WO2002073889A1 - Communications network - Google Patents
Communications network Download PDFInfo
- Publication number
- WO2002073889A1 WO2002073889A1 PCT/GB2002/000903 GB0200903W WO02073889A1 WO 2002073889 A1 WO2002073889 A1 WO 2002073889A1 GB 0200903 W GB0200903 W GB 0200903W WO 02073889 A1 WO02073889 A1 WO 02073889A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- node
- service
- network
- communications network
- nodes
- Prior art date
Links
- 238000004891 communication Methods 0.000 title claims description 21
- 238000000034 method Methods 0.000 claims abstract description 33
- 230000008569 process Effects 0.000 claims abstract description 26
- 230000000694 effects Effects 0.000 claims description 44
- 230000008676 import Effects 0.000 claims description 3
- 241000894006 Bacteria Species 0.000 abstract description 22
- 238000012545 processing Methods 0.000 abstract description 17
- 238000004088 simulation Methods 0.000 description 18
- 239000013612 plasmid Substances 0.000 description 14
- 230000004044 response Effects 0.000 description 12
- 230000035772 mutation Effects 0.000 description 11
- 230000010415 tropism Effects 0.000 description 11
- 230000001580 bacterial effect Effects 0.000 description 10
- 230000005012 migration Effects 0.000 description 9
- 238000013508 migration Methods 0.000 description 9
- 108090000623 proteins and genes Proteins 0.000 description 9
- 238000013459 approach Methods 0.000 description 8
- 230000007246 mechanism Effects 0.000 description 8
- 230000008901 benefit Effects 0.000 description 7
- 230000006978 adaptation Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 235000015097 nutrients Nutrition 0.000 description 6
- 230000003044 adaptive effect Effects 0.000 description 5
- 230000006399 behavior Effects 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 230000007613 environmental effect Effects 0.000 description 5
- 230000002068 genetic effect Effects 0.000 description 5
- 230000033458 reproduction Effects 0.000 description 4
- 241000251730 Chondrichthyes Species 0.000 description 3
- 230000004993 binary fission Effects 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 235000013305 food Nutrition 0.000 description 3
- 102000004169 proteins and genes Human genes 0.000 description 3
- 230000004083 survival effect Effects 0.000 description 3
- 210000004027 cell Anatomy 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000003278 mimic effect Effects 0.000 description 2
- 230000010076 replication Effects 0.000 description 2
- 239000011435 rock Substances 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 241000894007 species Species 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 108020004638 Circular DNA Proteins 0.000 description 1
- 235000008694 Humulus lupulus Nutrition 0.000 description 1
- 241000257303 Hymenoptera Species 0.000 description 1
- 241000256602 Isoptera Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000011681 asexual reproduction Effects 0.000 description 1
- 238000013465 asexual reproduction Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010367 cloning Methods 0.000 description 1
- 238000004138 cluster model Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000006353 environmental stress Effects 0.000 description 1
- 230000027555 hydrotropism Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000012528 membrane Substances 0.000 description 1
- 230000002503 metabolic effect Effects 0.000 description 1
- 230000000813 microbial effect Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 210000003463 organelle Anatomy 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 230000029553 photosynthesis Effects 0.000 description 1
- 238000010672 photosynthesis Methods 0.000 description 1
- 230000027870 phototropism Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000025508 response to water Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000035882 stress Effects 0.000 description 1
- 230000002459 sustained effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000035899 viability Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
Definitions
- This invention relates to the field of network management and in particular the management of complex data communication networks.
- Communication networks such as the Internet are probably the most complex machines built by civilization. The number of possible failure states in a major network is so large that even counting them is infeasible. Deciding the state that the network is in at any time with great accuracy is therefore not possible.
- data networks such as the Internet are subjected to a mixture of deterministic and stochastic load. The network's response to this type of traffic is chaotic, and thus the variation of network state is highly divergent and accurate predictions of network performance require knowledge of the current network state that is more accurate than can be obtained. " Future networks, which will ' have increased intelligence, will be even more complex and have less tractable management.
- a network management paradigm is required that can maintain network performance in the face of fractal demands without detailed knowledge of the state of the network, and can meet unanticipated future demands.
- Biologically inspired algorithms for example genetic algorithms and neural networks
- Stromatolites (S.Golubic "Stromatolites of Shark Bay” pp1 03-1 49 in Environmental evolution ed. Margulis and Olendzenski, MIT press 1 999) are rock mounds built by heterogeneous colonies of bacteria. Colonies of this type can be found today in Shark Bay, on the coast of Western Australia. Fossil colonies can also be found throughout earth history - currently the oldest known was found in rocks dated at 3.5Ga, also in Western Australia. The colonies exhibit sophisticated internal organization, that emerges from interactions between individual colony members. In comparison with higher organisms such as ants and termites, the interactions are relatively simple. It is known (see IW Marshall and C Roadknight, "Adaptive management of an active services network", British Telecom. Technol.
