USC Computer Science Technical Reports, no. 804 (2003)

PDF

Download

Open document

Flip pages

Download a page range

Ahmed Helmy. "Efficient provisioning for services in large-scale wireless-wired networks." Computer Science Technical Reports (Los Angeles, California, USA: University of Southern California. Department of Computer Science) no. 804 (2003).

Transcript (if available)

Content E FFICIENT P ROVISIONING FOR S ERVICES IN L ARGE - SCALE W IRELESS - W IRED N ETWORKS
A HMED H ELMY
D EPARTMENT OF E LECTRICAL E NGINEERING , U NIVERSITY OF S OUTHER C ALIFORNIA
1. O BJECTIVE
This document outlines an architectural design for future internets, that are envisioned to include
integrated wired and wireless (ad hoc) networks. Although such an environment presents several
research challenges, we believe that the promise in such networks substantially outweighs the risk,
and we shall show that the research directions proposed herein hold promise to overcome the
challenges. One major challenge in ad hoc networks – that differentiates such networks from the
Internet – is the lack of infrastructure (such as autonomous systems ( ASs) or DNS). Another major
challenge is mobility , which leads to a highly dynamic network that cannot be managed using
conventional networking techniques. Moreover, many services in ad hoc networks may be data-
centric . Meaning that the goal of the network may be to gather, store, and retrieve data with node id - independent semantics (e.g., spatial data). For such applications, conventional communication
paradigms based on node IDs (such as those used in today’s Internet, i.e., IP addresses) may not be
suitable. Therefore, we introduce a novel architecture for ad hoc networks that provides efficient
resource discovery (whether these resources are defined by physical node IDs, location, data, or in
general logical IDs). In doing so, this proposal provides the building blocks for providing services in
such emerging networks. We design our architecture to be self-configuring, scalable, efficient and
robust. In addition, we provide an architecture for the necessary ‘glue’ that ties ad hoc and wired
networks. Using a single mechanism (called multicast-based mobility) we provide efficient handoff
solutions for unicast and multicast scenarios.
Although the solutions set forth in this proposal may be generalized for classes of ad hoc
networks, we shall target specific example applications that illustrate the utility of such technology.
In particular, we choose two applications: (a) person tracking and location, and (b) wildfire fighting.
We believe that without our proposed architectures, these applications cannot be enabled efficiently
in large-scale mobile wireless networks using existing technologies. This will be explained further in
more detailed discussion of these applications, in addition to the related work section.
• Motivation
With the recent advances in microelectronics and communications, PDAs, wearable computers,
and embedded systems are predicted to proliferate and provide the future platform for ubiquitous
computing and communications. This will potentially enable several new classes of applications, such
as location specific services, person and object localization, and search and rescue missions.
These networks may scale up to tens of thousands of wireless mobile nodes, some of which may be
power constraint. Developing efficient protocols for such environments provides several research
challenges. First, the protocols need to be designed with power consumption in mind. One main
2 factor of power consumption is communication. Hence, power awareness necessitates designing
protocols with reduced communication overhead. Second, in such networks all nodes are potentially
mobile. Such mobility produces a highly dynamic network, in which network connectivity is
constantly changing. Unlike wired networks, these changes cannot be dealt with as failures, since
they are part of the normal operation of the network. Ad hoc network protocols must adapt
efficiently to mobility and network dynamics. Third, these networks may contain thousands of nodes.
Hence, the protocols designed must scale with network size (i.e., performance of these protocols
must be acceptable with the increase in number of nodes in the network). Fourth, the lack of
infrastructure in ad hoc networks creates a barrier for service provisioning (such as multicast,
location-based services or storage location). Without fixed servers with well-known addresses it is
not possible to bootstrap these services with existing technologies. The protocols designed must be
self-configuring and must enable efficient resource discovery without any reliance on existence of
well-known servers or infrastructure. Fifth, we do not expect services in ad hoc networks to be
necessarily linked to physical IDs (such as IP addresses). We envision various services (such as
location-based services) to be linked to logical IDs instead. The designed protocols must have
support for mapping logical IDs efficiently onto physical resources. This is what we call data-centric
networking. Finally, these networks will be integrated with the wired Internet. The architecture
proposed must allow efficient and seamless integration of ad hoc networks with the wired networks.
• Research hypothesis
The main architectural components of our design target large-scale ad hoc networks. As was
mentioned above, the infrastructure-less nature of such networks brings about the need for efficient
resource discovery. We provide two main mechanisms to achieve efficient resource discovery. The
first is based on the small world phenomenon and establishes a contact-based architecture in wireless
networks. We provide a unique approach for mobility-assisted contact selection that utilizes mobility
to improve performance. The second mechanism provides a robust rendezvous mechanism to
bootstrap services in such networks. To the best of our knowledge, this is the first work to address
service bootstrap in large-scale ad hoc networks. In addition, we provide a component to tie wireless
networks into wired networks efficiently using a multicast-based mobility scheme. Hence, the three
main components of our architecture are (I) contact based mechanisms, (II) rendezvous mechanisms,
and (III) multicast-based mobility.
I – Contact-based Mechanisms: Building a Small World in Mobile Wireless Networks
Self- configurability and infrastructure-less- ness of ad hoc networks provides resilience in the face
of network dynamics and failures. In the design of infrastructure-less networks, efficient resource
discovery is an essential component. Current approaches for resource discovery employ two main
techniques. The first is flooding (or broadcast) where a request is sent to every node in the network
to locate the resource. Many approaches for ad hoc networks depend on flooding for discovery (e.g.,
DSR [26] , AODV [27] ). This approach consumes a lot of communication resources and energy and
does not scale well with the size of the network. The second is cluster-based hierarchies [47] , where
nodes would form clusters, electing a cluster-head within each of these clusters to be responsible for
inter-cluster communication. Although in general they perform better than flooding, cluster-based
hierarchies suffer from susceptibility to major re-configuration with failure or movement of nodes
(especially cluster-heads). This is due to the complex coordination needed to setup the clusters.
Frequent re-configuration wastes network resources and introduces undesirable network transients.
3 Other approaches use landmark hierarchy [44] (e.g., LANMAR [46] and SCOUT [45] ). These
approaches do not perform well in highly dynamic, high mobility environments. The zone routing
protocol (ZRP) [4] [29] [30] uses the concept of zones with a table-driven protocol for intra-zone
routing and on-demand protocol for inter-zone routing. The on-demand protocol uses flooding
between borders of the zones (called ‘ bordercasting’). Although ZRP does not include complex
hierarchical schemes, and performs better than flat flooding, it still uses global flooding between
borders. In our approach we avoid bordercasting and instead use contacts, which proved to perform
much more efficiently as we will show. We shall discuss these approaches further in the related work
section.
Edge Node
Center Node
Internal Node
R S 5 1 4 3 6 7 2 Vicinity Radius
Figure 1 Vicinity from the perspective of node S as a center node. S knows about resources (internal and edge nodes) up to R hops away using table-driven protocol. Nodes at exactly R hops are called edge nodes.
To overcome the above limitations, we introduce a new loosely-coupled hierarchical architecture
for resource discovery that is based upon the small world phenomenon. In our architecture, a
wireless node knows about resources in its vicinity (up to R hops away). This is achieved through a
localized table-driven routing protocol (e.g., DSDV [25] ). Note that every node has its own vicinity,
and hence has its own view of the network. This is shown in Figure 1 . Unlike complex hierarchies,
our architecture does not require major re-configuration due to dynamics , each node simply changes
its own view of the network without coordination with other nodes. For out-of-vicinity resources, a
node queries a small number of nodes (called contacts ) outside its vicinity. Contacts increase the
network view of a node and act as shortcuts to form a small world in the wireless network. We call
our approach the ‘contact-based architecture for resource discovery’ ( CARD ). We have shown in
[2] [3] that CARD performs drastically better than the previous approaches. An interesting problem
introduced by such architecture is the selection and maintenance of the contacts. We address such
problem next.
Contact Selection and Maintenance Schemes
One approach for contact selection would be to send random recruiting messages outside of the
vicinity to build those contacts. However, random contact selection may lead to incomplete network
coverage and unpredictable performance. One improvement would be to choose contacts in a way
that minimizes vicinity overlaps between the contacts and the original node. At the same time, we
want to limit the distance at which the contacts may be selected (to reduce the maintenance
4 overhead)
1 . To reduce overlap we include a TTL (hop count) field in the message. Unfortunately,
once the selection message goes out of the vicinity, it has no sense of direction and it is not possible
to depend only on hop count to infer overlap information (since a node knows about other nodes
only within its vicinity). This leads to diminishing returns in terms of reachability with the increase in
number of contacts; as more contacts are selected the number of nodes reachable tends to saturate.
At the same time, looking for a good contact (with minimum overlap) becomes harder and the
protocol incurs increased overhead (due to search and backtracking). To ameliorate this problem we
include the list of edge nodes (those at the edge of the vicinity) in the contact selection message. This
provides nodes along the selection path with information to infer overlap, and hence leads to
increased reachability. This also leads to drastically reduced overhead due to reduction in
backtracking [2] . R S Z A B C
R R S Z contact
A B C (a) ( b)
Figure 2 . Lightweight contact maintenance protocol (a) Z is at the edge of S ’s vicinity with a route ‘ S-A-B-Z’. Z is moving out
of vicinity. (b) Z is no longer within S’ s vicinity, but it highly likely to exist in B’ s vicinity. The route ‘S-A-B-C-Z’ is
identified and Z is selected as a contact.
For contact maintenance, we use a lightweight routing protocol that leverages already-existing
vicinity information for each node. If a contact moves, then the previous hop node (in the original
path) will have a valid route to the contact so long as it is within R hops away ; a highly likely
scenario in most cases. This is shown in Figure 2 . In [2] [3] we present more details on this approach
and show that our protocol converges on stable contacts over time leading to reduced maintenance
overhead and better reachability.
We have compared our CARD approach to flooding and bordercasting (employed by ZRP) using
simulations over various topologies using random way point mobility and random queries. As shown
in Figure 3 , flooding performance degrades drastically with the scale of the network. Bordercasting
(i.e., ZRP) performs better than flooding, but the best performance was achieved using CARD . Even
after adding the overhead of contact selection and maintenance, CARD ’s total overhead is less than
‘third’ that of ZRP query overhead for large networks. This is achieved by avoiding border flooding
(or bordercasting) and instead using contact queries.

