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
  • H04L41/14—Network analysis or design
  • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L67/00—Network arrangements or protocols for supporting network services or applications
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L67/00—Network arrangements or protocols for supporting network services or applications
  • H04L67/01—Protocols
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L67/00—Network arrangements or protocols for supporting network services or applications
  • H04L67/01—Protocols
  • H04L67/10—Protocols in which an application is distributed across nodes in the network
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L67/00—Network arrangements or protocols for supporting network services or applications
  • H04L67/50—Network services
  • H04L67/52—Network services specially adapted for the location of the user terminal
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L67/00—Network arrangements or protocols for supporting network services or applications
  • H04L67/50—Network services
  • H04L67/535—Tracking the activity of the user
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L12/00—Data switching networks
  • H04L12/02—Details
  • H04L12/16—Arrangements for providing special services to substations
  • H04L12/18—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast
  • H04L12/1813—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast for computer conferences, e.g. chat rooms
  • H04L12/1827—Network arrangements for conference optimisation or adaptation

Definitions

• Embodiments of the invention relate to the field of networking; and more specifically, to the determination of recommended geographic server locations based upon characteristics of existing networks.
• a social network is a set of people (or organizations or other social entities) connected by a set of social relationships, such as friendship, co-working or information exchange relationship.
• OSNs Online Social Networks
• OSNs Besides handling traditional client-to-server requests, OSNs also need to handle highly interconnected data due to the strong community structure and human relationships among their end users, which often results in complex data sharing among users. Given the tremendous user population and frequent data access by these users, effective resource planning and provisioning strategies are of extreme importance to the performance and revenue of an OSN. In particular, selecting the most suitable locations to deploy server farms is one of the key steps in such resource management.
• a computer implemented method to determine a plurality of recommended geographic server locations by attempting to minimize both user-server latency and a latency of inter-user communications.
• the method includes acquiring geographic information for a plurality of users of a set of one or more networks.
• the geographic information for each of the plurality of users indicates a geographic location of that user.
• the method also includes acquiring relationship information for at least some of the plurality of users.
• the relationship information indicates those of the plurality of users that are connected on a network of the set of networks.
• the method includes transforming the geographic information and the relationship information into a graph including a plurality of nodes representing the plurality of users, and a plurality of edges connecting the plurality of nodes according to the relationship information.
• Each of the plurality of edges includes an edge weight.
• the method further includes generating a first plurality of clusters by performing a first clustering algorithm on the graph.
• Each cluster of the first plurality of clusters includes a centroid and a set of one or more nodes of the plurality of nodes.
• Each of the set of nodes is included in only one of the first plurality of clusters.
• the method further includes generating a second plurality of clusters by performing a second clustering algorithm.
• the second clustering algorithm includes iteratively examining pairs of clusters of the first plurality of clusters.
• the second clustering algorithm repeatedly swaps pairs of nodes between the pair of clusters when a swap of a pair of nodes will reduce a total cut weight of the graph, and locate each node of the pair of nodes, when swapped to the other cluster of the examined pair of clusters, within a defined maximum distance from the centroid of the other cluster to thereby bound user-to-server latency.
• the total cut weight is a sum of edge weights of edges that connect nodes in different clusters.
• the method further includes causing information describing geographic locations of centroids of the second plurality of clusters to be presented to a user as the plurality of recommended geographic server locations.
• a computer implemented method to determine a plurality of recommended geographic server locations by attempting to minimize both user-server latency and a latency of inter-user communications using a joint analysis approach based upon characteristics of a plurality of networks.
• the method includes acquiring geographic information for a plurality of users of the plurality of networks.
• the geographic information for a user of the plurality of users indicates a geographic location of the user.
• the method also includes acquiring relationship information for at least some of the plurality of users.
• the relationship information indicates those of the plurality of users that are connected on at least one of the plurality of networks.
• the method further includes transforming the geographic information and the relationship information into a plurality of graphs.
