers and services agency-wide, shall be managed by the agency CIO in a manner consistent with FITARA7 and OMB Memorandum M-15-14, “Management and Oversight of Information Technology.
On Page 6, Metric Target Values, it states; "Optimization, cost-savings, and closure metrics and goals apply to *_all federally-owned *_data centers, except high-performance computing (HPC) nodes.
Additionally, EO13693 Implementing Instructions states; Definitions – "Data centers" at "agency facilities" means all Federal data centers – data centers owned, operated, or maintained by or on behalf of the agency, irrespective of the size, as defined by OMB for data center reporting in the Integrated Data Collection (IDC). This definition includes contractor/commercial facilities containing government-funded IT equipment.
As currently written, there are multiple interpretations on data center applicability for DCOI. "Federally-owned" would exclude data centers operated by both government contractors and commercial so-location data centers.
Recommend the DCOI guidance memo include a section which clearly identifies which data centers are applicable to the DCOI guidance. Suggest using the definition in the EO13693 Implementing Instructions.

Metrics for non-tiered data centers

Page 6, footnote 21 "For non-tiered data centers, only automated monitoring of server utilization is required."

That seems confusing, and a major change vs. the "PUE is required for all data centers" language previous in the document.

Also it's unclear what "server utilization" represents, it could be:

• CPU activity (i.e. load)
• Energy Consumption
• Uptime (i.e. on 99.9999% of the time)
• ...

I would suggest striking the exception or expanding with more specificity.

HPC Nodes

RE: P.6, "Optimization, cost-savings, and closure metrics and goals apply to all federally-owned data centers, except high-performance computing (HPC) nodes."
And P.7 (Metric Target Values - Table 1)
Questions:

1. Please verify that a data center containing HPC nodes (regardless of number) is considered an HPC Data Center.
2. What metrics (if any) apply to Data Centers that contain both HPC and Non-HPC systems?
3. Please clarify exactly how HPC nodes should be addressed if data centers containing HPC nodes are NOT excluded from the inventory reporting requirements, but ARE excluded from the metrics and goals?

Data Center Energy Practitioner (DCEP)

The EO13693 states: "All existing and new data centers shall have at least one certified Data Center Energy Practitioner (DCEP) assigned to manage its performance." The DCOI guidance does not make any reference to this EO requirement. Since PUE is a key goal in DCOI, suggest including DCEPs as a recommended "best practice".

Closure and Optimization targets should acknowledge and build on past performance

The bottom of page 4 states the following:

"Under this memorandum, OMB sets closure and optimization targets that are applicable to each type of data center. Additionally, the terms “core” and “non-core” will no longer be used as the categorical benchmarks for OMB oversight."

It is suggested that text be added clarifying that these goals will be beyond accomplishments currently achieved by agencies in consolidating data centers.

Should cloud services be excluded from DCOI?

Cloud services are essentially a suite of services running over a physical data center infrastructure. When combined with the Cloud First policy, it would seem that DCOI (with a cloud exemption) would seek to quickly optimize data centers while simultaneously creating a growing portfolio of non-optimized data centers, existing behind the cloud label.

Cloud services should be included in this memorandum, should require PUE reporting, and be viewed as physical data centers - since - that is what they are underneath the set of services that reclassify the asset as a cloud service vs. a physical data center.

Clarification on Data Center Definition

RE: P.4, "rooms with at least one server, providing services (whether in a production, test, stage, development, or any other environment), are considered data centers."
Questions:

1. The previous definition of a data center was, "...a data center is now defined as a closet, room, floor or building for the storage, management, and dissemination of data and information. Such a repository houses computer systems and associated components, such as database, application, and storage systems and data stores. A data center generally includes redundant or backup power supplies, redundant data communications connections, environmental controls (air conditioning, fire suppression, etc.) and special security devices housed in leased (including by cloud providers), owned, collocated, or stand-alone facilities. Under this revised definition, neither square footage nor Uptime Institute tier classifications are required to define a facility as a data center." Given the definition of "Total gross square footage" and the new definition of a "data center", this potentially gives way to multiple "data centers" of less than 5 square feet. Is it OMB's intention that the inventory include such "data centers" that are dedicated to another purpose and only contain a server as a bi-product of another function and does not include associated components such as listed in the previous definition?
2. Agencies understand single servers need to be moved to a data center. Is there a more effective way to capture rooms with a single server or servers with no racks in order to monitor the consolidation of those servers in an appropriate data center without reporting each instance as an individual data center? (i.e. count the total number of physical servers not in a data center)