1 In [1] we have shown that limiting the distance of contacts to around 20%-30% of the network diameter does indeed result in a
small world graph and achieves the best reduction in average degrees of separation (ADS). ADS in our context denotes the average
number of nodes queried before reaching the target. Reducing ADS translates into reduction in search overhead.
5 0 1000
2000
3000
4000
5000
6000
7000
8000
250 500 1000
Number of nodes
Querying Traffic (packets) Flooding
Bordercasting
CARD
CARD Overhead
Figure 3 . Simulation comparison of the contact-based approach to flooding and bordercasting (employed by ZRP)
In [1] [6] we introduce another novel approach for contact selection that takes advantage of
mobility . In that approach, a node first selects candidate contacts from its vicinity. While these nodes
move and exchange messages with the original node (through the table-driven protocol) those nodes
with desirable energy and mobility patterns are kept as candidate contacts. As the candidate contacts
move out of the vicinity, such that they have minimum overlap with other contacts and the original
node, they are promoted to become contacts and are used in resource discovery. This approach is
illustrated in Figure 4 . Unlike other approaches that consider mobility a liability, this novel protocol
utilizes mobility to enhance its performance. In this sense, this protocol is quite unique as its
performance improves with mobility , whereas other works on ad hoc networks clearly state that their
performance degrades with mobility.
One drawback of this approach, is its dependence on mobility. Hence, in non-mobility or low-
mobility scenarios, this approach may perform poorly.
As part of this research, we propose to design, develop and evaluate a hybrid, mobility-
adaptive, approach that uses mobility-assisted protocols when suitable, but adapts by using efficient
contact selection schemes during non-mobility scenarios. In addition, we shall re-visit mechanisms
for minimum overlap detection. Currently, sending the list of edge nodes may incur high overhead.
We plan to use encoding techniques (such as Bloom filters [22] and their variants [23] [24] ) to detect
vicinity overlap statistics in a communication-efficient manner. Furthermore, we plan to evaluate the
above protocols extensively using various topologies, mobility models [18] and communication
patterns. This shall be done using network simulation (NS-2 [13] ) as a first step, then using a
laboratory test-bed for in-door and out-door environments in later stages.
6 Edge Node
Center Node
Internal Node
Mobility
R S 5 1 4 3 6 7 2 Vicinity Radius
Route