- a multi-service communications network comprising a plurality of clusters, each cluster comprising more than one node and being connected to one or more other clusters, each node comprising one or more nodal policy and being configured to process one or more services in accordance with said one or more nodal policy, each nodal policy comprising;
- a service request identifier said service request identifier determining the type of service request that may be processed by each respective node; and (ii) one or more service request criteria, said service request criteria determining whether a type of service request defined by said service request identifier will be processed by each respective node characterised in that, in use, un-processed service requests are forwarded to other nodes within a cluster and un-processed service requests which reach the perimeter of the cluster are forwarded to one of the connected clusters.
- a service request contains a nodal policy. If the activity indicator reaches an upper threshold then the node may replicate all of the nodal policies to generate a clone of the node.
- the node may import the policy contained in the service request and additionally if the activity indicator reaches a second lower threshold then the node may delete an enabled nodal policy and enable a dormant nodal policy. Alternatively, if the activity indicator reaches a second lower threshold then the node may exchange position with an adjacent node. Service requests may be injected into the uppermost layer of the cluster.
- Each node may further comprise a variable that encodes a preference for existing in a defined region of a cluster.
- the multi-service communications network may be implemented as described above using a data carrier comprising computer code means.
- Figure 1 shows a schematic depiction of a multi-service communications network according to the present invention
- Figure 2 shows a schematic depiction of the response of a multi-service communications network according to the present invention to different levels of network traffic
- Figure 3 shows a graphical depiction of the response of a multi-service communications network according to the present invention to the introduction of new service types to the network.
- Figure 4 is a graphical depiction of the simulation of a network cluster according to the present invention.
- Figure 5 is a graphical depiction comparing the performance of a network according to the present invention with a conventional network.
- FIG 1 shows a multi-service communications network 10 that comprises a number of inter-connected nodes 20. These nodes are capable of performing one or more of a number of processes which are needed to support the multi-service network 1 0, for example an ATM switch, WWW server, IP router, SMTP mail server, domain name server (DNS), etc.
- the users of the multiservice network 10 are divided into a number of communities 30. These communities are geographically dispersed and have different requirements from the multi-service network, in terms of the type of request and the number of requests made (in Figure 1 , the different types of requests are indicated by the letters a, b, c & d, with each letter representing a request for a different service and the frequency of each letter representing the number of requests made for that service).
- the number of communities 30 using the network 1 0 will vary and the nature and volume of network usage of each community will vary in a self-similar, deterministic way. Also, the number of services provided by the network will also vary over time as new services are introduced and some services become obsolete and are removed from the network.
- One approach to managing such a network would be to make all nodes capable of processing all types of requests and with each node having sufficient capacity to be able to process all of the requests received at that node.
- this would lead to a very inefficient use of resources as virtually every node would be over-dimensioned and have an excess of capacity, both in terms of the type of service requests and the number of service requests that each node would be able to process.
- Another approach would be to limit each node to processing a fixed subset of the different services supported by the network and providing a fixed capacity of processing capability for each service type at each node.
- Lamarkian was an 1 8th century French scientist who argued that evolution occurs because organisms can inherit traits, which have been acquired by their ancestors during their lifetime
- plasmids individuals capable of exchanging elements of their genetic material, known as plasmids, during their lifetime in response to environmental stress, and passing the acquired characteristics to their descendants.
- Plasmid migration allows much quicker reaction to sudden changes in influential environmental factors and can be modelled as a learning mechanism. Sustaining fitness in a complex changing fitness landscape has been shown to require evolution together with a fast learning mechanism.
- each node posses a set of policies, which mimic the functionality of a bacterial plasmid.
- the set of policies specifies which type of service request the node is able to process (e.g. a, b, c or d from Figure 1 ), and a rule, or rules, that determine whether the node will accept a service request or not.
- Each node has a certain number of policies and these policies determine how the node responds to the changing environment of the network (in the same way as the genetic material of a bacterium determines how that bacterium responds to its environment).
- the policies take the form ⁇ x,y,z ⁇ where: x is a function representing the type of service requested; y is a function which determines whether a service request is accepted dependent upon the number of service requests queued at the node; and z is a function which determines whether a service request is accepted dependent upon the activity level of the node.
- the 'value' that a node may derive from processing a service request would be receiving revenue from a user or network or service provider (this is analogous to a bacterium gaining energy from metabolising resources, such that the bacterium can survive and potentially reproduce) .
- the quantum of revenue will depend upon the type of service request which is processed by the node, with some service requests being more important and hence providing greater reward when they are processed.
- Each node may have any number of policies.
