15 Hardware and Wireless Environments

As described in the Using Powder section, Powder provides a wide variety of hardware and configured environments for carrying out wireless research, including conducted and over-the-air (OTA) RF resources as well as backend computation resources. These resources are spread across the University of Utah campus as shown in the Powder Map. Conducted RF resources are largely in the MEB datacenter, OTA resources all across campus, "near edge" compute nodes in the MEB and Fort datacenters, and additional "cloud" compute nodes in the Downtown datacenter (not shown on the map).

Following is a brief description of these resources. For further information about the capabilities of, and restrictions on, the individual OTA components, refer to the Radio Information page at the Powder portal.

15.1 Conducted attenuator matrix resources

A custom-built JFW Industries 50PA-960 attenuator matrix allows for configurable conducted RF paths between a variety of connected components. While the actual RF paths are fixed, and must be manually configured by Powder personnel, users can adjust the attenuation values on those paths dynamically.

The current configuration of paths is:

The connected components are:

• Two NI N300 SDRs each with two 10Gb Ethernet connections to the Powder wired network. The Powder nodes names are n300-1 and n300-2.

• One NI X310 SDR with two UBX160 daughter boards and two 10Gb Ethernet connections to the Powder wired network. The Powder node name is x310-1.

• Eleven NI B210 SDRs each with a USB3 connection to an Intel NUC. The NUC provides basic compute capability and a 1Gb Ethernet connection to the Powder wired network. The majority of the NUCs (nuc1 to nuc4, nuc8 to nuc10, nuc16) are:

  nuc5300		8 nodes (Maple Canyon, 2 cores)
  CPU		Core i5-53000U processor (2 cores, 2.3Ghz)
  RAM		16GB Memory (2 x 8GB DDR3 DIMMs)
  Disks		128GB SATA 6Gbps SSD Drive
  NIC		1GbE embedded NIC

  The remaining NUCs (nuc22, nuc23, nuc27) are newer, with up to four cores and 32GB of RAM.

The N300 and X310 SDRs have their 1 PPS and 10 MHz clock inputs connected to an NI Octoclock-G module to provide a synchronized timing base. Currently the Octoclock is not synchronized to GPS. The B210 SDRs are not connected to the Octoclock.

15.2 Paired Radio workbench resources

The Paired Radio workbench consists of two sets of two NI X310 SDRs each, cross connected via SMA cables with a fixed 30 dB attenuation on each path. Connections are shown in this diagram:

One pair, oai-wb-a1 and oai-wb-a2, has a single NI UBX160 daughter board along with a single 10Gb Ethernet port connected to the Powder wired network. The second pair, oai-wb-b1 and oai-wb-b2, has two NI UBX160 daughter boards and two 10Gb Ethernet ports connected to the Powder wired network.

All four X310 SDRs have their 1 PPS and 10 MHz clock inputs connected to an NI Octoclock-G module to provide a synchronized timing base. Currently the Octoclock is not synchronized to GPS.

15.3 Indoor OTA lab resources

The Indoor OTA environment consists of four NI B210 SDRs with antennas on one side of the lab, and four NI X310 SDRs with antennas on the opposite side of the lab:

The four B210 devices are each connected to a four-port Taoglas MA963 antenna in a 2x2 MIMO configuration, meaning that the TX/RX and RX2 ports on both Channel A and Channel B are connected to an antenna element. The B210 port to antenna port connectivity is as follows: Channel A TX/RX to port 1; Channel A RX2 to port 2; Channel B RX2 to port 3; Channel B TX/RX to port 4.

The four B210 devices are accessed via USB from their associated Intel NUC nodes, ota-nuc1 to ota-nuc4. All NUCs are of the same type:

Also connected to each of the NUCs is a RM520N COTS UE connected to a four-port Taoglas MA963 antenna. Allocating the associated NUC will provide access to the COTS UE.

On the other side of the room, two of the X310 SDRs, ota-x310-2 and ota-x310-3, also have 2x2 MIMO capability via Taoglas GSA.8841 I-bar antennas connected to each of their Channel 0, and Channel 1 TX/RX and RX2 ports. The other two X310s, ota-x310-1 and ota-x310-4 have two of the I-bar antennas each, connected to the TX/RX and RX2 ports of channel 0 only.

All radios (X310s and B210s) have their 1 PPS and 10 MHz clock inputs connected to Octoclock-G modules to provide a synchronized timing base. The X310 devices connect to one Octoclock, and the B210 devices connect to another, with the two Octoclocks interconnected. Currently the Octoclocks are not synchronized to GPS.