R R R R S 5 1 4 3 6 7 2 contact
contact
contact
(a) ( b)
R R R R S 5 1 4 3 6 7 2 contact
contact
( c)
Figure 4 Example of vicinity, contacts and effect of mobility: (a) Vicinity for source node S is shown (with radius R ). Edge
nodes are numbered (1-7). Nodes 1,3 and 6 are moving/drifting out of vicinity. (b) Radii for vicinities of the drifting nodes are
shown. S stays in contact with the drifting nodes, which enables it to obtain better network coverage with low overhead. (c)
After moving away, contact nodes drift up to a point where their vicinities no longer intersect with S ’s vicinity. In this
example, S maintains contact with those nodes not more than (2 R +1) hops away, i.e. nodes 3 and 6, and loses contact with
node 1 as it drifts farther than the contact zone.
Summary I:
We propose a contact-based architecture for efficient resource discovery in large-scale ad hoc
networks. Our architecture is based upon the small world phenomenon and leads to significant
reduction in degrees of separation, and hence reduces the query overhead required for discovery. We
have shown that our protocols improve upon existing approaches drastically. In addition, mobility-
assisted techniques promise even more improved performance with mobility. This architecture
provides the first component upon which we shall provide our service bootstrap schemes, discussed
next.
II- Rendezvous Mechanisms: Bootstrapping Services using Rendezvous Regions ( RR s)
The contact-based architecture presented above provides a base upon which we can build a
model for bootstrapping services in large-scale ad hoc networks. Our bootstrap model uses a new
7 concept that we call rendezvous regions ( RRs ). Our model extends to a large class of generic
services. Examples of such services include (but are not limited to) multicast session establishment
and directory, location-based services, and distributed storage and retrieval systems. The main idea is
to self-configure the nodes into dynamically promoted servers that share responsibility of updating
the service information. The only information the nodes need is a mapping scheme that maps service
identifiers into physical regions. To illustrate, we consider a multicast service example. The multicast
space is broken into group prefixes ( Gprefix ). Each Gprefix is assigned a RR . The mapping ‘ Gprefix
i ↔ RR
i ’ is provided to all the nodes (either as a simple table or a hash function). Using that mapping,
any sender and receiver for a group G can consistently determine the corresponding RR and send
update/query messages to it. The messages are forwarded to the RR with the aid of contacts using
approximate location information. For this rendezvous scheme, we assume that nodes know their
approximate locations
2 . A subset of the nodes that reside in RR
i are elected using local promotion
schemes to maintain Gprefix
i information. We call those elected nodes service discovery servers
( SDSs ). Figure 5 shows the basic mechanisms of this scheme. If one of the servers moves out of the
RR it would send a leave message that would trigger another node to promote itself as a server.
Hence, this mechanism is self-replenishing and adapts gracefully to mobility and network dynamics.
geo -cast
RR1 RR2 RR3
RR4 RR5 RR6
RR7 RR8 RR9
contact
S Sender to group G RR 1 Gprefix 1 RR 2 Gprefix 2 … ... RR n Gprefix n G Î
Gprefix
i «
RR
3 R Receiver for G
G Î
Gprefix
i «
RR
3 service discovery server
Figure 5 . Solving the multicast rendezvous problem: In scalable multicast members of a group need only know the group ID
( G ) without prior knowledge of other members. Our approach: (1) Network is broken into rendezvous regions (RR) and the
multicast address into group prefixes ( Gprefix). Nodes need only know mapping between Gprefix
«
RR . (2) In RRi servers are
dynamically elected locally to serve Gprefix i . (3) Senders use mapping to store their information at servers in RRi . (4)
Receivers use mapping to retrieve information from servers in RR . (5) Contacts are used to discover servers efficiently.
This rendezvous concept can also be applied to storage and retrieval networks where a file is
mapped into RR , and so on. Also, location-based services can be naturally supported by this scheme
where location-specific information would be stored at servers in RR containing that location. Parties
interested in such information use the query scheme provided above to reach the corresponding RR
efficiently and obtain the information.
In this proposal, we plan to further detail the design of our rendezvous scheme, provide protocol
specifications for the mapping algorithm, approximate geographic routing using contacts, and the