- Enabled nodes i.e. a node that is processing service requests
- Dormant policies may be re-enabled and enabled policies may be repressed, becoming dormant.
- User requests for service are received by the node(s) nearest the point of entry to the network from the user community generating the service request. If the node is capable of processing the request then the request joins a queue, with each node evaluating the items that arrive in its input queue on a 'first in, first out' principle. If the service request at the front of the queue matches an available policy the service request is processed, the node is 'rewarded' (i.e. revenue is generated for the network or service operator) and the request is then deleted from the queue. If the service request does not match any of the node's enabled policies then the request is forwarded to another node and no reward is given.
- Plasmid migration involves genes from healthy individuals being shed or replicated into the environment and subsequently being absorbed into the genetic material of less healthy individuals. If plasmid migration does not help weak strains increase their fitness they eventually die. Thus, if a node has a queue length or an activity indicator that reaches an upper threshold value then one of the node's policies may be made accessible to other nodes.
- a node may acquire one of these policies that has been donated by successful node (these threshold values (both upper and lower) need not be the same for the queue length as for the activity indicator) .
- a node If a node maintains the upper threshold value for the queue length or for the activity indicator for a given period of time (i.e. the node sustains its success) then the node can clone itself by producing another node having the same set of policies as the parent node. This is analogous to healthy bacterium with a plentiful food supply reproducing by binary fission to produce identical offspring.
- this cloning process may be initiated by the node's queue length or activity indicator reaching a second upper threshold value, this second upper threshold value being greater than the first upper threshold value.
- a node maintains the lower threshold value for the queue length or for the activity indicator for a given period of time (i.e. the idleness of the node is sustained) then some or all of the enabled policies of the node are deleted and any dormant policies are activated. If the node has no dormant polices then once all the enabled policies have been deleted then the node is switched off. This is analogous to bacterial death due to nutrient deprivation.
- nodal policies may be varied.
- policy mutation may involve the random alteration of a single value in a policy. If a policy were to have the form:
- Accept request for service c if activity indicator ⁇ 80%, or Accept request for service a if activity indicator ⁇ 60%, or
- the inventors have implemented a simulation of a network according to the present invention (and of the kind described above).
- the simulated network comprises six inter-connected clusters 21 , 22, 23, 24, 25 & 26 (see Figure 2) .
- Each cluster comprises 500 vertices at which a node may be situated, with 1 0 rows each containing 50 vertices.
- Each cluster is able to forward service requests to two of the other clusters and to also receive service requests from two other clusters.
- cluster 21 can forward requests to, and receive requests from, clusters 22 and 24.
- the simulation system supports up to one hundred different service types but in the interest of simplicity the following discussion will refer to a subset of four of these service types; A, B, C, D.
- the system is initialised by populating a random selection of vertices with enabled nodes. These initial nodes have a random selection of policies which define how each node will handle requests for service. These policies have a number of variables and are represented by the form ⁇ x,y,z ⁇ where: x is a function representing the type of service requested; y is a function between 0 and 200 which is interpreted as the value in a statement of the form
- a node having a number of policies is represented as ⁇ x ⁇ ,y ⁇ ,z ⁇ : ⁇ 2 ,y 2 ,Z2 : .. : xi,y ⁇ ,zi ⁇
- Requests are input to the system by injecting sequences of characters, which represent service requests, at each vertex in the array. If the vertex is populated by a node, the items join a node input queue. If there is no node at the vertex then the requests are forwarded to a neighbouring vertex. Each node evaluates the service requests that arrive in its input queue on a 'first in, first out' principle.
- each node may only process four requests per measurement period (henceforth referred to as an epoch). The more time a node spends processing requests, the busier it is seen to be and the greater the value of its activity indicator.
- the activity indicator can be determined by calculating the activity in the current epoch, for example, if the node processed three requests in the current epoch, generating 25 'points' of reward for each processed request, then the activity indicator would be 75. However in order to dampen any sudden changes in behaviour due to a highly dynamic environment it is preferred to combine the activity indicator at the previous epoch with the activity indicator for the current epoch. It has been found for the simulated network that a suitable ratio for the previous indicator to the current indicator is 0.8:0.2. For example, if in this epoch the node has processed three requests with each generating 25 points, and the node had an activity indicator of 65 for the previous epoch then the activity indicator for the present epoch will be 67.
- the inventors have implemented a new mechanism for plasmid migration by extending the definition of the service request to include . the functionality whereby a service request can carry a policy (in a manner analogous to an email message carrying an attachment) . If a node is not able to process a service request (and so must forward it) and a random number between 0-100 is less than it's activity indicator then a copy of a policy from that node is attached to the request. Alternatively, if a node is not able to process a service request and its activity indicator plus a constant (typically 5 %) is less than a random number between 0-1 00 then the node may remove take a policy from the policy attachment space, if there is one. If a node successfully processes a service request and its activity indicator plus a constant (currently 5 %)- is less than a random number between 0-100 then the node may also remove a policy from the policy attachment space if it holds a policy.