Add space between footnote numbers and text

This is a small aesthetic thing, but the numbers butt right up against the footnote text and it looks a little weird:

Tailor closure goal to each agency

In the middle of page 8, there is a sentence stating the following:

"Furthermore, agencies must close at least 60% of non-tiered data centers government-wide."

It is suggested that this goal be tailored to each agency depending on their size, mission profile and current progress against data center consolidation goals.

A high percentage of some agencies' non-tiered data centers are actually network closets in what are essentially field offices. These locations house network termination points as well as local computing (i.e., file and print) and are unlikely to close or be consolidated. At a minimum it is suggested they be excluded from the data center list and not included in the closure goal.

Review date may need updating

From https://datacenters.cio.gov/optimization/ -

This comprehensive review shall be completed by February 28, 2016, to align with the Integrated Data Collection (IDC) process.

That date is in the past, and will only be more in the past by the time public comment is over.

great initiative to leverage new technologies to drive data center consolidation

I applaud the Federal Government CIO's effort to refine the original data center consolidation guidance in order to leverage new technology that furthers the core goals to close and consolidate federal data centers. The commercial data center environment has made significant technological advances and proven new techniques that can be used by government CIOs to improve efficiencies, reduce overall data center costs, and improve security.

Failure to measure collateral Impacts of Closures of Data Centers

As part of data center consolidation efforts network servers at our workplace, a military base, were moved to another nearby base (between which there is not a fast connection established). There occurred an immediate, substantial, and distracting degradation in speed of access to network resources which appears to have affected everything from shared databases, to e-mail, to increased frequency of slow general Internet access. We are told the speeds meet the minimum requirements. Absent a national effort to upgrade U.S Internet speeds, can the projected benefits of data center consolidation be realized without such collateral impacts? Absent an effort to measure service quality at the client machine level, dollar savings claimed at the data center level seem suspect and myopic.

Clarification on Shared Service Providers

RE: P.3, "Data centers that OGP determines satisfactory in all of their operating standards will be designated as inter-agency shared services providers."
Questions:

1. What is the process for designating shared co-location data centers?
2. What are the minimum requirements for obtaining shared co-location data center designation? (i.e. what are the tests for determining the operating standards are satisfactory?)
3. If not yet determined, when will the answers to 1 & 2, above, be provided to agencies?

Request Data - not derived calculations

The table on page 7 reports a number of metrics that are derived, i.e. calculated from more fundamental data points.

Agencies should report the "foundational" numbers and allow others (e.g. GSA and OMB) to derive further calculations and insight from those metrics.

For example, "Energy Metering" is "GFA of Energy Metered Data Centers" / "GFA of All Data Centers"; Two agencies could report similar numbers, e.g. 1 but from vastly different places, e.g. Agency 1 has 1,000/1,000 and Agency 2 has 1,000,000 / 1,000,000 ---- simply requesting the derived calculation would obscure the opportunity to ask contextual questions on the distinction behind those very different numbers.

Also the table could use an appendix to help define specific details of calculations, for example:
* "Total OS" - this (unnecessarily) excludes virtual desktops but also lacks guidance on virtual containers (which are some-what like fractional OS's and not full VMs)

• "Server Utilization" - many different administrative commands can be used to report this information (potentially averaged over drastically different periods, or only representing a snapshot in immediate time). -- Provide standard commands and/or scripts for collecting and reporting.
• "Facility Utilization" - requests "total active rack count" - but agencies might have unpopulated (or unpowered) racks that are not "active" but still continue to draw power (such as for a UPS or cooling).
  - In addition "active" would need defining, e.g. is COOP that's not used but powered on "active"?

datacenters listserve isn't set up yet

From https://datacenters.cio.gov/community/ -

Executive agencies are encouraged to join the Data Centers listserv by emailing [email protected] with no subject and "subscribe datacenters" in the body (from a .gov or .mil email address only.)