2 Since this scheme utilizes contacts, the location information used need only be approximate and rough, thus this scheme is more
robust than other geographic based routing in the face of imprecise location information.
8 directory service bootstrap mechanism. We will further investigate enhanced robustness by using
multiple mappings (or hash functions), and peer-to-peer network concepts (e.g., CAN [37] , Chord [38] , freenet [39] ). Multiple mappings would lead to more replication of data in the network,
which will also incur extra overhead. There is a spectrum of possibilities between replicating the data
in all (or some) RR s, or only replicating forwarding pointers to the data and having the query follow
the pointers. We shall investigate various replication/query strategies and study such design trade-off.
Furthermore, we plan extensive evaluation of the proposed algorithms and comparison with other
geographic routing schemes (e.g., LAR [35] , GLS [28] and GPSR [36] ). One aspect that we plan to
investigate (that has not been investigated before) is the performance of these protocols under
imprecise location information. We predict that our scheme (that operates using contacts on the
order of 2R hops apart) is more robust to errors and imprecision in location measurement and
estimation than other geographic routing schemes. This is yet to be tested.
Summary II:
We propose a rendezvous mechanism to bootstrap services in large-scale ad hoc networks.
Our mechanism uses simple mapping between service resources into physical regions. Those regions
hold dynamically elected servers responsible for maintaining the service resource. Consistent
mapping allows different nodes to rendezvous at a region. This scheme is robust to mobility and
errors in location estimation. Example services supported include multicast, storage-retrieval systems
(such as peer-to-peer networks), and location-based services.
III- Multicast-based Mobility (M&M) to Integrate Wired and Wireless Networks
Now that we have outlined several techniques that address efficient resource discovery and
service bootstrap in large-scale ad hoc networks, we need to provide an efficient architecture by
which mobile wireless nodes can communicate with wired nodes through the Internet infrastructure.
A possible architecture for such heterogeneous network is to have multiple ad hoc networks, each of
which may be connected to the Internet through one or more base stations (or access routers), and is
considered as a LAN in that sense. One of the main problems associated with such architecture is the
handoff problem; the problem of maintaining continuous communication while moving between
access points. Several micro mobility protocols have been proposed to address this problem in the
context of IP mobility (i.e., last hop wireless). All such solutions, however, target unicast data only.
By contrast, we provide a solution that enables efficient handoff for both unicast and multicast data.
We call our solution multicast-based mobility (M&M). Mobile nodes in ad hoc networks may be
deployed to carry out specific tasks (e.g., search and rescue or disaster relief), and hence multicast
communication plays an important role in such networks to enable their efficient collaboration. In [7]
we introduced the main architecture. In [19] [20] [21] we introduced detailed mechanisms for using
M&M for micro-mobility ( intra -domain) and show that it outperforms most micro-mobility (e.g.,
hierarchical MIP [15] , seamless handoff [50] , Hawaii [17] ), while achieving very similar performance
(in terms of loss and delay during handoff) to the best micro-mobility approach (cellular IP ‘CIP’
[16] ). Our architecture also allows M&M to co-exist with Mobile IP (MIPv4 [9] ) and MIPv6 [11] as
inter-domain protocols. As was pointed before, M&M is also suitable for multicast support to mobile
users while CIP was designed for unicast only. Furthermore, M&M can support reactive and
proactive handoff schemes that minimize handoff delays and losses, that no other unicast-based
scheme can provide [21] . For a campus network to use M&M, the only requirement is to deploy
multicast within that campus network or domain. Similarly, other micro-mobility approaches (e.g.,
9 CIP and Hawaii [17] ) require deployment of protocol-aware routers within that domain. Also,
dynamic configuration and address discovery (i.e., DHCP) may be used, as it would be for CIP or
Hawaii, to obtain the regional mobile address from the visited sub-network. In addition, M&M (by
virtue of using multicast routing) is more robust than other micro-mobility architectures in the face of
failure of the root- router
3 . Following we shall illustrate how M&M is used to provide efficient handoff for unicast traffic,
(when Mobile IP is the inter-domain mobility protocol). In basic multicast-based mobility, each
mobile node (MN) is assigned a multicast address, instead of a unicast address . The MN, throughout
its movement, joins this multicast address through locations it visits. Senders wishing to send to the
MN send their packets to its multicast address, instead of unicast. Because the movement will be to a
geographical vicinity, it is highly likely that the join from the new location, to which the mobile
recently moved, will traverse a small number of hops to reach the already-established multicast
distribution tree. Hence, performance during handoff improves considerably. An overview of this
architecture is given in Figure 6 . As the MN moves, it joins to the assigned multicast address through
the new access router. Once the MN starts receiving packets through the new location, it sends a
prune message to the old AR to stop the flow of the packets down that path. Thus completing the
smooth handoff process.
Sender ( S ) Join
Prune
Mobile Node ( MN )
S MN

S MN
(a) ( b) (c)
Figure 6 . Multicast-based mobility. As the MN moves, as in (b) and (c), the MN joins the distribution tree through the new
location and prunes through the old location.
To allow for gradual deployment of our scheme and co-existence with MIP, we limit our
architecture to a domain. When a mobile node moves into a new domain, it is assigned a unicast
address that is unique within the domain, called regional care of address (RCOA). The RCOA is
chosen from a pre-defined subnet address used only for mobile nodes. This is needed to identify
packets sent to mobile nodes. The RCOA maps to a multicast care of address (MCOA). The MCOA
is used for routing packets within the domain, so there is no need to assign COA at every subnet.
These multicast addresses are domain- scoped for micro-mobility (i.e., they have only local
significance within the domain). In our scheme there is a one-to-one mapping between an RCOA and