- a service request can carry a policy
- values for queue length and the time-averaged activity indicator are used as the basis for determining whether policies are added or discarded or if replication occurs, and evaluation is performed every five epochs.
- Both of the threshold values were 30 in this example and it is clear that this value is only an example, other threshold values may be selected and there is no need to have the threshold value for the queue length to be equal to that of the activity indicator.
- the threshold values will determine the number of nodes which can reproduce to occupy the system and should be chosen suitably to match the performance requirements of the system. If the node continues to exceed the threshold for a number evaluation periods (i.e.
- any dormant policies exist these are enabled, but if there are no dormant policies then the node is switched off. This is analogous to bacterial death by nutrient deprivation. If an imported policy does not increase the activity indicator of the node such that the node is no longer idle then a further policy may be imported from a service request having an attached policy; if there is no such policy to be imported then the node may be deleted.
- a visualisation environment was created for this implemented simulation.
- the visualisation environment provides an interface where network load and other parameters can be varied in many ways, thereby allowing stresses to be introduced into the system in a flexible manner. For instance, the ratio of requests for the services can be made to vary over time, as can the overall number of requests per unit time.
- Rules governing reproduction and evolution, including plasmid migration (as described above) were introduced into the simulation, in an attempt to force the nodes to model the behaviour of bacterium in a changing environment. For the simulations discussed below the four different policies had time to live (TTL) values of 2, 5, 10 and 20 hops respectively and were injected into the clusters in equal proportion.
- TTL time to live
- the injection rate was approximately 25 % of the maximum allowed by the simulation model, leading to approximately 50% of the vertices in the clusters to be populated at any given time.
- the clusters are arranged in a rectangular array of 50 x 10 vertices with all requests for services arriving at the nodes along one of the longer edges of the cluster, with requests arriving with equal probability at the 50 nodes in the uppermost row of the cluster (it will be understood that the orientation of the rectangular cluster and the arrival of the requests at the uppermost layer of the cluster are arbitrary.
- the service requests could arrive at the lowest layer of the cluster and travel upwards through the cluster, or the cluster could be rotated through 90 ⁇ with the requests arriving at either of the longer edges. Equally, the requests could arrive at one the shorter edges of the cluster.
- the size and proportion of the cluster are merely exemplary and are not critical to the working of this embodiment of the invention) .
- Figure 3 shows the probabilities for a service request being forwarded from node X. The choice of these probabilities is designed to promote the migration of forwarded requests down through a cluster thus, it " will be seen that the general tendency of service requests will be to migrate down through the array before they arrive at a node that is capable of processing the service request.
- the request is forwarded to the uppermost layer of one of the clusters to which its original cluster may forward requests. If there is more than one cluster to which a request may be forwarded then one may be chosen randomly, or on a turn-wise basis or some cluster parameter may be used as the basis for a decision.
- the service request is injected into a column that preferably is chosen randomly, although it may be injected into the column corresponding to the column from which it left its originating cluster (additionally, if un-processed service requests are forwarded beyond the width of a cluster then the request may be forwarded to a different cluster) .
- Occupying a position in the uppermost layers of the array will provide an advantage to all nodes, but some nodes will be more advantaged than others by this. It is obvious that nodes that process predominantly short TTL requests will be better positioned at or near to the supply of service requests to the network, but nodes processing predominantly long TTL requests may also benefit by being near the point of request origin.
- a request having a long TTL may be processed equally well after passing through several layers of the network as if it had been processed at the edge of the network, but although a node may be thriving on the ninth layer of the array if it produces clones that populate the eight layer of the array then the clones will start taking some of the parent's potential requests, thereby decreasing the viability of the parent node.
- a penalty was introduced to a node for allowing a request to 'die' in its queue. If a node accepts a request onto its queue that has exceeded its TTL by the time the node attempts to process it, the node receives a penalty, for example a reduction of the activity indicator by 4 points.
- a more indirect penalty was created by penalising nodes that passed on requests with short times to live after realising they do not contain a policy to process the request. It is very anti-social to forward a near 'dead' request to the bottom of a neighbouring node's pile of requests (given the aforementioned timeout penalty) so nodes are penalised for forwarding a request based on the remaining TTL of that request. This should put evolutionary pressure on nodes to restrict the maximum length of their queues, as there is no penalty for not accepting a request. This penalty function has an effect on the optimal position of a node.
