Internet DRAFT - draft-coras-lisp-re
draft-coras-lisp-re
Network Working Group F. Coras
Internet-Draft A. Cabellos-Aparicio
Intended status: Informational J. Domingo-Pascual
Expires: May 4, 2016 Technical University of Catalonia
F. Maino
cisco Systems
D. Farinacci
lispers.net
November 1, 2015
LISP Replication Engineering
draft-coras-lisp-re-08
Abstract
This document describes a method to build and optimize inter-domain
LISP router distribution trees for locator-based unicast and
multicast replication of EID-sourced multicast packets.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on May 4, 2016.
Copyright Notice
Copyright (c) 2015 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
Coras, et al. Expires May 4, 2016 [Page 1]
Internet-Draft LISP-RE November 2015
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Definition of Terms . . . . . . . . . . . . . . . . . . . . . 3
3. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4. Overlay Signaling . . . . . . . . . . . . . . . . . . . . . . 5
4.1. RTR Registration . . . . . . . . . . . . . . . . . . . . 6
4.2. ETR/RTR Subscription . . . . . . . . . . . . . . . . . . 6
4.3. ETR/RTR Unsubscription . . . . . . . . . . . . . . . . . 7
5. Overlay Management . . . . . . . . . . . . . . . . . . . . . 8
5.1. RLOC Failure and Unreachability . . . . . . . . . . . . . 8
5.2. Other Overlay Management Considerations . . . . . . . . . 8
5.3. Automated Computation of RTR Level . . . . . . . . . . . 9
5.3.1. Algorithm for Computing Optimized Distribution Trees 9
6. Security Considerations . . . . . . . . . . . . . . . . . . . 10
7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10
8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 11
9. References . . . . . . . . . . . . . . . . . . . . . . . . . 11
9.1. Normative References . . . . . . . . . . . . . . . . . . 11
9.2. Informative References . . . . . . . . . . . . . . . . . 12
Appendix A. MADDBST heuristic . . . . . . . . . . . . . . . . . 12
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 12
1. Introduction
The Locator/Identifier Separation Protocol (LISP) [RFC6830] provides
the mechanisms for the separation of Location and Identity semantics
presently overloaded by IP addresses. The split results in the
creation of two namespaces that unambigously identify edge-site
network objects, Endpoint IDs (EIDs), and core routing objects,
Routing LOCators (RLOCs). Apart from aiding the scalablity of the
core routing infrastructure, the decoupling also enables the
(re)implementation of new or existing inter-domain routing
mechanisms.
One such mechanism is inter-domain IP source-specific multicast (SSM)
[RFC4607]. In this sense, [RFC6831] defines the procedures carried
out for delivering multicast packets from a source host in a LISP
site to receivers residing in the same domain or in other LISP or
non-LISP sites when an underlying multicast infrastructure exists.
The signaling protocol it specifies for conveying (S-EID,G) state and
building the distribution tree that connects the source ITR and the
receiving ETRs is PIM [RFC4601]. An alternative method that uses
Coras, et al. Expires May 4, 2016 [Page 2]
Internet-Draft LISP-RE November 2015
Map-Requests for propagating (S-EID,G) state from ETRs to the ITR is
established in [I-D.farinacci-lisp-mr-signaling].
Although desirable to use multicast routing in the core network when
available, a mismatch between the multicast capabilities of receiver
ETRs and source ITR might impede their interconnection. In such a
case, unicast RLOC encapsulation is necessary to deliver multicast
packets directly to the ETRs. This however leads to high ITR head-
end replication for large sets of ETRs. Therefore, to reduce the
replication load of the ITR and scale the service with the number of
multicast receivers, the ITR may choose to offload replication to a
set of RTRs.
The current document describes how multicast RTRs can be used to
build an inter-domain distribution tree rooted at the ITR that can
perform unicast and/or multicast encapsulated replication of
multicast packets. This concept, of distributing the replication
load from ITR to other RTRs downstream on the core overlay
distribution tree, is known as Replication Engineering or LISP-RE.
