[RFC] Feature Proposal – Data center technical attributes

General talk Tagging general discussion
,
1

Hello everyone,

I’m opening the RFC phase for my proposal “Data center technical attributes”.

This proposal adds four optional tags for telecom=data_center features:

  • data_center:tier
  • data_center:total_power
  • data_center:IT_power
  • data_center:IT_area

These tags are intended to capture technical characteristics of data centers, such as redundancy tier, facility power capacity, IT load, and usable IT floor area.

The goal is to make data center mapping more useful for infrastructure analysis, energy research, and capacity planning.

I’d really appreciate comments, and whether the proposed units and usage rules are clear enough.

Please discuss the proposal on its Wiki Talk page.
Thank you.

2

Several concerns about this: First I’m hearing of the Uptime Institute. The tag name should have a reference to the standard in use. This is extremely difficult to identify without entering an access-controlled site.

I know that several of the hyperscalers consider these numbers to be trade secrets and will go after anyone caught leaking them. (Note: specifically, “how many computers they own” and “how much of the datacenter power is going to the computers” except as published for publicity purposes)

3

Also, the proposal seems to encourage tagging this by reference to the Uptime Institute website. But the Terms of Use of the website appear to directly prohibit such use. Does OSM have permission to use this data?

4

DWG have previously been approached to remove details of data centres from OSM due to security concerns. Companies involved have said that there is nothing on the building involved to show that they operate it, or that it’s a data centre, so it shouldn’t be advertised!

5

Can I ask was that to remove specific tags defining the characteristics of the data centre, or to obscure the fact that it’s a data centre in the first place?

6

Both to take their name off the building & also to remove any reference to a data centre.

I must admit that I don’t remember the eventual outcome of the issue though!

7

Perhaps ‘data_center:uptime_tier’ would be more appropriate. What do you think?

8

You’re right. I just check the Terms in use of the Uptime Institute and it seems against republishing their content.
I think I’ll to change the description in the proposal to only indicate the uptime tier from the operator website or documentation

9

I agree that some information may be treated as confidential by operators. My intention is not to encourage anyone to leak internal or non-public information. For example colocation data centers publish their informations, so for this type of center, I believe it will be possible to store it.

10

I’ve mapped quite a few data centres in Ireland: Open Infrastructure Map

In the EU, it might be difficult to hide things like power demand, as an Environmental Impact Assessment Report is likely required and likely to be made public for all but the smallest data centres. Of course, I would imagine power usage is commercially sensitive due to the environmental impact, not inter-operator rivalry. Available power capacity might not relate directly to processing capacity, as cooling and housekeeping also use power, that will vary from site to site.

Other statistics about data centers like usable IT floor area might be obscured due to the use of stacked modular units within the data halls.

It might be useful to determine what existing tag combinations people are using with telecom=data_center, industrial=data_centre, building=data_center, and telecom=*, communication*=** and man_made=* generally. Given the limited number of data centers worldwide and in specific location, it’s not easy to get information from TagInfo, so a suitable Overpass-Turbo query might be needed.

11

Simply changing the mapping instructions is still using their proprietary tier system, which will not be a worldwide standard.

1 Like
12

May I ask why you said it will not be a worldwide standard? Or is there any other organism which is in the process of being recognised?

Oh okay. I’ll remove it from the proposal then. Thanks!

13

I’ve updated the proposal by removing the ‘tier’ tag and renaming the others:

  • data_center:total_power → data_center:input
  • data_center:IT_power → data_center:load
  • data_center:IT_area → data_center:floor_area
14

I think the documentation should also emphasize that the figures used are the “critical” / “design maximum” power loads, not the typical power loads. (This is the correct values to map, because the design maximum is how you size your power wires, while the typical load varies when computers turn on and off.)

1 Like
15

Hi everyone,

I’ve updated the proposal based on the feedback so far. The main change is a renaming of the power-related tags: data_center:input and data_center:load are now data_center:power:input and data_center:power:load.

This is mainly to make the intended meaning clearer for people who may not be familiar with OSM tagging conventions, and to reduce the risk of misunderstanding what these tags are meant to describe. I also expanded some of the definitions and explanations.

Thanks again for all the comments, and please let me know if you have any other concerns.