• Each graph of the plurality of graphs represents one network of the plurality of networks.
• Each graph of the plurality of graphs includes a plurality of nodes representing those users of the plurality of users that belong to the one network, and a plurality of edges connecting the plurality of nodes according to the relationship information.
• Each of the plurality of edges includes an edge weight.
• the method also includes generating, for each graph of the plurality of graphs, a plurality of clusters for that graph by performing a clustering algorithm.
• Each of the plurality of clusters includes a centroid.
• the method further includes identifying a first recommended geographic server location by ranking a set of centroids according to the frequency of occurrence of each centroid in the set of centroids (where the set of centroids includes all of the centroids of the plurality of clusters of each of the plurality of graphs), and identifying a centroid of the ranked set of centroids having the highest occurrence as representing the first recommended geographic server location.
• a server end station configured to determine a plurality of recommended geographic server locations by attempting to minimize both user-server latency and a latency of inter-user communications.
• the server end station includes an information acquisition module.
• the information acquisition module is configured to acquire geographic information for a plurality of users of a set of one or more networks.
• the geographic information for each user of the plurality of users indicates a geographic location of that user.
• the information acquisition module is also configured to acquire relationship information for at least some of the plurality of users.
• the relationship information indicates those of the plurality of users that are connected on at least one network of the set of networks.
• the server end station also includes a transformation module configured to transform the geographic information and the relationship information into a graph including a plurality of nodes representing the plurality of users and a plurality of edges connecting the plurality of nodes according to the relationship information. Each of the plurality of edges includes an edge weight.
• the server end station also includes a server placement module. The server placement module is configured to generate a first plurality of clusters by performing a first clustering algorithm on the graph. Each cluster of the first plurality of clusters includes a centroid and a set of one or more nodes of the plurality of nodes. Each node of the set of nodes is included in only one cluster of the first plurality of clusters.
• the server placement module is further configured to generate a second plurality of clusters by performing a second clustering algorithm.
• the second clustering algorithm includes iteratively examining pairs of clusters of the first plurality of clusters.
• the second clustering algorithm also includes for each examined pair of clusters, repeatedly swapping pairs of nodes between the pair of clusters when a swap of a pair of nodes will reduce a total cut weight of the graph (wherein the total cut weight is a sum of edge weights of edges that connect nodes in different clusters) and locate each node of the pair of nodes, when swapped to the other cluster of the examined pair of clusters, within a defined maximum distance from the centroid of the other cluster to thereby bound user-server latency.
• the server end station further includes a presentation module configured to cause information describing geographic locations of centroids of the second plurality of clusters to be presented to a user as the plurality of recommended geographic server locations.
• Figure 1 illustrates a system for determining a plurality of recommended geographic server locations using multiple objectives according to one embodiment of the invention
• Figure 2 illustrates a "joint before” approach for determining a plurality of recommended geographic server locations using aggregated user graphs from a plurality of networks according to one embodiment of the invention
• Figure 3 illustrates portions of a "joint after” approach for determining a plurality of recommended geographic server locations using centroids from within clusters from different network graphs according to one embodiment of the invention
• Figure 4 illustrates additional portions of the "joint after” approach presented in Figure 3 according to one embodiment of the invention
• Figure 5 illustrates a high-level view of an approach for determining a plurality of recommended server locations according to one embodiment of the invention
• Figure 6 illustrates two possible cost functions useful in particular clustering algorithms according to one embodiment of the invention
• FIG. 7 illustrates three server placement algorithms according to one embodiment of the invention.
• Figure 8 illustrates a flow for determining a plurality of recommended geographic server locations by attempting to minimize both user-server latency and a latency of inter-user communications according to one embodiment of the invention.
• Figure 9 illustrates determine a plurality of recommended geographic server locations by attempting to minimize both user-server latency and a latency of inter-user communications using a joint analysis approach based upon characteristics of a plurality of networks according to one embodiment of the invention.
• references in the specification to "one embodiment,” “an embodiment,” “an example embodiment,” etc., indicate that the embodiment described may include a particular feature, structure, or characteristic, but every embodiment may not necessarily include the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the art to effect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described.