Since unicast replication in such overlays can be suboptimal when
compared to the underlay network, methods to optimize packet delivery
over the distribution tree are also presented.
This specification does not define the mechanisms used to build
(S-EID,G) state in source and receiver domains, nor does it describe
the messages used to propagate such state from receiver ETRs to
source ITR. What it defines is how (S-EID,G) state is built in the
ITR, RTRs and ETRs participating in the overlay distribution tree.
2. Definition of Terms
The terminology in this document is consistent with the definitions
in [RFC6830] and [RFC6831] however, it is extended to account for
LISP-RE concepts:
Delivery Group (DG): This is the outer destination address of a
packet when LISP encapsulating a multicast packet with an EID
source within a multicast packet.
Re-encapsulating Tunnel Router (RTR): An RTR is a router that
implements the re-encapsulating tunnel function detailed in
Section 8 of the main LISP specification [RFC6830]. Such router
performs packet re-routing by chaining ETR and ITR functions,
whereby they first remove the LISP header of ingressing packets
and then prepend a new one prior to forwarding them.
Unicast Replication: Is the notion of replicating a multicast packet
with an EID source address at an ITR or RTR by encapsulating it
Coras, et al. Expires May 4, 2016 [Page 3]
Internet-Draft LISP-RE November 2015
into a unicast packet. That is, the oif-list of a multicast map-
cache entry can not only have interfaces present for link-layer
replication and multicast encapsulation but also for site-facing
interfaces and unicast encapsulation.
Overlay Distribution Tree: A degree-constrained spanning tree that
represents the path followed by unicast and/or multicast
encapsulated multicast packets from the root (ITR) to the leaves
(ETRs) through intermediary nodes (RTRs). The ITR and RTRs
unicast and/or multicast replicate packets to their tree children.
LISP Replication Node: A router (either the ITR or an RTR)
participating and replicating packets downstream in the
distribution tree.
Multicast Ingress Tunnel Router (ITR): An ITR as specificed in
[RFC6830] that supports multicast and participates as the root in
the distribution tree. In this document we use the term "ITR" to
mean a multicast capable ITR.
Multicast Egress Tunnel Router (ETR): An ETR as specified in
[RFC6830] that participates as a leaf in the distribution tree.
ETR are the only members of the tree that do not unicast
replicate. In this document we use the term "ETR" to mean a
multicast capable ETR.
Multicast Re-encapsulating Tunnel Router (RTR): An RTR as specified
in [I-D.farinacci-lisp-te] that participates as an intermediary
node in the distribution tree. In this document we use the term
"RTR" to mean a multicast capable RTR.
Replication Target: A multicast channel-id (S-EID,G) or a set of
multicast channel-ids (S-EID-prefix,G).
Joining-OIF-list: Represents a collection of state per multicast
routing table entry at an RTR or ETR that is created by received
Map-Request/Join-Request.
Forwarding-OIF-list: Represents the outgoing RLOC list a multicast
router stores per multicast routing table entry such that it knows
to which RLOCs to replicate multicast packets. Although the
Joining-OIF-list contains sufficient information to allow the
forwarding of encapsulated multicast packets, using it is
inefficient. Thereby, an RTR implementation may wish to build an
efficient Forwarding-OIF-list. Ways of implementing a Forwarding-
OIF-list are out of the scope of this document.
Upstream: Towards the root of the tree.
Coras, et al. Expires May 4, 2016 [Page 4]
Internet-Draft LISP-RE November 2015
Downstream: Away from the root of the tree.
3. Overview
This document describes a method to diminish the ITR's replication
load by using RTRs to build an inter-domain distribution tree. The
tree is managed by the source domain's Map-Server. RTRs join the
overlay due to either manual or automatic configuration and advertise
to the Map-Server their availability to replicate traffic for a
multicast channel (S-EID,G). Out of all the RTRs registering for the
same multicast channel, the Map-Server builds one mapping and
organizes the RLOCs in a multi-level hierarchy. The hierarchy is
rooted at the ITR and computed based on the configured information
RTRs register or by means of local policy and algorithms. ETRs
always join the overlay as leaves and their attachment prompts the
creation of a path, which traverses the RTR hierarchy, from the ITR.