But I gave this a shot with my .gov and got:

No LISTSERV list by the name of "DATACENTERS" is known to exist.

You may want to remove this until that works, or just go and create the listserve before the policy text is widely promoted.

When to provide justifications for data center expansion

At the bottom of page 2, there as a sentence that states:

Likewise, any expenditure of appropriated funds toward initiating a new data center or significantly expanding an existing data center must be accompanied by the same written justification.

Does this imply that the same justification needs to be provided when funds are budgeted and again when they are spent?

Clarification on Energy Metering and PUE

RE: P.5, "Agencies shall install automated energy metering tools and shall use these to collect and report to OMB energy usage data in their data centers. The March 19, 2015, Executive Order 13693, “Planning for Federal Sustainability in the Next Decade,” requires agencies to install and monitor advanced energy meters in all data centers by September 30, 2018."
And Table 1, P.7, "[Total Data Center Energy Used / Total IT Equipment Energy Used]"
Questions:

1. Please define the term "energy usage" relative to the data that must be submitted in the inventory fields that inform the PUE metric (i.e. Total Data Center Energy Used and Total IT Equipment Energy Used).
2. Since the Green Grid's guidelines provide various methods for measurement, the resulting field data can vary significantly.
   a) Will OGP be defining the salient characteristics of "advanced energy meters" for consistancy across procurements?
   b) Will existing electrical metering that doesn't meet those salient characteristics have to be replaced?
   c) Are "advanced energy meters" intended to meter more than electricity?
   d) How is advanced energy metering intended to address multi-occupancy facilities and campus environments that share infrastrucures?
   e) Will OGP be publishing standard guidelines for placement of power meters and a consistent methodogy for the calculation of Data Center PUE values?
3. Please address the potential unintended consequence of compromising IT data availability and security as a result of having to eliminate data centers because their geographic location cannot meet the PUE target because of inherent climate factors.

Cost Savings and Avoidance Goal for FY 2018 seems aggressive

Goal 2, at the bottom of page 7, states the following:

"Agencies shall, by the end of fiscal year 2018, reduce government-wide annual costs attributable to physical data centers by at least 25%, relative to the fiscal year 2016 IT Infrastructure Spending data submitted to the Federal IT Dashboard."

Considering the significant progress some agencies have already made since 2010, a blanket 25% reduction on top of current reductions seems very aggressive. Maybe this goal should be customized to each agency based on their portfolio and current progress?

The recently re-released GAO report (GAO-16-323) http://www.gao.gov/products/GAO-16-323 states:

"The 24 agencies participating in the Federal Data Center Consolidation Initiative have collectively made progress on their data center closures efforts. As of November 2015, agencies identified a total of 10,584 data centers, of which they reported closing 3,125 through fiscal year 2015. Notably, the Departments of Agriculture, Defense, the Interior, and the Treasury accounted for 84 percent of these total closures. Agencies are also planning to close an additional 2,078 data centers—for a total of 5,203—by the end of fiscal year 2019. In total, 19 of the 24 agencies reported achieving an estimated $2.8 billion in cost savings and avoidances from fiscal years 2011 to 2015. In particular, the Departments of Commerce, Defense, Homeland Security, and the Treasury accounted for about $2.4 billion (or about 86 percent) of the total. Further, 21 agencies collectively reported planning an additional $5.4 billion in cost savings and avoidances, for a total of approximately $8.2 billion, through fiscal year 2019."

For those agencies that had the lion's share of the reported savings, it seems as if an additional 25% reduction may be too aggressive.

Virtualization is a punishing legacy metric

Now that we're entering the Windows and Linux Software Container era, your "Virtualization" metrics and targets are outdated, and they way it is presented would appear to punish agencies for implementing the latest technology.

Consider two scenarios with 4 hosted applications:

1. Legacy OS Virtualization architecture = 1 Host OS + 4 heavy weight OSs to provide users 4 applications.
2. Contemporary Software Container architecture = 1 Host OS + 0 VM OSs + 4 lightweight software Containers to provide users the same 4 applications.