• Coupled is used to indicate that two or more elements, which may or may not be in direct physical or electrical contact with each other, co-operate or interact with each other.
• Connected is used to indicate the establishment of communication between two or more elements that are coupled with each other.
• An electronic device e.g., an end station, a network element, a computer / computing system / computing device stores and transmits (internally and/or with other electronic devices over a network) code (composed of software instructions) and data using computer-readable media, such as non-transitory tangible computer-readable media (e.g., computer-readable storage media such as magnetic disks; optical disks; read only memory; flash memory devices) and transitory computer-readable transmission media (e.g., electrical, optical, acoustical or other form of propagated signals - such as carrier waves, infrared signals).
• non-transitory tangible computer-readable media e.g., computer-readable storage media such as magnetic disks; optical disks; read only memory; flash memory devices
• transitory computer-readable transmission media e.g., electrical, optical, acoustical or other form of propagated signals - such as carrier waves, infrared signals.
• such electronic devices typically include a set of one or more processors coupled to one or more other components, such as one or more non-transitory machine-readable media (to store code and/or data), user input/output devices (e.g., a keyboard, a mouse, a touchscreen, and/or a display), and network connections (to transmit code and/or data using propagating signals).
• the coupling of the set of processors and other components is typically through one or more busses and bridges (also termed as bus controllers).
• a non- transitory computer-readable medium of a given electronic device typically stores instructions for execution on one or more processors of that electronic device.
• One or more parts of an embodiment of the invention may be implemented using different combinations of software, firmware, and/or hardware.
• a network element e.g., a router, switch, bridge
• a network element is a piece of networking equipment, including hardware and software, which communicatively interconnects other equipment on the network (e.g., other network elements, end stations).
• Subscriber end stations are computing devices (e.g., servers, workstations, laptops, netbooks, palm tops, mobile phones, smartphones, multimedia phones, Voice Over Internet Protocol (VOIP) phones, user equipment, terminals, portable media players, GPS units, gaming systems, set-top boxes) that access content/services provided over the Internet and/or content/services provided on virtual private networks (VPNs) overlaid on (e.g., tunneled through) the Internet.
• VOIP Voice Over Internet Protocol
• the content and/or services are typically provided by one or more end stations (e.g., server end stations) belonging to a service or content provider or end stations participating in a peer to peer service, and may include, for example, public webpages (e.g., free content, store fronts, search services), private webpages (e.g., username/password accessed webpages providing email services), and/or corporate networks over VPNs.
• end stations e.g., server end stations
• subscriber end stations are coupled (e.g., through customer premise equipment coupled to an access network (wired or wirelessly)) to edge network elements, which are coupled (e.g., through one or more core network elements) to other edge network elements, which are coupled to other end stations (e.g., server end stations).
• An online social network is an online service, platform, or website that facilitates the building of social networks or social relations among its users.
• OSNs are general purpose and provide social networking services for a wide variety of users, while others provide such services for users having particular interests, such as users with common hobbies, activities, backgrounds, or real-life connections.
• OSNs often include a representation of each user (often a profile, including biographic data, photographs, and other information) and a set of his/her social links (e.g. connections to other users, organizations, entities, etc.).
• Many OSNs are web-based and provide means for users to interact through the use of computing devices and the Internet. Certain embodiments of the invention described herein involve server placement for OSNs; however, other embodiments of the invention apply to server placement for other scenarios such as networks of users, including but not limited to users of a cellular network, wired network, wireless network, etc.
• Embodiments of the invention utilize a scalable server placement algorithm based upon graph partitioning.
• Some embodiments employ clustering techniques that partition the whole client space into non-overlapping groups (or clusters) according to a dual purpose, wherein user locations in a cluster are both topologically close to the centroid of the cluster (i.e. which indicate the best suitable locations to minimize user- to-server latency) and wherein each centroid is topologically close to other centroids (i.e. which indicate the best suitable locations to minimize server-to-server latency).
• Embodiments of the invention in implementing the scalable server placement algorithm, take inter-user data sharing / communication into consideration.
• Some embodiments also utilize publicly available social network data to make better resource provisioning decisions for social networks that do not currently exist.