- a node processes predominantly requests having a long TTL, occupying a prominent position in the network would mean it would have to forward all requests having a short TTL and incurring penalties for this. It is advantageous for these nodes to reside where the short TTL requests have already been processed, i.e. in the lower layers of the array. Although these penalties have been implemented by directly reducing the activity indictor of the node, it is also possible to penalise nodes by reducing the time window that they have to process requests
- the simulation also includes an additional mechanism that is an analogy of tropism, which is the response of plants and lower order animal forms to a stimulus that has a greater intensity from one direction than another.
- tropism is the response of plants and lower order animal forms to a stimulus that has a greater intensity from one direction than another.
- most plant seeds exhibit phototropism (a response to light) by sending shoots to maximise the effects of photosynthesis and hydrotropism (a response to water) by sending root systems to maximise water uptake from their environment.
- Each node has a tropism value encoded within its genome, which can be within the range 0 to 1 , and which can effect the manner in which the node reproduces by binary fission (note that the tropism value is not transferred during plasmid migration).
- the tropism value of a node may be altered by a random mutation.
- a node has a tropism value of greater than 0.5, its offspring will spawn directly to the left (at a 25 % probability), directly to the right (at a 25 % probability) or directly above the parent node (at a 50% probability) .
- the node's tropism value is 0.5 or less then its offspring will spawn directly to the left (at a 25% probability) , directly to the right (at a 25 % probability) or directly below the parent node (at a 50% probability). This allows some strains of nodes to evolve a preference for living in the upper levels of the array and other strains to prefer a life in the lower levels of the array.
- Those nodes that have a low tropism value are likely to process predominantly long TTL service requests due to their preference for the lower levels of the array, whilst those nodes having a high tropism value are likely to possess policies enabling them to process predominantly short TTL service requests due to their preference for the upper levels of the array.
- a node that is about to 'die' may sometimes change position with the node directly above or directly below them, allowing a node to move in a preferred direction where it may find more suitable conditions even if it's path is blocked by another node.
- the direction of the move is dependent upon the tropism value of the under-performing node with low tropism value nodes moving down the array whilst high tropism value nodes will move up the array. Simulation has shown that a suitable probability for allowing such a positional switch to occur is 1 0% .
- Figure 4 shows some simulation results for a cluster from a network according to the present invention. Those nodes without a numerical indicator are of a general nature whilst those with a numerical indicator have adapted to handle service requests with particular TTL values with the indictor increasing in value as the TTL limit increases.
- Figure 4a is a snapshot of the simulation after it has been run for 1 5 epochs and the random nature of the distribution of nodes is very much apparent.
- Figures 4b and 4c are snapshots taken after a longer period of simulation (200 and 800 epochs respectively) and they show the layering that occurs, with the type 1 and type 2 nodes predominating towards the top of the cluster and the type 3 and type 4 nodes predominating towards the bottom of the cluster.
- FIGS 5a 6k 5b show the results of a comparison of the performance of an adaptive network according to the present invention and a fixed network (in which each cluster is randomly populated with nodes which contain a fixed number of policies, which do not evolve, are not able to mimic plasmid exchange and can not replicate) .
- Figure 5a shows the consumption efficiency of the colony against simulated time. As might be expected the properties of the random colony do not change with time. For the network according to the present invention, two timescales of evolution are apparent.
- FIG. 5b shows the effect of progressively concentrating nutrient input into fewer sub-colonies.
- the x-axis has an even distribution of input (at 80% max as in Figure 5a) into the top edge of all six clusters. The input is concentrated until the supply to one cluster is the max possible, and the other clusters are fed so that the total number of service requests remains constant. It is clear that at all times the evolved network is more effective than the random network.
- This final graph is particularly interesting since the network is behaving in exactly the way that a good distributed load balancing algorithm should. It is spreading load across available resources, ensuring a high probability of success (comparable to real computing systems at high load), whilst minimizing the resources required.
- control parameters are merely exemplary and are provided to illustrate the present invention.
- the optimal values of the different parameters will vary from system to system depending upon the number and type of service requests that are to be processed by the network.
- Networks according to the present invention have the advantage that they can be upgraded to provide additional services by inserting new policies defining these services into the network. The simplest method of achieving this is to allow nodal policies to mutate so that a policy can process the new type of service request(s).
- a policy mutates such that it can process a new service request type then the node will have an abundant food source, which will aid the distribution of the newly mutated policy through the nodal population, by plasmid migration and/or by reproduction.
- Another method of enabling nodes to process new types of service request is to modify the policy set of an existing node so that it can handle the new service type. In order to do this it will be necessary to disable the node for a short period of time, but the distributed nature of the network means that other nodes will react to accommodate the missing node. However, it would be easy to introduce a new policy type by attaching it to one or more service requests as they are injected into the clusters.