The path is built at receiver request by incrementally linking all
distribution tree levels, starting at the joining ETR up to the
source ITR.
The way the distribution tree is built has several benefits. First,
it ensures that packets in the source domain do not reach the ITR if
no ETR is joined. Second, it ensures that packets are forwarded from
ITR to all ETRs without mapping database lookups since the state that
defines the distribution tree, i.e., the replication hierarchy, is
created prior to forwarding/replicating the packets. Finally, the
multicast source is allowed to roam since a first level RTR, when
informed of the roam event, can do a new database lookup to find the
new ITR to join to.
It is worth pointing out that because of the receiver-initiated
approach multicast employs to build distribution trees, whereby
receivers join upstream sources, LISP-RE operates backwards from LISP
point of view. That is, ETRs are the ones to send Map-Requests to
discover potential upstream parents and the ITR answers with Map-
Replies to joining downstream clients.
4. Overlay Signaling
This section describes the signaling the ITR, RTRs and ETRs use in
order to participate in the overlay and build a distribution tree.
The signaling messages used are described in
[I-D.farinacci-lisp-mr-signaling] and [RFC6831].
Coras, et al. Expires May 4, 2016 [Page 5]
Internet-Draft LISP-RE November 2015
4.1. RTR Registration
RTR participation in the overlay is condition by the configuration of
a replication target, a multicast channel (S-EID,G) or set of
channels (S-EID-prefix,G), the RTR is to perform replication for.
Once configured, manually or by automated mechanisms, an RTR Map-
Registers its replication target with merge-semantics to the
appropriate Map-Server. In the registration it also provides its
list of RLOCs to be used by overlay peers and a set of corresponding
weights and priorities. If present, information about the level of
the hierarchy where the RTR should attach is also conveyed by means
of an Replication List Entry canonical address [I-D.ietf-lisp-lcaf].
Due to the merge-semantics, the Map-Server aggregates all RTR
originated Map-Register messages in a single, per replication target
mapping. If no level information is provided or if so configured,
the Map-Server should use local policy to compute a hierarchy and
associate a level within it to each entry in the list (more details
in Section 5.3). It should be noted that the entries that are
pointed to in the resulting mapping are not RLOCs but Replication
List entries.
4.2. ETR/RTR Subscription
When an ETR creates (S-EID,G) state from a site based multicast join,
i.e., its oif-list goes non-empty, it must send an upstream Join
request. If the ETR does not have multicast connectivity to its
upstream and unicast replication must be performed, the ETR requests
that a path from ITR to itself, over the RTR hierarchy be
constructed. The following procedure is followed to build the path:
1. ETR sends a Map-Request/Join-Request for (S-EID,G) multicast
channel to the mapping database system which further ensures its
delivery to the authoritative Map-Server.
2. The Map-Server looks up the mapping associated to (S-EID,G) and,
out of the distribution tree hierarchy encoded within, it selects
a set of leaf RTRs, i.e., members of the level furthest away from
the ITR, with spare replication capacity. The set of potential
parents is encoded in a new (S-EID, G) mapping the Map-Server
conveys to the ETR in a Map-Reply.
3. From the list it receives, the ETR selects the best upstream RTR
RLOC according to local policy, taking into account the
associated priorities and weights and sends to the owning RTR a
Map-Request/Join-Request for (S-EID,G). If the ETR itself has
multiple RLOCs it wishes to use in the overlay, it may convey
them all to the upstream RTR encoded in the Map-Reply field of
Coras, et al. Expires May 4, 2016 [Page 6]
Internet-Draft LISP-RE November 2015
the Map-Request/Join-Request together with associated priorities
and weights.
4. The RTR stores the ETR's subscription information in the join-
oif-list associated to (S-EID,G) and inserts the RLOC obtained
after evaluating the priorities and weights in the oif-list for
(S-EID,G). It then confirms the ETR's subscription with a Map-
Reply.
5. If not already a member of (S-EID,G), the RTR initiates it's own
attachment to the distribution tree by repeating the steps 1-4.