• Embodiments of the invention help improve the performance and reduce operational costs of future (i.e. not yet existing) as well we current online social network services.
• Use of the invention may also reduce energy consumption and carbon emission of such services, especially at their hosting data centers and the communication networks, and help make better usage of computing resources and preserve natural resources.
• updating a user profile for one user may trigger notifications to be sent to multiple other users.
• the subscriptions between connected users triggers the data transfer among storage entities within the OSN.
• these servers reside in different sites, the traffic will have to traverse the Internet, causing additional inter-site network consumption and/or congestion, delay, and additional bandwidth cost.
• selecting servers to minimize cost such a factor should be taken into consideration. Therefore, these connections fundamentally transform the problem's mathematics: in addition to connections between clients and the servers to store their data, there are connections in the communication graph between users' data.
• OSNs are believed to reflect the real-life social relationships of people. Therefore, there are some factors that have been common among most OSNs. Users may be connected (i.e. have a "relationship") based on one relation only, e.g., as members of the same organization. Alternatively, users may maintain a multiplex tie based on many relations, such as sharing information, giving financial support and attending conferences together. These ties reflect the real- life social relationships of people and person-to-person interaction. Moreover, a person's social personality is likely to stay consistent across communities and environments. For instance, a user who is active in one social network is likely to be active in another.
• Embodiments of the invention utilize public information from existing, established Online Social Networks (OSNs), which provides sufficient information to provide intelligent server placement suggestions to new born social network applications.
• OSNs Online Social Networks
• Embodiments of the invention utilize a solution that is formulated as an optimization problem of minimizing a number of required replicas.
• G (V, E) denote the social graph of an OSN, with node set V representing users, and edge set E representing friendship relationships among users.
• social graph G is an undirected symmetric graph with weighted edges, i.e., if (u, v) 6 E, then (v, u) 6 E.
• Each edge is associated with a weight wy representing the number of relationship between user u and v.
• wy 1 if the two nodes stand for a single user.
• wy is the number of pairwise friendships between this group of users.
• a cost function assigns a real number to any given clustering of G. The goal is to find a clustering that minimizes a given cost function. For instance, the cost function could be the sum of the distance between each node and its centroid, or it could be the negative sum of edge weights between clusters.
• Two common ways to partition a graph include using an agglomerative approach, which initializes each element to belong to its own cluster and proceeds to merge clusters until a certain terminating condition is met, and using a partitive clustering approach that starts with a single cluster containing all elements and proceeds by splitting clusters.
• Embodiments of the invention utilize graph partitioning processes that include optimizing for multiple objectives.
• Embodiments of the invention define two objectives to capture both latency and inter-site traffic cost, although in other embodiments these formations can be easily modified to account for other types of performance metrics and costs.
• the center of any cluster "C” is defined as the following, where v is used to enumerate all nodes in the cluster:
• Centroid (C) -
• Centroid (C) -
• P is a given partition solution
• dist(u, v) is the weight of edge (u, v), which is defined to be the user-to-centroid latency.
• nei UJ - is defined as node u's most connected j th neighbor according to the definition of connections among users.
• the value j is used to enumerate all of u's neighbors.
• the value 8(u, v) is a binary variable to denote if u and v are in the same partition.
• inter-cluster traffic cost function 604 "g” measures inter-cluster connectivity to be minimized
• intra-cluster latency cost function 602 "f” measures intra-cluster difference, which should also be minimized.
• Figure 1 illustrates a system for determining a plurality of recommended geographic server locations using multiple objectives according to one embodiment of the invention.
• This system represents one embodiment of the invention that implements the general process depicted in Figure 5, which illustrates a high-level view of an approach for determining a plurality of recommended server locations.
• data from one or more OSNs (502A-502N) is acquired by an information acquisition module 108 using a set of ports 120 (e.g. physical network interfaces) and transformed into one or more graphs by a transformation module 1 12.
• a server placement module 114 which utilizes one or more clustering algorithms (highlighted above) to cluster the graphs, which may generate one or more recommended geographic server locations based upon the clustering.
• a presentation module 1 16 may display the results to a user or transmit the results so that a user is so informed. In embodiments, this transmission is one or more of transmitting the results to a display, transmitting the results to a separate device for viewing or further processing, or transmitting the results to another device (e.g., a printer) to create a physical document or element.