- QoS quality of service
- the users may choose a suitable level of QoS, or a cost level that they are prepared to pay, with this choice then being translated into a suitable combination of service request value and/or time to live.
- the QoS process may be more transparent, with users selecting directly the service request value and/or time to live for some or all of their network transactions.
- ANNs large scale active network nodes
- DPS dynamic proxylet server
- JVMs Java virtual machines
- Each DPS controls and implements the algorithmic rules discussed above that determine the activity indicator levels at which nodes may reproduce, export nodal policies, import nodal policies, etc.
- the JVMs are the nodes (i.e. analogous to bacterium) with the proxylets (or policies pointing to the proxylets and authorising execution) providing the nodal policies (i.e. the genes of the bacteria).
- the proxylets are mostly multi-user devices, but they may be installed and used by a single user.
- a proxylet may be installed by a user who then proceeds to load the proxylet by making suitable service demands (this is analogous ' to placing a nodal policy into a node and then injecting some requests for the service it represents). Only at very low traffic loads will a given proxylet not be present at all the ANNs, but this gives rise to efficient network resource utilisation as proxylets are run only in response to user demand and are run at a convenient network location.
- the nodes of the simulated network do not have any awareness of the concept of an ANN or the boundaries between the different ANNs that comprise a network. The main reasons for this is to minimise the complexity of the nodal operations and because of the fuzzy nature of the boundaries in a cluster model.
- the simulation indicates that such an active network should be capable of implementation and that reasonable levels of performance achieved.
- the ability to manage the deployment of new services over such networks has also been successfully simulated.
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP02703711A EP1368932A1 (en) | 2001-03-13 | 2002-03-01 | Communications network |
US10/468,283 US20040071147A1 (en) | 2001-03-13 | 2002-03-01 | Communications network |
CA002438540A CA2438540A1 (en) | 2001-03-13 | 2002-03-01 | Communications network |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP01302291 | 2001-03-13 | ||
EP01302291.8 | 2001-03-13 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2002073889A1 true WO2002073889A1 (en) | 2002-09-19 |
Family
ID=8181783
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/GB2002/000903 WO2002073889A1 (en) | 2001-03-13 | 2002-03-01 | Communications network |
Country Status (4)
Country | Link |
---|---|
US (1) | US20040071147A1 (en) |
EP (1) | EP1368932A1 (en) |
CA (1) | CA2438540A1 (en) |
WO (1) | WO2002073889A1 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007052254A1 (en) * | 2005-11-01 | 2007-05-10 | Eci Telecom Ltd. | Access system for the provisioning of different communication services, and method for using same |
Families Citing this family (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7478162B2 (en) * | 2000-02-08 | 2009-01-13 | British Telecommunications Public Limited Company | Communications network |
GB0022561D0 (en) * | 2000-09-14 | 2000-11-01 | British Telecomm | Communications network |
US7703018B2 (en) * | 2003-05-22 | 2010-04-20 | International Business Machines Corporation | Apparatus and method for automating the diagramming of virtual local area networks |
US8782654B2 (en) | 2004-03-13 | 2014-07-15 | Adaptive Computing Enterprises, Inc. | Co-allocating a reservation spanning different compute resources types |
US20070266388A1 (en) | 2004-06-18 | 2007-11-15 | Cluster Resources, Inc. | System and method for providing advanced reservations in a compute environment |
CA2571654A1 (en) * | 2004-06-22 | 2005-12-29 | British Telecommunications Public Limited Company | Wireless ad hoc network |
DE502004003405D1 (en) * | 2004-07-23 | 2007-05-16 | Alcatel Lucent | Method for providing services in a telecommunication network and telecommunication network and network nodes therefor |
US8176490B1 (en) | 2004-08-20 | 2012-05-08 | Adaptive Computing Enterprises, Inc. | System and method of interfacing a workload manager and scheduler with an identity manager |
CA2586763C (en) | 2004-11-08 | 2013-12-17 | Cluster Resources, Inc. | System and method of providing system jobs within a compute environment |
US9075657B2 (en) | 2005-04-07 | 2015-07-07 | Adaptive Computing Enterprises, Inc. | On-demand access to compute resources |
US8863143B2 (en) | 2006-03-16 | 2014-10-14 | Adaptive Computing Enterprises, Inc. | System and method for managing a hybrid compute environment |