An important difference at step 2, the Map-Server replies to a
joining RTR with a list of RTRs in the adjacent upstream layer,
as opposed to a list of leaf RTRs, like in the case of an ETR
join. This procedure may recurse upstream up to when the ITR or
an RTR already attached to the distribution tree is joined. On
completion, there should exist a path from ITR to joining ETR.
6. If the ITR is already member of (S-EID,G) the process stops.
Otherwise, the ITR sends a PIM join to the intra-domain multicast
source ensuring the creation of a path from the multicast source
to the receiver end-hosts.
If at any point, when creating a link between two adjacent layers,
native multicast replication can be performed, instead of unicast
replication, the router joining its upstream could set as source of
the Map-Request/Join-Request a delivery group. However, group naming
must be coordinated between the participating parties in this case,
if core network replication is to be exploited.
4.3. ETR/RTR Unsubscription
When an ETR's oif-list goes empty a Map-Request/Leave-Request is sent
to the upstream RTR which will result in the removal of the ETR's
associated entry from the RTR's oif-list. The procedure is repeated
by the RTR, and it may recurse upstream, if its own oif-list also
goes empty.
When an RTR with active dowstreams departs, it should first change
the priority of the RLOCs it registers with the Map-Server to 255 and
set its locators as unreachable in the RLOC-Probing replies it sends
downstream. Finally, once all adjacent lower level members have sent
Map-Request/Leave-Request messages the RTR can stop registering
(S-EID,G) with the mapping database system and thus leave the
overlay.
Coras, et al. Expires May 4, 2016 [Page 7]
Internet-Draft LISP-RE November 2015
5. Overlay Management
5.1. RLOC Failure and Unreachability
RLOC failure is detected at control-plane level through RLOC-probing
[RFC6830] by both upstream and downstream routers. When an RTR
detects the failure of an downstream RLOC, it ceases replicating
towards it. The affected RLOC is removed from the forwarding-oif-
list and marked as unreachable in the join-oif-list. If a backup
RLOC was provided by the downstream router in the Map-Request/Join-
Request, it is instead inserted in the forwarding-oif-list and the
failure results in no packet loss.
The routers downstream of a failed RTR RLOC, or who lost connectivity
to said RLOCs, remove their Map-Request/Join-Request associated state
and reperform the join procedure. Packet loss in this case must be
solved by out-of-band mechanisms that are out of the scope of the
current document.
5.2. Other Overlay Management Considerations
An overloaded RTR, i.e., one whose fan-out can not be increased,
should change the priority of the RLOCs it registers with the mapping
database system to 255. In such a situation, the Map-Server updates
the associated mapping and informs all routers having requested it
about the change through solicit Map Request (SMR) messages. Both
new ETRs attaching to the distribution tree and those already
connected but reperforming the join procedure must not use the RLOCs
with a priority of 255 as specified in [RFC6830]. However, routers
having performed Join-Requests prior to the change should not break
their existing connections to the affected RTR.
All routers part of an (S-EID,G) multicast channel should re-evaluate
their attachment point to the distribution tree whenever the Map-
Server updates the associated mapping. This ensures the overlay
member routers attach to the best suited parent when new RTRs join or
previously attached ones stop being overloaded. Change of a parent
should be done following a "make before break" procedure.
Specifically, the router changing attachment point first connects to
the new parent and only afterwards sends the Leave-Request.
When a downstream RTR subscribes to a set of channel-ids (S-EID-
prefix,G) using multiple RLOCs in a load-balancing configuration, the
upstream RTR may choose to load-split channel-ids (S-EID,G) over the
given set of RLOCs.
Coras, et al. Expires May 4, 2016 [Page 8]
Internet-Draft LISP-RE November 2015
5.3. Automated Computation of RTR Level
Operators wishing to automate the RTR joining procedure may wish to
use an algorithm for computing an optimized distribution tree. The
algorithm could be implemented in the Map-Server and its output
should be used to associate to all RTRs a level in the distribution
tree. Due to the centralized management, on-line switching between
algorithms may be possible in accordance to the required distribution
tree performance. However, their use of such algorithms is dependent
on the presence of overlay topological information. Ways of
obtaining topological information will be discussed in future
versions of this document.