• the presentation module 1 16 is enabled to automatically or semi-automatically reserve or allocate server resources at the recommended geographic server locations, such as by enabling one or more virtual machine and/or network instances at or near the recommended geographic server locations.
• the process may also continue to a joint analysis module 118 configured to utilize data from multiple OSNs when determining recommended geographic server locations.
• the joint analysis module 1 18 may include functionalities of transformation module 1 12 or server placement module 1 14, which will allow the joint analysis module 1 18 to generate its own results, which are shifted to the presentation module 1 16. However, following 516, the joint analysis module 1 18 may cause the server placement module 114 to be used again. In some embodiments, the joint analysis module 1 18 or server placement module 1 14 may immediately (see line 512) be invoked with OSN data (502A- 102N). This may occur, in some embodiments, using OSN data (502A-102N) stored by a user information storage location 1 10 of a computing device 106, as depicted in Figure 1.
• an information acquisition module 108 of the computing device 106 acquires 150 information about a plurality of users 102A-102N and user relationships 104 of one or more networks 103A-103N.
• the networks 103A-103N are OSNs.
• this acquired information includes a set of user profiles from the OSNs, as well as a list of other users with whom each user has relationships with on the network (i.e. "friends" of each user, or some other designation of connected users).
• the acquired information is maintained in a user information storage location 1 10 of the computing device 106.
• the transformation module 1 12 attempts to deduce a geographic location from the user profile for each user (e.g. 102A). Many users explicitly register their current location with the network, while many users enter their birth location, and yet others enter affiliations that reveal a user geographic location. Interestingly, a non-trivial fraction of users' locations are automatically entered by the smart phone applications in the form of longitude and latitude.
• the transformation module 1 12 first retrieves all possible hints on location from the crawled data. The transformation module 1 12 may then pre-process the crawled data to correct typographical errors, eliminate ambiguity, and combine any same location having multiple representations into one (e.g., "California" and "CA” and "Cal.”).
• the transformation module 112 then translates the user location strings into latitude- longitude coordinates using a geocoding database (internal or external to the computing device 106) or a geocoding API, including but not limited to APIs such as the Google Geocoding API or the Yahoo! Maps Web Services Geocoding API.
• a geocoding database internal or external to the computing device 106
• a geocoding API including but not limited to APIs such as the Google Geocoding API or the Yahoo! Maps Web Services Geocoding API.
• the acquired information and the deduced geographic coordinates are transformed by the transformation module 1 12 into a user graph 126 having a plurality of nodes 136 and a plurality of edges 138 connecting some of those nodes.
• the location of each node 136 is based upon the deduced geographic coordinates of the users, and the edges 138 exist between any two nodes 136 that represent users that have a connection on the OSN or OSNs (based upon the acquired user relationship 104 data).
• a latency map is constructed between any pair of users. While in one embodiment, each latency value is a round-trip delays collected for all users. However, in many embodiments due to limited access to end hosts at a large scale, a hypothetical direct link latency is calculated to approximate the latency between two users using the deduced geographic locations. Thus, in an embodiment, the latency between two users is the transmission latency of a hypothetical direct link (e.g. optical, metallic) between those users.
• locations are treated as points and are clustered into regions using the well-known Vincenty's formula, which is used to calculate geodesic distances between points. (Except for networks that use circuit-switching, geographic distance correlates well with the minimum network delay.
• each individual user is represented as one node 136 in the user graph 126
• users physically near each other will share similar network topological properties. Therefore, in an embodiment of the invention users are aggregated to groups to form an aggregated user graph 128.
• users located within a range of latitude and longitude from each other e.g., within 2 degrees of latitude and longitude
• a node weight 141 is assigned to each node (e.g. 139) based upon the number of users in the group represented by the node 139.
• the top left aggregated node 139 contains a node weight 141 of '3' because there are three nodes from the original user graph 126 that were located within a particular distance of each other and thus were aggregated together.
• an edge weight 137 is assigned to an edge that represents the total number of relationships (e.g. friendship connections) between users in the two connected groups.
• These weights are represented in Figure 1 using thick lines (for heavier comparative weights) or thin lines (for lighter/smaller comparative weights), but in embodiments of the invention the weights are integers, real numbers, binary values, or the like.