3 The root- router, as defined in Hawaii or CIP, is the router at which the micro mobility tree is rooted and through which
traffic destined to mobile nodes funnels. This constitutes a single-point-of-failure. To alleviate such a problem, these micro-
mobility protocols need to introduce replication or redundancy solutions that, in turn, introduce a new set of consistency problems
during failure or healing of routers. M&M uses the underlying multicast routing protocol (e.g., PIM-SM) that already has built-in
mechanism for dynamic election of rendezvous-points ( RPs). M&M does not specify a single root- router, rather it uses the border
routers to map the packets destined to mobile nodes into multicast packets. Failure of the RP or a border router will automatically
trigger dynamic mechanims (in the underlying multicast routing) to adapt gracefully and recover from failure.
10
MCOA. When a mobile node moves into a new domain it is assigned RCOA by the access router
(AR) and the mobile performs inter-domain handoff; i.e., it registers the RCOA with its home agent
(HA) for MIPv4 or the corresponding node (CN) for MIPv6 with route optimization. The AR
automatically infers the multicast address (MCOA) for the mobile node from the assigned unicast
address (RCOA) through a straightforward algorithmic mapping (as described in [21] ). The AR then
triggers a Join message for MCOA to establish the multicast tree. Packets destined to the MN’s
home address are tunneled to its RCOA by the HA. When these packets arrive in the foreign domain,
they are identified by the border router (BR) as being destined to a mobile node. As shown in Figure
7 , the BR maps the destination unicast address to the multicast address and transmits the packets to
the MN down the multicast tree. In this case the multicast tree consists of the branch leading to the
mobile node (explicit join mechanisms are used, and so no broadcast is performed).
AR1
AR2
BR
Internet
AP
RCOA
MCOA RCOA
MCOA Algorithmic mapping
unicast
multicast
Mobile Node
Figure 7 . High level architectural view: Data packet is unicast over the Internet destined to the RCOA and arrives at the
border router (BR) for the mobile node. The BR intercepts the packet and performs algorithmic mapping from the RCOA to
MCOA. The packet is then multicast within the domain to the mobile node.
If a mobile node goes into ‘sleep’ or power-saving mode, when it wakes up it uses its RCOA to
contact the nearest access router, the access router maps the RCOA into MCOA using algorithmic
mapping, and joins to the multicast tree. We have evaluated our protocol using detailed NS-2
simulations over a rich set of topologies, mobility scenarios, and cell overlap conditions. We note
that geographic vicinity (in mobility) may not necessary lead to network vicinity (i.e., handing off to
a new access router that is directly connected to the previous access router). This was considered in
our evaluation. Based on our previous studies, M&M holds a lot of promise for a very efficient
handoff scheme, achieving low loss and delay during handoff, and traversing, on average, the least
number of hops to the already established tree [20] [21] . M&M’s application and evaluation so far
have been studied for single-hop (last-hop) wireless. As part of this proposal we shall investigate its
application as a mechanism to tie large-scale ad hoc networks to the Internet efficiently. Note that
multicast addresses are logical addresses (unlike unicast IP addresses), and thus are quite suitable for
use with our overall scheme described above (based on contacts and RR s). Another aspect that will
be studied is the interoperability between existing Internet protocols, M&M and ad hoc network
protocols. Since our architecture allows operation with no change to the mobile node (the access
router does the multicast functionality), then we expect smooth interoperation with existing
protocols. The M&M can be used seamlessly to provide efficient handoff. A mobile connecting to an
ad hoc network that, in turn, connects to the Internet through a new access router, would register its
RCOA with the access router which maps it to the MCOA and triggers a join, so on. This, and the
multicast scenarios, shall be investigated further and will be carried out through simulations and test
11
bed experimentation. We may also investigate M&M between different heterogeneous LANs (e.g.,
802.11 and GPRS). Another intriguing problem would be to explore the operation of M&M over a
combination of IPv4 and IPv6.
Summary III:
We propose a multicast-based mobility scheme to integrate wired and wireless networks. Our
approach leverages multicast technology to enable very efficient handoff performance between
access routers. This was shown through extensive simulation and comparison studies. This
architecture may be used for unicast and multicast traffic and we consider it a suitable choice for the
glue required between wired and wireless ad hoc networks.
IV- Potential Target Applications
To illustrate the utility of our proposed protocols and architectures we discuss two of its
potential applications; namely person/object tracking and wildfire fighting. In this work we propose
to tailor the design, parameter setting and trade- offs in our architecture to each of the above
applications. This will be performed as part of our efforts to provide a systematic procedure that can
be followed for other applications that belong to the same (or similar) classes of applications.
A- Tracking people and objects using ad hoc networks
Tracking technologies are quite useful and sometimes are crucial in emergency situations.
Unfortunately, usefulness of such technologies is limited to the coverage of the supporting systems
(e.g., the cellular-network provider). The ability to configure the tracking nodes in an ad hoc
network fashion could potentially increase coverage of the tracking system drastically. An
interesting application of tracking systems is in monitoring and tracking of people (including children
and elderly people). Recent advances in electronics and nano-technology have enabled new kinds of
biomedical sensing. Utilizing such technologies, we envision near-future devices that monitor vital
signs (e.g., pulse, blood pressure, temperature and blood glucose levels). This enables a device to
detect alarming changes in ones health. These devices will have computing and wireless
communication capabilities. Networking such devices in an ad hoc network and reporting health and
location information during emergencies may prove quite valuable in providing much needed proper
and medical assistance. This vision may constitute the heart of the future, fully automated,
emergency 911 system. Without a scalable ad hoc architecture that adapts gracefully to mobility,
such application cannot be realized. We believe that our mobility assisted contact-based architecture
is the only existing scheme that possesses such desirable features. Furthermore, a rendezvous
mechanism is necessary to provide scalable services in such a network. Our RR architecture is the
only architecture, of which we are aware, that provides such a method without requiring accurate
location information of all nodes. To enable such network to be connected to the Internet the M&M
protocol may be used.
B- Efficient wildfire fighting using ad hoc networks
Wildfires often cover extended uninhabited geographical areas. The spread of such fires
depends on several factors (e.g., wind, fuel and temperature) some of which are affected by the fire
12
itself (e.g., fire-generated winds). These factors, among others (such as slopes, moisture distribution
and canopy), are location-specific. Weather forecasts and wind measures in general provide helpful
information in fighting such fires. However, currently such information is not real-time nor is it
location specific. Ad hoc networks of sensing devices can potentially increase the efficiency of
wildfire fighting by providing real-time information about temperature, moisture, and other factors
mentioned above, along with location information. This information would help predict fire spread,
increase efficiency in containing the fire and potentially save lives and property. The sensing devices
may be deployed rapidly using airplanes, and they self-configure using our contact-based
architectures. The measured information may need to be replicated rapidly for robustness (in case of
device damage due to fire), and the network may be queried for specific data as needed. This data
will help in predicting future fire spread, and hence will aid in effectively allocating fire fighting
resources. To enable efficient operation of such a network we propose to use RRs to provide robust
replication of monitored data in case of damage due to fire. Also, we propose the CARD (non-
mobility) protocol to provide efficient query of the network for the stored/replicated data. In case the
above monitoring network is to be linked to a mobile network of PDAs (e.g., carried by firemen), an
adaptive ( CARD + mobility-assisted) protocol may be used efficiently. The network may also be tied
to a fire fighting management center (possibly through the Internet) using M&M.
2. R ELATIONSHIP T O O THER R ESEARCH & P RACTICE
• Related Work
In wireless ad hoc networks perhaps the simplest form of resource discovery is global flooding.