This metric punishes agencies that adopt contemporary and more efficient architectures, while promoting the use of more legacy architectures.

Recommend changing "Total OS/Total Physical Servers" to "Total Software Containers/Total Physical Servers." If folks are still using OS virtualization, you'd have to give them an instruction that they can count 1 VM as 1 software container, albeit a very heavy one.

Alternately I would recommend removing this metric all together as the problem OS virtualization is trying to solve is how do we make better use of our hardware capacity, i.e. the next metric Server Utilization.

Availability metric

A data center is the foundation service for IT. Data centers must continuously deliver their services. An availability metric would provide insight as to the effectiveness of the spending for the data center. This is a metric used throughout industry and well understood.

Annual Strategic Plan

Overall I think this is a pretty good policy. I have a few comments though.

First, you have a section entitle "Annual Strategic Plan" (page 9) but the strategic plan is intended to cover a period of 3 years. It shouldn't be "annual" if it is covering a three-year term. But this issue is related to my next point.

Second, and more importantly, it would be a shame if the data center strategic plan was another one-off plan that didn't integrate with the other IT plans in the organization. The strategy for data centers must work in concert with the strategy for cloud, for security and for mission application development. For this reason I would strongly recommend that this section be reorganized to require a discussion of the data center strategy as a required component of the Information Resources Management (IRM) Strategic Plan which is required every four years under the Paperwork Reduction Act. Doing this will marry the timing (4 years) up with GPRA-M and force consistency among the strategic plan of the Department and the IT plan.

Third, I would strongly recommend a section on transparency. You have a section on compliance reporting and you are proposing to collect data on a quarterly basis. I think that agency performance should be displayed on the IT Dashboard on a quarterly basis at least in the aggregate if not much more detailed. We should be holding agencies accountable to achieving the closure targets as well as the performance metrics identified in Table 1.

Finally, what are the repercussions for agencies that don't play ball? For example, the Department of Energy is woefully under-reporting its number of data centers. Heretofore they have been getting a free pass. How does OMB propose to hold those agencies accountable? It doesn't seem fair that some agencies go through this effort to do the right thing and other agencies ignore the policy with no negative impact.

Adjust date for data center inventory review

On page 4, the following sentence indicates a date in the past. This needs to be adjusted to allow for revisions to the inventory based on the new data center definition.

"Agencies shall perform a comprehensive review of their data center inventories and continue to maintain complete and updated data center inventories. This comprehensive review shall be completed by February 28, 2016, to align with the Integrated Data Collection (IDC) process."

Development Freeze for New and Current Data Centers

The DCOI guidance states:
"Beginning 180 days after issuance of this memorandum, agencies may not budget any funds or resources toward initiating a new data center or significantly expanding an existing data center without approval from OMB OFCIO."
Since FITARA makes the Agency CIO responsible and accountable for the Agency's IT investments, recommend that the Agency CIO be the approval authority for any new or significantly expanding data center.

Link the links in footnotes

I don't have the energy to go through and PR each one, but footnotes with URLs would benefit from linking the URLs. For example:

Applicability to HPC is unclear

Page 6 states "Optimization, cost-savings, and closure metrics and goals apply to all federally-owned data centers, except high-performance computing (HPC) nodes."

So HPC seems to, sensibly, be out of scope for DCOI - however introducing this exception so far into the guidance may cause confusion around interpretation; particularly in cases where HPC nodes are co-habitants of an agency data center --- i.e. then they would affect the overall metrics (e.g. PUE) but somehow be excepted.

