User's Manual

Understanding Silver Spring Networks Access Points 1 About Access Points
Understanding Silver Spring Networks Access Points Rev 3 22 January 2013 Silver Spring Networks 16
Responding to a WAN Failure
An AP might not be reachable because the device cannot dial out on the WAN. The device
continues to try dialing out until it is either able to connect, or it has dialed 308 times. If it can
connect, the device notifies the event management program that it has successfully
recovered. If it cannot, then the modem resets itself. The behavior of an AP repeatedly
becoming unreachable, and then recovering shortly after, is called flapping.
A small amount of flapping is normal. Because a device can recover on its own, wait for one
day of downtime, or one to two days of frequent flapping before investigating further.
Filtering and reviewing the event management program mail alerts regularly can help to
identify devices that need attention. The standard operating procedure regarding collector
failure/management is shown in Figure 5 on page 15.
Multicast / Unicast Communications
Commands sent from head-end monitoring applications (NEM, GridScape, or HCM)
through the Access Point (AP) are unicast, as the receipt of each packet must be
acknowledged. However, in similar fashion to multicast, commands can be sent to a
statically or dynamically defined group of RF devices for ease of manageability. In addition,
the HES sends “Jobs” out in parallel to Access Points, that then in turn, send commands out
to endpoints. The resulting effect is parallel processing that accomplishes a multicast-like
service.
Networking
Each NIC in Silver Spring RF devices registers with two IPv6 addresses – one for each of two
APs: a primary and secondary AP. When one dies, another is found. From a topology
perspective, each AP is its own IPv6 subnet. So NICs are in two different subnets - one for
each AP.
The end-to-end protocol within the NAN is AES 128/256 bit, crypto IPv6. As the Access Point
understands the full topology of the underlying mesh, the IPv6 packet is source routed from
the AP to each end device – meaning that the full communication path from the Access Point
to the device is specified, hop by hop, in the packet. The NAN endpoint to Access Point
routing decisions are made hop-by-hop (that is, the NAN endpoint will send the IPv6 packet
to its neighbor that has the least cost route to the Access Point. Then this neighbor will
forward the packet to its neighbor with the best route until the packet reaches the Access
Point. (The Access Point acquires its understanding of the full mesh topology because it
sends route advertisement messages at randomized intervals.)
NAN system one way latency is measured at 50ms per hop between any points in the NAN.
The Silver Spring network design expects on average, no more than 6 hops to any end point
(such as a meter or Data Link Control, DLC, device). This would typically yield 300ms of
NAN latency to any RF or DLC device. In addition, an Access Point can process up to 10
packets per second. Each DLC command consists of a single packet, with one associated
ACK packet. With no other AMI traffic traversing an Access Point, an Access Point could
process 100 DLC commands and ACKs, in approximately 30 seconds.
The cellular modem has store-and-forward GPS reporting capability with accuracy no less
than 10 meters.