This scheme does not scale well. It also uses broadcast, which is usually unreliable at the data link
layer (e.g., in 802.11). The synchronization of the broadcasts may lead to severe collisions and
medium contention. Hence, it is our design goal to avoid global flooding. Expanding ring search uses
repeated flooding with incremental TTL. This approach also does not scale well. Much of the work
on routing protocols in ad hoc networks uses some form of flooding or ring search (e.g., DSR [26] , AODV [27] , ODMRP [34] ).
Other approaches in ad hoc networks that address scalability employ hierarchical schemes based
on clusters or landmarks (e.g., LANMAR [46] , SCOUT [45] and [47] ). These architectures,
however, require complex coordination between nodes, and are susceptible to major re-configuration
(e.g., adoption, re-election schemes) due to mobility or failure of the cluster-head or landmark.
Furthermore, usually the cluster-head becomes a bottleneck. Hence, in general we avoid the use of
complex coordination schemes for hierarchy formation, and we avoid using cluster-heads.
In the GLS architecture [28] , all nodes know a grid map of the network. Nodes recruit location
servers to maintain their location. Nodes update their location using an ID-based algorithm. Nodes
looking for location of a specific ID use the same algorithm to reach a location server with updated
information. This is a useful architecture when a node knows the network grid map, knows its own
location, and knows the ID of the node it wishes to contact. Performance of such mechanism
degrades with mobility as the location update messages are increased with every move. Furthermore,
a source node may be looking for a target resource residing at a node with an ID unknown to the
source node, in which case GLS fails.
The algorithm proposed in [48] and [49] uses global information about node locations to establish
short cuts or friends, and uses geographic routing to reach the destination. Knowledge of the
13
locations of all nodes over time under mobility conditions is infeasible. Also, the destination ID (and
location) must be known in advance, which may not be the case in resource discovery.
In ZRP [4] [29] [30] the concept of hybrid routing is used, where table-driven routing is used intra-
zone and on-demand routing is used inter-zone. Border-casting (flooding between borders) is used to
discover inter-zone routes, which may not scale well. A good feature in ZRP is that a zone is node-
specific. Hence, there is no complex coordination susceptible to mobility as in cluster-head
approaches. We use the concept of zone in our architecture. However, we avoid border-casting by using contacts out-of-zone. The main concepts upon which contacts were designed (small world
graphs) are fundamentally different than ZRP’s bordercasting. We have compared the performance
of ZRP and the contact-based approach through simulations. The contacts based approach incurs
significantly lower overhead and has much more desirable scalability characteristics than ZRP.
In [5] an architecture based on intelligent agents is introduced for resource discovery in ad hoc
networks. The concept of domains is used and global cluster-head election (using flooding) is needed
to define a domain. This approach does not scale well in number of nodes in the network due to
repeated global flooding. That architecture was not designed for a mobile network.
Our mobility-assisted contact selection scheme takes advantage of mobility, unlike any other
previous work in this area. In this sense it is quite unique, and we expect that its performance
improve with mobility, whereas all previous works clearly state that their performance degrades with
mobility.
To the best of our knowledge, our rendezvous region scheme ( RRs) described above, is the first
to address explicitly the issues of service bootstrapping (e.g., multicast) in large-scale ad hoc
networks. The idea of rendezvous points ( RPs) was proposed (partially by the PI) for PIM-SM [8]
for multicast in the Internet. However, discovery and management of the RP is fundamentally
different in ad hoc networks. First, a single RP may fail and hence we introduce multiple RPs.
Second, to avoid flooding RP information (as done in PIM-SM) we instead use a mapping from
group prefixes into rendezvous regions. Third, mobility of the RPs may result in their migration from
the respective RRs. This necessitates a dynamic election mechanism that we designed for our RR
architecture.
Comparison of our M&M architecture to other micro-mobility techniques has already been
presented in the main body of the document. To clarify further, M&M achieves handoff performance
comparable to the best existing micro-mobility protocols. Furthermore, the fact that the multicast
maybe sent to multiple (more than two) access routers , allows us to perform efficient reactive
handoff (where a mobile node moves out of coverage and re-connects to a base station). Other
micro-mobility approaches may allow bi -casting (sending to two base stations at a time), but this
requires apriori knowledge of the movement, or connectivity to both base stations. Hence, reactive
handoff cannot be handled well using bi- casting . We argue that allowing existing micro-mobility
approaches to send to more than two base stations requires re-design of those protocols to re-invent
what multicast has already established [8] , in terms of loop duplication and black hole prevention
mechanisms, among others. In addition, M&M naturally supports multicast traffic to the mobile
nodes.
For object tracking, in SCOUT [45] an architecture was presented that is based on hierarchy
formation. Using concepts borrowed from landmark hierarchy [44] , where wireless devices self-
14
configure in a multi-level hierarchy of parent nodes and children nodes. Each level is associated with
a radius to which the device advertises itself. To configure the hierarchy complex mechanisms for
promotion, demotion, and adoption are used. These mechanisms are susceptible to major re-
configurations with mobility. This is mentioned clearly in their work. The root nodes of the hierarchy
use global flooding to send advertisements. These advertisements are sent periodically. If the root
nodes fail or move, new root nodes may be elected, and all nodes in the network may need to re-map
all tracked objects. This does not scale well under dynamic conditions. The work presents two
schemes that may be supported by the hierarchy, called SCOUT-AGG and SCOUT-MAP. In
SCOUT-AGG object information is aggregated up the hierarchy. Queries travel up the hierarchy tree
until the object is found. This query scheme may degenerate to flooding if the object summaries in
many devices in the hierarchy indicate that they may have the object. SCOUT-MAP uses indirection
through a device locator. A hashing scheme is used to route to the device locator (which has
information about the tracking device). The hash depends on the number and identity of children of
each device in the hierarchy that will be involved in this routing. Hence, a change in the number or
identity of children for any of the en-route devices will cause re-hashing. Children for any devices at
any level often change with mobility. A re-hash of the objects and their device locators is needed
with mobility, in addition to re-configuration of the hierarchy. The performance of the SCOUT
architecture degrades drastically with node mobility and network dynamics.
• Complementary Research
Advances in technologies and architectures that enable low-power, low-cost computing and
communications will significantly help the realization of large-scale ad hoc networks that we are
addressing in this research. In addition, research on mobility modeling [18] will provide richer
evaluation and test scenarios for our protocols. Scalable simulation tools (e.g., NS-2 [13]) provide a
powerful virtual environment to study design trade- offs and protocol performance. Research on
small wireless devices that enable monitoring human health and vital signs, or those that enable
monitoring temperature, wind, humidity or slope, will in turn enable applications similar to those we
mention in this document. For example, the digital angel project [33] provides small devices (e.g.,
watches) for medical monitoring. However, they provide limited service (using specific cellular
service provider) to track persons and objects. Coverage, however, is limited to that ISP’s coverage.
Ad hoc networks will potentially increase such coverage drastically; coverage that is crucially needed
in cases of emergency. We plan to leverage digital angel’s sensing and GPS capabilities when
adequate. Also, for fire fighting the smart dust project at UC-Berkeley [32] is building devices that
will enable monitoring forests for wild fires [31] . That project concentrates on the device technology
while our project deals with protocol and architectural design. Hence, the two efforts are
complementary.
Research on security in ad hoc networks (e.g., TESLA [41] , BEBA [42] , and Ariadne [40]) may
prove to be useful in providing security for our protocols and architectures. TESLA is based on
efficient symmetric cryptography, and does not use one-way functions that are expensive to
compute. It can be used to provide secure authentication in point-to-point communication and