5.3.1. Algorithm for Computing Optimized Distribution Trees
The current document does not recommend an algorithm for computing
optimized distribution trees. However, it provides as an example a
low computation cost heuristic, which, in the scenarios simulated in
[LCAST], can produce latencies between the ITR and the multicast
receivers close to unicast ones. Its choice is to be influenced by
operational requirements and the hardware constraints of the
equipment in charge of running it. Future experiments might result
in a recommendation.
In what follows, we use the term "distance" when referring to a
relative length or amplitude of a metric, observed on a path
connecting two points, but when the exact nature of the metric is of
no interest.
Considering as goal the delivery of content for delay sensitive
applications, the function the algorithm minimizes is the maximum
distance (e.g. latency or number of AS hops) from a multicast
receiver to the ITR source. Notice that the reference is the
multicast receiver host and not an ETR. Thus, what matters in
deciding a member's position in the distribution tree is not solely
its distance to the ITR but also the number of multicast receivers it
serves. Then, a router close to the source but serving few receivers
might find itself lower in the distribution tree than another with a
slightly higher distance to the source but with a larger receiver
set. The algorithm optimizes the quality of experience for multicast
receivers and not for tunnel routers.
The problem described above, that searches for a minimum average
distance, degree-bounded spanning tree (MADDBST), can be formally
stated as:
Definition: Given an undirected complete graph G=(V,E), a designated
vertex r belonging to V, for all vertices v in V, a degree bound
Coras, et al. Expires May 4, 2016 [Page 9]
Internet-Draft LISP-RE November 2015
d(v) <= dmax, dmax a positive integer, a vertex weight function
c(v) with positive integer values, and an edge weight function
w(e) with positive values, for all edges e in E. Let W(r,v,T)
represent the cost of the path linking r and v in the spanning
tree T. Find the spanning tree T of G, routed at r, satisfying
that d(v) <= dmax and the distance to the source per multicast
receiver is minimized.
The heuristic used to solve this problem works by incrementally
growing a tree, starting at the root node r, until it becomes a
spanning tree. For each node v, not yet a tree member, it selects a
potential parent node u in the tree T, such that the distance per
receiver to r, is minimized. At each step, the node with the
smallest metric value is added to the tree and the parent selection
is redone. The pseudocode of the heuristic is provided in
Appendix A.
[SHI] and [BAN] have previously defined and solved similar
optimization problems. Shi et al. [SHI] also prove that a
particular instance of the problem, where all vertices have weight 1,
is NP-complete for degree constraints 2 <= dmax <= |V|-1.
The algorithm can optimize an unicast overlay however, it should not
be used to optimize multicast underlay delivery. As a result, if
multicast is used as underlay between part of the overlay members,
once one of the members of such Delivery Group is added to the
distribution tree, the others should be marked as attached also.
These nodes should receive multicast encapsulated multicast packets
from the chosen node over the underlying multicast distribution tree.
Finally, since the RTRs do not replicate packets for multicast
receiver hosts, prior to applying the MADDBST heuristic, a Minimum
Spanning Tree (MST) algorithm should be used to compute the RTR
distribution tree. In this case, the MADDBST heuristic should start
attaching ETRs having as input the tree resulting from MST.
6. Security Considerations
Security concerns for LISP-RE the same as for [RFC6831] and
[I-D.farinacci-lisp-mr-signaling].
7. IANA Considerations
This memo includes no request to IANA.
Coras, et al. Expires May 4, 2016 [Page 10]
Internet-Draft LISP-RE November 2015
8. Acknowledgements
The authors would like to thank Noel Chiappa for his technical and
editorial commentary.
9. References
9.1. Normative References
[I-D.farinacci-lisp-mr-signaling]
Farinacci, D. and M. Napierala, "LISP Control-Plane
Multicast Signaling", draft-farinacci-lisp-mr-signaling-06
(work in progress), February 2015.