US9231886B2 (en) | 2005-03-16 | 2016-01-05 | Adaptive Computing Enterprises, Inc. | Simple integration of an on-demand compute environment |
WO2006112980A2 (en) | 2005-03-16 | 2006-10-26 | Cluster Resources, Inc. | Reserving resources in an on-demand compute environment from a local compute environment |
US9225663B2 (en) | 2005-03-16 | 2015-12-29 | Adaptive Computing Enterprises, Inc. | System and method providing a virtual private cluster |
US7480281B2 (en) * | 2005-11-30 | 2009-01-20 | International Business Machines Corporation | Method for improving cluster bring-up in a distributed topology liveness system |
EP1821489A1 (en) * | 2006-02-20 | 2007-08-22 | Alcatel Lucent | Method of managing real-time services and corresponding packet-based telecommunications network |
US8041773B2 (en) | 2007-09-24 | 2011-10-18 | The Research Foundation Of State University Of New York | Automatic clustering for self-organizing grids |
US11720290B2 (en) | 2009-10-30 | 2023-08-08 | Iii Holdings 2, Llc | Memcached server functionality in a cluster of data processing nodes |
US10877695B2 (en) | 2009-10-30 | 2020-12-29 | Iii Holdings 2, Llc | Memcached server functionality in a cluster of data processing nodes |
US10719777B2 (en) | 2016-07-28 | 2020-07-21 | At&T Intellectual Propery I, L.P. | Optimization of multiple services via machine learning |
US10404549B2 (en) | 2016-07-28 | 2019-09-03 | At&T Intellectual Property I, L.P. | Applying machine learning to heterogeneous data of existing services to generate a new service |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0889656A2 (en) * | 1997-06-12 | 1999-01-07 | Nortel Networks Corporation | Real time control architecture for admission control in communications network |
Family Cites Families (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5341477A (en) * | 1989-02-24 | 1994-08-23 | Digital Equipment Corporation | Broker for computer network server selection |
IL99923A0 (en) * | 1991-10-31 | 1992-08-18 | Ibm Israel | Method of operating a computer in a network |
NZ330166A (en) * | 1993-11-30 | 2000-01-28 | British Telecomm | Communications network with selected links allocated on priority basis, and reconfiguration in the event of agent failure |
US5940622A (en) * | 1996-12-11 | 1999-08-17 | Ncr Corporation | Systems and methods for code replicating for optimized execution time |
US5963447A (en) * | 1997-08-22 | 1999-10-05 | Hynomics Corporation | Multiple-agent hybrid control architecture for intelligent real-time control of distributed nonlinear processes |
GB2332809A (en) * | 1997-12-24 | 1999-06-30 | Northern Telecom Ltd | Least cost routing |
JP3439337B2 (en) * | 1998-03-04 | 2003-08-25 | 日本電気株式会社 | Network management system |
US6272544B1 (en) * | 1998-09-08 | 2001-08-07 | Avaya Technology Corp | Dynamically assigning priorities for the allocation of server resources to completing classes of work based upon achievement of server level goals |
US6728748B1 (en) * | 1998-12-01 | 2004-04-27 | Network Appliance, Inc. | Method and apparatus for policy based class of service and adaptive service level management within the context of an internet and intranet |
US6393474B1 (en) * | 1998-12-31 | 2002-05-21 | 3Com Corporation | Dynamic policy management apparatus and method using active network devices |
US6449650B1 (en) * | 1999-02-01 | 2002-09-10 | Redback Networks Inc. | Methods and apparatus for deploying quality of service policies on a data communication network |
US6473851B1 (en) * | 1999-03-11 | 2002-10-29 | Mark E Plutowski | System for combining plurality of input control policies to provide a compositional output control policy |
US7145869B1 (en) * | 1999-03-17 | 2006-12-05 | Broadcom Corporation | Method for avoiding out-of-ordering of frames in a network switch |
US6628610B1 (en) * | 1999-06-28 | 2003-09-30 | Cisco Technology, Inc. | Methods and apparatus for managing a flow of packets using change and reply signals |
US6701342B1 (en) * | 1999-12-21 | 2004-03-02 | Agilent Technologies, Inc. | Method and apparatus for processing quality of service measurement data to assess a degree of compliance of internet services with service level agreements |
US7478162B2 (en) * | 2000-02-08 | 2009-01-13 | British Telecommunications Public Limited Company | Communications network |
US6477150B1 (en) * | 2000-03-03 | 2002-11-05 | Qualcomm, Inc. | System and method for providing group communication services in an existing communication system |
US6982955B1 (en) * | 2000-03-28 | 2006-01-03 | British Telecommunications Public Limited Company | Communications network |
US6976090B2 (en) * | 2000-04-20 | 2005-12-13 | Actona Technologies Ltd. | Differentiated content and application delivery via internet |
US6985442B1 (en) * | 2000-07-26 | 2006-01-10 | Lucent Technologies Inc. | Technique for bandwidth sharing in internet and other router networks without per flow state record keeping |