Suggestions:

1. Introduce the exception earlier
2. Outline guidance in the case where HPC systems are not in a wholly dedicated facility.
3. Also explicitly state if federally-funded R&D centers are considered "federally-owned" (even if operated on behalf of the government)

Performance Metric - Server Utilization & Automated Monitoring

RE: Table 1 (Server Utilization & Automated Monitoring), P.7, "(%) Percent of time busy (measured as 1 – percent of time spent idle), measured directly by continuous, automated monitoring software, discounted by the fraction of data centers equipped with automated monitoring. [Average Server Utilization / Percent of Physical Servers Equipped with Automated Monitoring]"
Question:

1. Does this formula include or exclude VM hosts?
2. The metric for server utilization effectively prohibits reaching the 65% target for enterprise data centers that require High Availability (servers can't be more than 50% loaded to retain High Availability) for mission critical systems. At what point is the importance of High Availability for mission critical systems less important than meeting this metric?
3. Even in the absence of any High Availability requirement, if 100% of all data centers in the inventory (regardless of size or complexity) do not have these tools, it is highly unlikely this metric can be met without significantly increasing spending. Which is a higher priority, meeting this metric target or meeting the cost savings and avoidance metric target?
4. What is the measurement timeframe for the target server utilization of 65%? (e.g. measured across an 8-hour operational period, 24-hour operational period, or other?)

Adjust calculations for optimization targets to reflect scope of included data centers

Table 1 on the top of page 7 includes 4 metrics applicable to Tiered data centers only, and only 1 metric applicable to non-Tiered. It is suggested that the formulas be clarified to indicate the population of data centers included in the calculation. For example, the Energy Metering metric is currently defined as:

[Gross Floor Area of Energy Metered Data Centers] / [Gross Floor Area of All Data Centers]

Since the text under the table states:

"Only the Server Utilization & Automated Monitoring optimization metric shall apply to non-tiered data centers."

it is assumed that the Energy Metering metric only applies to tiered data centers and hence the formula should be defined as:

[Gross Floor Area of Energy Metered **Tiered** Data Centers] / [Gross Floor Area of All **Tiered** Data Centers]

Similarly, this clarification (of tiered vs. non-tiered) should be applied to the formulas for the following metrics as well:

• Power Usage Effectiveness (PUE)
• Virtualization
• Facility Utilization

Is 60 days enough time to develop DCOI Strategic Plans?

The top of page 10 states:

"Agencies are required to publish their Strategic Plans in a machine-readable JSON format at agency.gov/digitalstrategy under a section entitled, “Data Center Optimization Initiative Strategic Plans,” within 60 days of the issuance of this memorandum. OMB will provide instructions to agencies, including a schema, at management.cio.gov within 30 days of the issuance of this memorandum."

This implies agencies will only have 30 days from when OMB releases the instructions, to write their plans. Considering the effort involved, and what can be lengthy time frames for agencies to "Clear" their submissions, this seems overly optimistic.

Will cost savings/ avoidance goals be set with agency input?

Under Table 2 (at the top of page 8) there is a sentence that states:

"Within 30 days after publication of this document, OMB OFCIO will set individual cost savings and cost avoidance goals for each agency."

It is suggested that the sentence be modified as follows to indicate agencies will have had a chance to help with setting their individual goals:

"Within 30 days after publication of this document, OMB OFCIO will *_work with each agency to *_set individual cost savings and cost avoidance goals."

Automated Infrastructure Management Procurements

RE: P.6, "Coordinating with the GSA Federal Acquisition Service (FAS) to create and maintain an inventory of acquisition tools and products specific to the technology and services"
And Table 1 (Server Utilization Metric), "[% Percent of time busy / % of Physical Servers Equipped with Automated Monitoring]"
And Table 1 (Server Virtualization Metric), "[Total OS / Total Physical Servers]"
Questions:

1. Please clarify the requirement for "Automated Infrastructure Management tools" relative to the data that must be submitted in the inventory fields that inform the Server Utilization and Server Virtualization metrics.
2. Are these tools intended to be able to differentiate between physical and virtualized OS? (Capability to do this may be dependent on individual agency naming conventions or other factors that inherently prevent differentiation.)
3. Since "Automated Infrastructure Management tools" can vary significantly:
   a) Will OGP be defining the salient characteristics of "Automated Infrastructure Management tools" for consistancy across procurements?
   b) Will existing "Automated Infrastructure Management tools" that doesn't meet those salient characteristics have to be replaced?
   c) Will OGP be publishing standard guidelines for "Automated Infrastructure Management tools" and a consistent methodogy for use in required reporting?
4. When will OGP issue the acquisition vehicle(s) to support agency needs for "Automated Infrastructure Management tools"?