• the server placement module 1 14 utilizes the aggregated user graph 128 to determine recommended geographic server locations 124 using a variety of placement algorithms. While in one embodiment an ultimate goal is to minimize user-to-server latency and a latency of inter-user communications simultaneously, the following description presents three separate algorithms for determining the recommended geographic server locations 124, which may be more suited for certain applications based upon the constraints and preferences involved: one algorithm for minimizing user-to-server latency 702, one algorithm for minimizing a latency of inter-user communications 704, and one algorithm to optimize for both 706.
• Figure 7 illustrates these three server placement algorithms according to one embodiment of the invention.
• the server placement module 1 14 is configured to utilize an algorithm for primarily minimizing user-to-server latency 702.
• the algorithm 702 is based upon a k-means or k-medoids clustering algorithm, but with several enhancements.
• the server placement module 1 14 begins with k elements as the centroid.
• the value of k is selected by the user to signify the number of recommended geographic server locations 124 that are sought by the user.
• the server placement module 114 iteratively selects a new centroid to minimize the distance between samples in the same group to the centroid of the group.
• a centroid of cluster Ci all the points U j belonging to this cluster are computed with center a in step 2.
• the server placement module 1 14 re- computes the new center C for each cluster P; according to the new grouping of Xj.
• the complexity of this algorithm is 0( ⁇ ), where k is the final number of clusters, n is the total data points, and ⁇ is the number of iterations.
• the value ⁇ has been significantly reduced through a careful selection of k and the initial set based on realistic conditions.
• the server placement module 1 14 doesn't simply apply the classic k-means or k-medoids style clustering solutions blindly, but carefully selects the initial sets according to our problem context.
• the server placement module 1 14 starts the selection of known data center locations of popular OSNs and uses them as the initial candidates (i.e. centroids) for the algorithm, which efficiently reduces the number of iterations of clustering.
• the server placement module 1 14 for each k, the server placement module 1 14 always includes the set of servers selected in the k - 1 experiment in the initial set. The remaining is selected based on ranking of user population.
• the algorithm takes the sample density information (e.g., the weights) into consideration, instead of treating each data point as equivalent.
• the server placement module 1 14 is configured to utilize an algorithm for primarily minimizing a latency (or, an average latency) of inter-user communications 704.
• a latency or, an average latency
• the algorithm 704 requires that the partition needs to be as balanced as possible.
• the balance property can also help providing balanced load and best resource utilization. For example, if one best location is used to serve all the users, it may easily create bandwidth bottlenecks.
• the algorithm 704 is based upon a minimum-cut clustering algorithm with modifications.
• the algorithm 704 is reproduced below:
• the server placement module 114 finds the best pair of nodes vi 6 CI and V2 £ C2 to exchange to maximize the gain.
• the server placement module 1 14 executes this procedure recursively until no further gain can be obtained by changing any pairs of nodes.
• the target of the algorithm 704 is still a partitioning of k clusters. Lines 9- 10 further partition the two initial two clusters Ci and C 2 , respectively. However, their targets are reduced to k- 1 and k-2 accordingly.
• the server placement module 1 14 is configured to utilize an algorithm for minimizing both user-to-server latency and a latency of inter-user communications 706, which is an extension of both the algorithm for minimizing user- to-server latency 702 and the algorithm for minimizing a latency of inter-user communications 704.
• the algorithm 706 is reproduced below:
• the server placement module 1 14 initializes the partition using the algorithm for minimizing user-to-server latency 702 and then iteratively switches nodes to reduce the cuts similar to that of the algorithm for minimizing a latency of inter-user communications 704.
• the server placement module 1 14 imposes maximum latency constraints between the centroid and the new node if the move is to be performed. While in some embodiments the algorithm 706 may finish with a local optimal solution instead of a global optimal one, given the NP hard property of the problem, the algorithm is much more efficient and scalable than other integer programming based solutions.
• the server placement module 1 14 is depicted as configured to utilize the algorithm for minimizing both user-to-server latency and a latency of inter-user communications 706, which, in an embodiment, includes two sub- algorithms, each performing operations similar to algorithms 702 and 704. Accordingly, the server placement module 1 14 includes two first clusters obtained from performing the first sub-algorithm 130, which depicts two potential clusters (and centroids 134A) generated by the algorithm that attempted to minimize user-to-server latency. Then, the server placement module 1 14 performs the second sub-algorithm and generates another two clusters as a result 132.
• this process entails examining pairs of nodes 135 from different clusters and swapping these nodes if and only if the cut size is reduced and the distance of each swapped node is within a maximum latency constraint distance from the centroid (e.g. 134B) of the new cluster.