broadcast communication. Ariadne is a protocol based on TESLA that provides secure ad hoc
routing. It is a variant of DSR, and hence has similar shortcomings in terms of scaling. We shall
leverage these protocols and investigate ways in which they may be applied to our architecture. One
possible direction to design secure scalable ad hoc networks is to create a small world of trust . The
15
main idea is to establish security relationships with selected nodes (such as the contacts used in our
architecture), then use those relationships to establish further security relationships. For example,
node A establishes a secure relationship with node B using TESLA. If node B already has a secure
relationship with node C , then A can trust C by transitivity. Then the problem of establishing a
secure relationship between two nodes becomes that of finding a chain of secure nodes leading to a
node that both nodes consider as secure. In order to solve this problem efficiently we need to reduce
the degrees of security separation between the nodes. This is achieved by using the small world
concept using secure contacts, similar to what we described above. One possible problem is that
establishing a secure relationship (using TESLA) requires synchronization between nodes, which is
hard to achieve with distant nodes, especially if not all nodes are GPS capable. To overcome this
problem, nodes start by establishing secure relationships with their direct neighbors, with which they
may establish synchronization easily [43]. As those secure neighbors move away, they in turn
establish secure relationships with new neighbors, and the small world of trust gets constructed
(similar to the way we establish a small world of contacts above).
References
[1] A. Helmy, “ Small Large-Large Scale Wireless Networks: Mobility-Assisted Resource Discovery”, USC-TR, July 2002. (Submitted for Review)
[http://ceng.usc.edu/~helmy]
[2] N. Nahata, P. Pamu, S. Garg, A. Helmy, "Efficient Resource Discovery for Large Scale Ad hoc Networks using Contacts", ACM SIGCOMM
Conference (Refereed poster), August 2002. [http://ceng.usc.edu/~helmy]
[3] S. Garg, P. Pamu, N. Nahata, A. Helmy, "Contact Based Architecture for Resource Discovery (CARD) in Large Scale MANets", USC-TR, July
2002. (Submitted for Review). [http://ceng.usc.edu/~helmy]
[4] M. Pearlman, Z. Haas, “Determining the optimal configuration for the zone routing protocol”, IEEE JSAC, p. 1395-1414, vol. 17, 8, Aug 1999.
[5] J. Liu, Q. Zhang, W. Zhu, J. Zhang, B. Li, “A Novel Framework for QoS-Aware Resource Discovery in Mobile Ad Hoc Networks”, IEEE ICC ’02.
[6] A. Helmy, "Architectural Framework for Large-Scale Multicast in Mobile Ad Hoc Networks", IEEE ICC ’02. [http://ceng.usc.edu/~helmy]
[7] A. Helmy, "A Multicast-based Protocol for IP Mobility Support", ACM Second International ACM SIGCOMM Workshop on Networked Group
Communication (NGC 2000), Palo Alto, November 2000. [http://ceng.usc.edu/~helmy]
[8] D. Estrin, D. Farinacci, A. Helmy, D. Thaler, S. Deering, V. Jacobson, M. Handley, C. Liu, P. Sharma, “Protocol Independent Multicast – Sparse
Mode (PIM-SM): Protocol Specification”, RFC 2362/2117 of IETF/IDMR , March '97/'98. [http://ceng.usc.edu/~helmy]
[9] C. Perkins, “IP Mobility Support” , RFC 2002, Internet Engineering Task Force , October 1996.
[10] C. Perkins, D. Johnson, “Route Optimization in Mobile IP”, Internet Draft, Internet Engineering Task Force, February 2000.
[11] C. Perkins and D. Johnson, “Mobility Support in IPv6” , Proceedings of MobiCom’96 , November 1996.
[12] D. Johnson, C. Perkins, “Mobility Support in IPv6”, Internet Draft, Internet Engineering Task Force, March 2000.
[13] L. Breslau, D. Estrin, K. Fall, S. Floyd, J. Heidemann, A. Helmy, P. Huang, S. McCanne, K. Varadhan, Y. Xu, H. Yu, "Advances in Network
Simulation", IEEE Computer, vol. 33, No. 5, p. 59-67, May 2000. [http://ceng.usc.edu/~helmy]
[14] A. T. Campbell and J. Gomez IP Micro-Mobility Protocols, ACM SIGMOBILE Mobile Computer and Communication Review (MC2R), 2001
[15] E. Gustafsson, A. Jonsson, C. Perkins, Mobile IP Regional Registration, Internet-draft, draft-ietf-mobileip-reg-tunnel-02, March 2000.
[16] A. Campbell, J. Gomez, S. Kim, A. Valko, C. Wan, Z. Turanyi, “Design, implementation, and evaluation of cellular IP” IEEE Personal
Communications , Volume: 7 Issue: 4 , Page(s): 42 –49, Aug. 2000.
[17] R. Ramjee, T. La Porta, L. Salgarelli, S. Thuel, K. Varadhan, L. Li, “IP-based access network infrastructure for next-generation wireless data
networks”, IEEE Personal Communications, Volume: 7 Issue: 4, Page(s): 34 –41, Aug. 2000.
[18] F. Bai, N. Sadagopan, A. Helmy, "IMPORTANT: A framework to systematically analyze the Impact of Mobility on Performance of RouTing
protocols for Adhoc NeTworks", USC-CS-TR-02-765 , July 2002. (Submitted for Review). [http://ceng.usc.edu/~helmy]
[19] A. Helmy, “State Analysis and Aggregation Study for Multicast-based Micro Mobility”, IEEE International Conference on Communications (ICC),
May 2002. [http://ceng.usc.edu/~helmy]
[20] A. Helmy, and M. Jaseemuddin, "Efficient Micro-Mobility using Intra-domain Multicast-based Mechanisms (M&M)", USC-CS-TR-01-747 , Aug
2001. [http://ceng.usc.edu/~helmy]
[21] A. Helmy, M. Jaseemuddin, Ganesha Bhaskara, "Efficient Micro-Mobility using Intra-domain Multicast-based Mechanisms (M&M)", July 2002.
(Submitted for Review). [http://ceng.usc.edu/~helmy]
[22] Burton Bloom , “Space/time trade- offs in hash coding with allowable errors ”, CACM , 13(7) :422-426, July 1970.
[23] M. Mitzenmacher, “ Compressed Bloom Filters ”, PODC, 2001.
[24] J. Byers, J. Considine, M. Mitzenmacher, S. Rost, “Informed Content Delivery Across Adaptive Overlay Networks”, ACM SIGCOMM , Aug. 2002.
[25] C.E. Perkins and P. Bhagwat, "Highly Dynamic Destination-Sequenced Distance-Vector Routing (DSDV) for Mobile Computers", Comp. Comm.
Rev., Oct. 1994, pp.234-244.
[26] David B. Johnson, Davis A. Maltz, "The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks" October 1999 IETF Draft.
[27] Charles E. Perkins, Elizabeth M. Royer, Samir R. Das, "Ad Hoc On-demand Distance Vector Routing", October 99 IETF Draft.
[28] J. Li, J. Jannotti, D. Couto, D. Karger, R. Morris, "A Scalable Location Service for Geographic Ad Hoc Routing (GLS/Grid)", ACM Mobicom 2000.
[29] Z. Haas, M. Pearlman, "The Zone Routing Protocol (ZRP) for Ad Hoc Networks", IETF Internet draft for the Manet group, June '99.
[30] Z. Haas, M. Pearlman, "The Performance of Query Control Schemes for the Zone Routing Protocol", ACM SIGCOMM '98.
[31] P. Eng, “Tiny Fire Marshals”, abcnews, Sept. 13, 02. http://abcnews.go.com/sections/scitech/CuttingEdge/cuttingedge020913.html.
[32] Smart dust project, UC-Berkeley. http://robotics.eecs.berkeley.edu/~pister/SmartDust/
[33] Digital Angel Corporation, http://www.digitalangel.net.
16
[34] S. Lee, M. Gerla, C. Chiang, "On-demand multicast routing protocol", IEEE WCNC, p. 1298-1302, vol. 3, 1999.
[35] Y. Ko, N. Vaidya, "Location-aided routing (LAR) in mobile ad hoc networks", Wireless Networks 6, 4, p. 307-321, July 2000.
[36] B. Karp, H. Kung, "GPSR: Greedy Perimeter Stateless Routing for Wireless Networks", MobiCom 2000.
[37] S. Ratnaswamy, P. Francis, M. Handley, R. Karp, and S. Shenker. A scalable content- addressable network. ACM SIGCOMM, 2001.
[38] I. Stoica, R. Morris, D. Karger, F. Kaashoek, and H. Balakrishnan. Chord: A peer-to-peer lookup service for internet applications. ACM
SIGCOMM, 2001.
[39] I. Clarke, O. Sandberg, B. Wiley, and T.W. Hong. Freenet: A distributed anonymous information storage and retrieval system in
designing privacy enhancing technologies. Int’l Workshop on Design Issues in Anonymity and Unobservability, LNCS 2009, 2001.
[40] Y. Hu, A. Perrig, D. Johnson, “ Ariadne: A Secure On-Demand Routing Protocol for Ad Hoc Networks”, Mobicom 2001.
[41] Adrian Perrig, Ran Canetti, Dawn Song, and J. D. Tygar. Efficient and Secure Source Authentication for Multicast. In Network and Distributed
System Security Symposium, NDSS ’0 1, pages 35–46, February 2001.
[42] Adrian Perrig, Ran Canetti, J.D. Tygar, and Dawn Song. Efficient Authentication and Signing of Multicast Streams over Lossy Channels. In IEEE
Symposium onSecurity and Privac y, pages 56–73, May 2000.
[43] J. Elson, L. Girod, D. Estrin, “ Fine-Grained Network Time Synchronization using Reference Broadcasts”, Fifth Symposium on Operating Systems
Design and Implementation (OSDI) 2002.
[44] P. F. Tsuchiya, "The Landmark Hierarchy: A new hierarchy for routing in very large networks", CCR, Vol. 18, no. 4, pp. 35-42, Aug. 1988.
[45] S. Kumar, C. Alaettinoglu, D. Estrin, “SCOUT: Scalable object tracking through unattended techniques”, ICNP 2000.
[46] P. Guangyu, M. Gerla, X. Hong, “LANMAR: landmark routing for large scale wireless ad hoc networks with group mobility”, MobiHOC ’00, p. 11-
18, 2000.
[47] B. Das, V. Bharaghavan, “Routing in ad-hoc networks using minimum connected dominating sets”, in Proc. IEEE ICC ’97.
[48] L. Blazevic, S. Giordano, J .-Y. Le Boudec “ Anchored Path Discovery in Terminode Routing”. Proceedings of the Second IFIP-TC6 Networking
Conference (Networking 2002), Pisa, May 2002.
[49] J. Kleinberg, “Navigating in a small world”, Nature, 406, Aug. 2000.
[50] A. O’Neill, G. Tsirtsis, S. Corson, Edge Mobility Architecture, Internet-draft, draft-oneill-ema-02.txt, July 2000.