15.4 Rooftop Base-station resources

Powder currently has seven rooftop base stations around the main University of Utah campus, as highlighted on this map:

(For a better sense of scale and location relative to all resources, see the Powder Map.) The siteids (used for node naming) for each base station are (clockwise from top right): Hospital (hospital), Honors (honors), Behavioral (bes), Friendship (fm), Browning (browning), MEB (meb), and USTAR (ustar).

Each Powder base station site includes a climate-controlled enclosure containing radio devices connected to Commscope and Keysight antennas. Each device has one or more dedicated 10G links to aggregation switches at the Fort Douglas datacenter. These network connections can be flexibly paired with compute at the aggregation point or slightly further upstream with Emulab/CloudLab resources.

15.5 Dense-deployment Base-station resources

Powder currently has five street-side and one rooftop base station located in a dense configuration around one common shuttle route on the main University of Utah campus. The base stations are highlighted in red and the shuttle route shown in green on this map:

(For a better sense of scale and location relative to all resources, see the Powder Map.) The siteids) (used for node naming) for each base station are (clockwise from upper left): NC Wasatch (wasatch), NC Mario (mario), Moran (moran), Guest House (guesthouse), EBC (ebc), and USTAR (ustar).

Powder dense deployment base station locations contain one or more radio devices connected to a Commscope antenna and a small compute node connected via fiber to the Fort Douglas aggregation point.

15.6 Fixed-endpoint resources

Powder has ten "fixed endpoint" (FE) installations which are enclosures permanently affixed to the sides of buildings at roughly human height (5-6ft). The endpoints are highlighted in red on the following map:

(For a better sense of scale and location relative to all resources, see the Powder Map.) The siteids) (used for node naming) for each FE are (clockwise from top right): Moran (moran), EBC (ebc), Guest House (guesthouse), Sage Point (sagepoint), Madsen (madsen), Central Parking Garage (cpg), Law 73 (law73), Bookstore (bookstore), Humanities (humanities), and WEB (web).

Each Powder FE enclosure contains an ensemble of radio equipment with complementary small form factor compute nodes. Unlike base stations, FEs do not have fiber backhaul connectivity. Instead they use commercial cellular and Campus WiFi to provide seamless access to resources.

Both NUC compute nodes are:

15.7 Mobile-endpoint resources

Powder has twenty "mobile endpoint" (ME) installations which are enclosures mounted on a rear shelf inside of campus shuttle buses.

As with MEs, each Powder FE enclosure contains radio equipment and an associated small form factor compute nodes and they use commercial cellular and Campus WiFi to provide seamless access to resources. However unlike FEs, not all MEs are likely to be available at any time. Only shuttles that are running and on a route are available for allocation. For a map of the bus routes and their proximate Powder resources see the Powder Map.

The Supermicro Compute Node is:

15.8 Near-edge computing resources

As previously documented, all base stations are connected via fiber to the Fort Douglas datacenter. The fibers are connected to a collection of CWDM muxes (one per base station) which break out to 1Gb/10Gb connections to a set of three Dell S5248F-ON Ethernet switches. The three switches are interconnected via 2 x 100Gb links and all host 10Gb connections from the rooftop and dense base-stations. One switch also serves as the aggregation switch with 100Gb uplinks to the MEB and DDC datacenters that contain further Powder resources and uplinks to Emulab and CloudLab.

There are 19 servers dedicated to Powder use:

All nodes are connected to two networks:

• A 10 Gbps Ethernet “control network”—this network is used for remote access, experiment management, etc., and is connected to the public Internet. When you log in to nodes in your experiment using ssh, this is the network you are using. You should not use this network as part of the experiments you run in Powder.

• A 10/40 Gbps Ethernet “experimental network”. Each d740 node has two 10Gb interfaces, one connected to each of two Dell S5248F-ON datacenter switches. Each d840 node has two 40Gb interfaces, both connected to the switch which also hosts the corresponding mMIMO base station connections.

15.9 Cloud computing resources

In addition, Powder can allocate bare-metal computing resources on any one of several federated clusters, including CloudLab and Emulab. The closest (latency-wise) and most plentiful cloud nodes are the Emulab d430 nodes:

which have multiple 10Gb Ethernet interfaces and 80Gbs connectivity to the Powder switch fabric.