[I-D.farinacci-lisp-te]
Farinacci, D., Kowal, M., and P. Lahiri, "LISP Traffic
Engineering Use-Cases", draft-farinacci-lisp-te-09 (work
in progress), September 2015.
[I-D.ietf-lisp-lcaf]
Farinacci, D., Meyer, D., and J. Snijders, "LISP Canonical
Address Format (LCAF)", draft-ietf-lisp-lcaf-11 (work in
progress), September 2015.
[RFC4601] Fenner, B., Handley, M., Holbrook, H., and I. Kouvelas,
"Protocol Independent Multicast - Sparse Mode (PIM-SM):
Protocol Specification (Revised)", RFC 4601,
DOI 10.17487/RFC4601, August 2006,
<http://www.rfc-editor.org/info/rfc4601>.
[RFC4607] Holbrook, H. and B. Cain, "Source-Specific Multicast for
IP", RFC 4607, DOI 10.17487/RFC4607, August 2006,
<http://www.rfc-editor.org/info/rfc4607>.
[RFC6830] Farinacci, D., Fuller, V., Meyer, D., and D. Lewis, "The
Locator/ID Separation Protocol (LISP)", RFC 6830,
DOI 10.17487/RFC6830, January 2013,
<http://www.rfc-editor.org/info/rfc6830>.
[RFC6831] Farinacci, D., Meyer, D., Zwiebel, J., and S. Venaas, "The
Locator/ID Separation Protocol (LISP) for Multicast
Environments", RFC 6831, DOI 10.17487/RFC6831, January
2013, <http://www.rfc-editor.org/info/rfc6831>.
Coras, et al. Expires May 4, 2016 [Page 11]
Internet-Draft LISP-RE November 2015
9.2. Informative References
[BAN] Banerjee, S., Kommareddy, C., Kar, K., Bhattacharjee, B.,
and S. Khuller, "Construction of an efficient overlay
multicast infrastructure for real-time applications",
INFOCOM , 2002.
[LCAST] Coras, F., Cabellos, A., Domingo, J., Maino, F., and D.
Farinacci, "Lcast: Software-defined inter-domain
multicast", Computer Networks , 2014.
[SHI] Shi, S., Turner, J., and M. Waldvogel, "Dimensioning
server access bandwidth and multicast routing in overlay
networks", NOSSDAV , 2001.
Appendix A. MADDBST heuristic
INPUT: G = (V,E); r; dmax; w(u,v); c(v); u, v in V
OUTPUT: T
FOREACH v in V DO
delta(v) = w(r,v)/c(v);
p(v) = r;
END FOREACH
T takes (U = {r}, D={});
WHILE U != V DO
LET u in U-V be the vertex with the smallest delta(u);
U = U U {u}; L = L U {(p(u),u)};
FOREACH v in V-U DO
delta(v) = infinity;
FOREACH u in U DO
IF d(u) < dmax and
W{r,u,T} + w(u,v)/c(v) < delta(v) THEN
delta(v) = W{r,u,T} + w(u,v)/c(v);
p(v) = u;
END IF
END FOR
END FOR
END WHILE
Figure 1
Authors' Addresses
Coras, et al. Expires May 4, 2016 [Page 12]
Internet-Draft LISP-RE November 2015
Florin Coras
Technical University of Catalonia
C/Jordi Girona, s/n
BARCELONA 08034
Spain
Email: fcoras@ac.upc.edu
Albert Cabellos-Aparicio
Technical University of Catalonia
C/Jordi Girona, s/n
BARCELONA 08034
Spain
Email: acabello@ac.upc.edu
Jordi Domingo-Pascual
Technical University of Catalonia
C/Jordi Girona, s/n
BARCELONA 08034
Spain
Email: jordi.domingo@ac.upc.edu
Fabio Maino
cisco Systems
Tasman Drive
San Jose, CA 95134
USA
Email: fmaino@cisco.com
Dino Farinacci
lispers.net
Email: farinacci@gmail.com
Coras, et al. Expires May 4, 2016 [Page 13]