Asset Metadata

Creator Helmy, Ahmed (author)

Core Title USC Computer Science Technical Reports, no. 804 (2003)

Alternative Title Efficient provisioning for services in large-scale wireless-wired networks (title)

Publisher Department of Computer Science,USC Viterbi School of Engineering, University of Southern California, 3650 McClintock Avenue, Los Angeles, California, 90089, USA (publisher)

Tag OAI-PMH Harvest

Format 16 pages (extent), technical reports (aat)

Language English

Unique identifier UC16270326

Identifier 03-804 Efficient Provisioning for Services In Large-Scale Wireless-Wired Networks (filename)

Legacy Identifier usc-cstr-03-804

Format 16 pages (extent),technical reports (aat)

Rights Department of Computer Science (University of Southern California) and the author(s).

Internet Media Type application/pdf

Source 20180426-rozan-cstechreports-shoaf (batch), Computer Science Technical Report Archive (collection), University of Southern California. Department of Computer Science. Technical Reports (series)

Access Conditions The author(s) retain rights to their work according to U.S. copyright law. Electronic access is being provided by the USC Libraries, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name USC Viterbi School of Engineering Department of Computer Science

Repository Location Department of Computer Science. USC Viterbi School of Engineering. Los Angeles\, CA\, 90089

Repository Email csdept@usc.edu

Inherited Values

Title Computer Science Technical Report Archive

Description Archive of computer science technical reports published by the USC Department of Computer Science from 1991 - 2017.

Coverage Temporal 1991/2017

Repository Email csdept@usc.edu

Repository Name USC Viterbi School of Engineering Department of Computer Science

Repository Location Department of Computer Science. USC Viterbi School of Engineering. Los Angeles\, CA\, 90089

Publisher Department of Computer Science,USC Viterbi School of Engineering, University of Southern California, 3650 McClintock Avenue, Los Angeles, California, 90089, USA (publisher)

Linked assets

Computer Science Technical Report Archive