Clarification on Energy Metering and PUE

RE: P.5, "Agencies shall install automated energy metering tools and shall use these to collect and report to OMB energy usage data in their data centers. The March 19, 2015, Executive Order 13693, “Planning for Federal Sustainability in the Next Decade,” requires agencies to install and monitor advanced energy meters in all data centers by September 30, 2018."
And Table 1, P.7, "[Total Data Center Energy Used / Total IT Equipment Energy Used]"
Questions:

1. Please define the term "energy usage" relative to the data that must be submitted in the inventory fields that inform the PUE metric (i.e. Total Data Center Energy Used and Total IT Equipment Energy Used).
2. Since the Green Grid's guidelines provide various methods for measurement, the resulting field data can vary significantly.
   a) Will OGP be defining the salient characteristics of "advanced energy meters" for consistancy across procurements?
   b) Will existing electrical metering that doesn't meet those salient characteristics have to be replaced?
   c) Are "advanced energy meters" intended to meter more than electricity?
   d) How is advanced energy metering intended to address multi-occupancy facilities and campus environments that share infrastrucures?
   e) Will OGP be publishing standard guidelines for placement of power meters and a consistent methodogy for the calculation of Data Center PUE values?
3. Please address the potential unintended consequence of compromising IT data availability and security as a result of having to eliminate data centers because their geographic location cannot meet the PUE target because of inherent climate factors.

Conflicting requirements

Clarify data center definition

On page 4, the section entitled "Classification of Physical Data Centers" contains the following text:

"For the purposes of this memorandum, rooms with at least one server, providing services (whether in a production, test, stage, development, or any other environment), are considered data centers. However, rooms containing only routing equipment, switches, security devices (such as firewalls), or other telecommunications components shall not be considered data centers."

Are rooms/ closets containing local file and print services considered data centers? By the definition above, it seems as if they would be, yet by themselves they do not offer any consolidation opportunity. Agencies with geographically distributed field offices (i.e., HUD, Justice, Transportation, EEOC, FAA, Treasury, DHS, etc.) will most likely have a room in that office where wide area network terminations exist as well as local file and print services. It is suggested that these be specifically excluded to simplify and focus each agency's list on viable opportunities for optimization.

Data Center listserv address needs to be updated

The Community Support section at the bottom of page 10 states:

"Executive agencies are encouraged to join the Data Centers listserv by emailing *[email protected] *_with no subject and “subscribe datacenters” in the body (from a .gov or .mil email address only.)"

The above email address appears to need to be updated.

Define "Significantly Expanding"

In the section entitled " Transition to Cloud and Data Center Shared Services" it is suggested that the term "significantly expanding" be defined. Is this physical expansion as measured by square footage, an increase in IT spending associated with a data center, an increase in physical # of servers, or some other measure? What if the data center in question is a consolidation point and will grow as other locations converge into this one?

Reporting language on strategic plans is confusing about format

On https://datacenters.cio.gov/reporting/ -

Agencies are required to publish their Strategic Plans in a machine-readable JSON format at agency.gov/digitalstrategy under a section entitled, "Data Center Optimization Initiative Strategic Plans," within 60 days of the issuance of this memorandum.

The sentence calls for JSON, but says that their strategic plans should be published under a "section", which is too ambiguous to be actionable if referring to real JSON.

But maybe it's meant to reference just the human-readable version? The sentence references a URL without .json at the end.

For example, the White House has a redirect to a PDF:
https://www.whitehouse.gov/digitalstrategy

And no JSON to be found at all (404):
https://www.whitehouse.gov/digitalstrategy.json

But GSA has both:

• http://www.gsa.gov/digitalstrategy
• http://www.gsa.gov/digitalstrategy.json

It seems like this sentence should address the machine-readable (JSON) and human-readable requirements separately and clearly.

Will data center closure goals be set with agency input?

The bottom of page 8 states:

"Within 30 days after publication of this document,** OMB OFCIO will share with each agency its individual goal for data centers closures**, specifying the respective number of tiered and non-tiered data centers the agency must close."

It is suggested that paragraph be changed to indicate OMB will work with each agency to set their respective goal for closures.