GB0022561D0 (en) * | 2000-09-14 | 2000-11-01 | British Telecomm | Communications network |
-
2002
- 2002-03-01 US US10/468,283 patent/US20040071147A1/en not_active Abandoned
- 2002-03-01 CA CA002438540A patent/CA2438540A1/en not_active Abandoned
- 2002-03-01 EP EP02703711A patent/EP1368932A1/en not_active Withdrawn
- 2002-03-01 WO PCT/GB2002/000903 patent/WO2002073889A1/en not_active Application Discontinuation
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0889656A2 (en) * | 1997-06-12 | 1999-01-07 | Nortel Networks Corporation | Real time control architecture for admission control in communications network |
Non-Patent Citations (3)
Title |
---|
CHU-SING YANG; MON-YEN LUO: "Building an adaptable, fault tolerant, and highly manageable web server on clusters of non-dedicated workstations", PARALLEL PROCESSING, 2000. PROCEEDINGS. 2000 INTERNATIONAL CONFERENCE ON, 21 August 2000 (2000-08-21) - 24 August 2000 (2000-08-24), Los Alamitos, CA, pages 413 - 420, XP002175975 * |
GOLDSZMIDT G S: "LOAD MANAGEMENT FOR SCALING UP INTERNET SERVICES", NOMS '98. 1998 IEEE NETWORK OPERATIONS AND MANAGEMENT SYMPOSIUM. NEW ORLEANS, LA, FEBR. 15 - 20, 1998, IEEE NETWORK OPERATIONS AND MANAGEMENT SYMPOSIUM, NEW YORK, NY: IEEE, US, vol. 3 CONF. 10, 15 February 1998 (1998-02-15), pages 828 - 835, XP000793430, ISBN: 0-7803-4352-2 * |
I.W. MARSHALL, C. ROADKNIGHT: "Adaptive Management of an active Service Network", BT TECHNOLOGY JOURNAL, vol. 18, no. 4, 1 October 2000 (2000-10-01), XP002175974 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007052254A1 (en) * | 2005-11-01 | 2007-05-10 | Eci Telecom Ltd. | Access system for the provisioning of different communication services, and method for using same |
Also Published As
Publication number | Publication date |
---|---|
CA2438540A1 (en) | 2002-09-19 |
US20040071147A1 (en) | 2004-04-15 |
EP1368932A1 (en) | 2003-12-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20040071147A1 (en) | Communications network | |
US7478162B2 (en) | Communications network | |
US7346704B2 (en) | Layered communications network with adaptively distributed nodal policies | |
US6982955B1 (en) | Communications network | |
Marshall et al. | Adaptive management of an active service network | |
CN109255513A (en) | A kind of power telecom network scene work order dispatching method | |
Gu et al. | Service function chain deployment and network flow scheduling in geo-distributed data centers | |
Marshall et al. | Provision of quality of service for active services | |
Wei et al. | GRL-PS: Graph embedding-based DRL approach for adaptive path selection | |
Nitash et al. | An artificial bee colony algorithm for minimum weight dominating set | |
Roadknight | Communications network | |
Di Pietro et al. | Evolution of self-organised division of labour driven by stigmergy in leaf-cutter ants | |
Boonyopakorn et al. | A hybrid GA and tabu search approach to find optimal node placement in IP networks | |
Champrasert et al. | Exploring self‐optimization and self‐stabilization properties in bio‐inspired autonomic cloud applications | |
Zhang et al. | Efficient population diversity handling genetic algorithm for QoS-aware web services selection | |
Roadknight et al. | Future Network Management-A bacterium inspired solution | |
Oates | Autonomous management of distributed information systems using evolutionary computation techniques | |
Marshall et al. | Management of future data networks | |
Oates | Evolutionary algorithm performance profiles on the adaptive distributed database management problem | |
Houtzager et al. | An evolutionary approach to optimal web proxy cache placement | |
Yazbek et al. | Adaptive strategies of multi-objective optimization for greener networks | |
Champrasert et al. | Building Self-Configuring Data Centers with Cross Layer Coevolution. | |
Chai et al. | Emergence of specialization from global optimizing evolution in a multi-agent system | |
Marshall et al. | Emergent Quality of service-A Bacterium Inspired Approach. | |
Jha | A Routing Algorithm Inspired from a Distributed Autonomous Multi-Agent System-The Ant Colony |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ OM PH PL PT RO RU SD SE SG SI SK SL TJ TM TN TR TT TZ UA UG US UZ VN YU ZA ZM ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2002703711 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2438540 Country of ref document: CA |
|
WWE | Wipo information: entry into national phase |
Ref document number: 10468283 Country of ref document: US |
|
WWP | Wipo information: published in national office |
Ref document number: 2002703711 Country of ref document: EP |
|
REG | Reference to national code |
Ref country code: DE Ref legal event code: 8642 |
|
NENP | Non-entry into the national phase |
Ref country code: JP |
|
WWW | Wipo information: withdrawn in national office |
Country of ref document: JP |