• this second set of clusters 132 is shows with a first swapped node 'A' 131A and a second swapped node 'B' 13 IB, which reveals that the swap occurred because the cut size was reduced and each node was within the maximum latency constraint distance from the centroid (e.g. 134B).
• each centroid e.g.
• the presentation module 1 16 upon completion of the algorithm 706 by the server placement module 1 14, the presentation module 1 16 utilizes port(s) 120 to cause the recommended geographic server locations 124 to be presented to a user, but in other embodiments, the presentation module 1 16 utilizes a display port 122 to display the recommended geographic server locations 124 to a user on a display device.
• the display port 122 is a physical interface of the computing device 106 used to transmit visual information to a user, either through a wired or wireless interface.
• the computing device 106 includes a joint analysis module 1 18. As depicted, the joint analysis module 1 18 interacts with the server placement module 114 and the presentation module 1 16; however, in other embodiments (not depicted) the joint analysis module 1 18 is used with other modules and used in a different order than that depicted.
• the joint analysis module 1 18 allows the computing device 106 to utilize information from multiple existing OSNs in additional ways when determining recommended geographic server locations 124.
• One use enabled by the joint analysis module 1 18 is depicted in Figure 2 and provides "joint before” analysis of OSN data, and another use enabled by the joint analysis module 1 18 is depicted in Figure 3 and Figure 4 and provides "joint after" analysis of OSN data.
• Figure 2 illustrates a "joint before” approach for determining a plurality of recommended geographic server locations using aggregated user graphs from a plurality of networks according to one embodiment of the invention.
• the joint analysis module 1 18 first combines all the input nodes and edges to create a super graph:
• this is depicted as the first, second, and third aggregated user graphs (202, 204, and 206) as being combined into "joint before” super graph 208.
• the aggregated user graphs (202, 204, and 206) are not first created; rather, individual nodes and edges from the different OSNs are added one by one to create the "joint before” super graph 208.
• the server placement module 1 14 takes G (the "joint before” super graph
• Figure 3 and Figure 4 illustrate portions of a "joint after" approach for determining a plurality of recommended geographic server locations using centroids from within clusters from different network graphs according to one embodiment of the invention.
• the server placement module 1 14 is configured to generate graph partitions for multiple networks 350.
• Figure 3 depicts this as a first, second, and third network clustered graph (302, 304, and 306).
• Each of these network clustered graphs includes clusters (e.g. 301) and a centroid (e.g. 31 1) for each cluster.
• the server placement module 1 14 identifies the centroids from each networked clustered graph as 303, 305, and 307.
• the joint analysis module 1 18 is then tasked with selecting a centroid having the largest occurrence 352.
• the joint analysis module 1 18 consolidates the centroids 308 and then ranks the centroids 310 according to their occurrence.
• the centroid with the largest occurrence 312 is selected - here, a centroid in both PI 302 and P2 304 is the same - at (8,3) - and thus is the centroid having the largest occurrence 312.
• This largest occurrence centroid 312 is identified as the first recommended server location 314.
• the coordinate or location of the centroid may already represent latitude and/or longitude or another geographic indication, or it may need to be transformed from some other representation into a geographic location.
• the process would end. However, in an embodiment the process will repeat until k recommended server locations are calculated. Thus, optionally (assuming the value of "k" is larger than T), the process proceeds to 454 to remove nodes covered by the selected centroid. At this point, in an embodiment the joint analysis module 1 18 removes all nodes and edges covered by the selected centroid from each graph 302, 304, and 306 to generate updated network graphs 402, 404, 406.
• the joint analysis module 1 18 is configured to remove the nodes and edges covered by the selected centroid by removing the cluster (and all nodes within) associated with the centroid from the graphs (here, removing a cluster from each updated network graph 302 and 304). For those graphs not including that centroid (e.g. 306), the cluster located closest to the selected centroid (or the cluster having a centroid located closest to the selected centroid) is removed. Thus, the process continues with the server placement module 114 utilizing the updated network graphs 402, 404, and 406 to again generate graph partitions for the multiple networks 350. This results in new clusters (e.g., 420) and centroids (e.g.
• Figure 8 illustrates a flow 800 for determining a plurality of recommended geographic server locations by attempting to minimize both user-server latency and a latency of inter-user communications according to one embodiment of the invention.
• the operations of this and other flow diagrams will be described with reference to the exemplary embodiments of the other diagrams. However, it should be understood that the operations of the flow diagrams can be performed by embodiments of the invention other than those discussed with reference to these other diagrams, and the embodiments of the invention discussed with reference these other diagrams can perform operations different than those discussed with reference to the flow diagrams.
• the flow 800 includes acquiring geographic information for a set of one or more users of a set of one or more networks.
• the geographic information for a user of the plurality of users indicates a geographic location of the user.
• each network of the set of networks is an online social network.
• relationship information is acquired for at least some of the plurality of users. The relationship information indicates those of the plurality of users that are connected on at least one network of the set of networks.
• the geographic information and the relationship information are transformed into a plurality of graphs 806.
• Each graph of the plurality of graphs represents one network of the plurality of networks.
• Each of the graphs includes a plurality of nodes representing those users of the plurality of users that belong to the one network, and a plurality of edges connecting the plurality of nodes according to the relationship information.
• Each of the plurality of edges includes an edge weight.
• a first plurality of clusters is generated by performing a first clustering algorithm on the graph 808.
• Each cluster of the first plurality of clusters includes a centroid and a set of one or more nodes of the plurality of nodes.
• Each node of the set of nodes is included in only one of the first plurality of clusters.
• a second plurality of clusters is generated by performing a second clustering algorithm 810.
• the second clustering algorithm includes iteratively examining pairs of clusters of the first plurality of clusters.
• the second clustering algorithm also includes, for each examined pair of clusters, repeatedly swapping pairs of nodes between the pair of clusters when a swap of a pair of nodes satisfies two conditions.
• the first condition is that the swap would reduce a cut weight of the graph, wherein the cut weight is a sum of edge weights of edges that connect nodes in different clusters.
• the second condition is that the swap would locate each node of the pair of nodes, when swapped to the other cluster of the examined pair of clusters, within a defined maximum distance from the centroid of the other cluster.
• FIG. 9 illustrates a flow 900 to determine a plurality of recommended geographic server locations by attempting to minimize both user-server latency and a latency of inter-user communications using a joint analysis approach based upon characteristics of a plurality of networks according to one embodiment of the invention.
• the flow 900 includes acquiring geographic information for a plurality of users of a plurality of networks.
• the geographic information for a user of the plurality of users indicates a geographic location of the user.
• the plurality of networks is a plurality of online social networks.
• relationship information is acquired for at least some of the plurality of users. The relationship information indicates those of the plurality of users that are connected on at least one of the plurality of networks.
• Each graph of the plurality of graphs represents one network of the plurality of networks.
• Each of the graphs includes a plurality of nodes representing those users of the plurality of users that belong to the one network, and a plurality of edges connecting the plurality of nodes according to the relationship information.
• Each of the plurality of edges includes an edge weight.
• the flow 900 includes generating, for each graph of the plurality of graphs, a plurality of clusters for that graph by performing a clustering algorithm.
• Each cluster of the plurality of clusters includes a centroid.
• a first recommended geographic server location is identified.
• a set of centroids is ranked according to the frequency of occurrence of each centroid in the set of centroids. This set of centroids includes all of the centroids of the plurality of clusters of each of the plurality of graphs.
• a centroid of the ranked set of centroids that has the highest occurrence is identified as representing the first recommended geographic server location.
• Such future devices will likely have embedded communication and data processing capabilities, and with the help of software and services, these devices will conceptually be able to communicate with each other as human friends do on OSNs today.
• communications must be optimized and thus the techniques described herein will be applicable.
• these networks also must solve the server placement problem by considering messaging latencies, communication patterns (or social strengths) of nodes, and costs of data centers; accordingly, the systems, methods, and apparatuses disclosed are equally applicable and beneficial in these environments.

Landscapes

• Engineering & Computer Science (AREA)
• Computer Networks & Wireless Communication (AREA)
• Signal Processing (AREA)
• Computer Hardware Design (AREA)
• General Engineering & Computer Science (AREA)
• Information Transfer Between Computers (AREA)
• Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Priority Applications (3)

Applications Claiming Priority (2)

Publications (2)

Family

ID=50001039

Family Applications (1)

Country Status (5)

Cited By (2)

Families Citing this family (70)

Family Cites Families (8)

• 2012
  • 2012-11-21 US US13/684,125 patent/US9083757B2/en active Active
• 2013
  • 2013-10-24 WO PCT/IB2013/059633 patent/WO2014080304A2/en not_active Ceased
  • 2013-10-24 IN IN2941DEN2015 patent/IN2015DN02941A/en unknown
  • 2013-10-24 CN CN201380060956.4A patent/CN104995870B/zh not_active Expired - Fee Related
  • 2013-10-24 EP EP13824165.8A patent/EP2923459B1/en active Active
• 2015
  • 2015-06-05 US US14/732,009 patent/US9426233B2/en active Active

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events