Abstract: The illustrative embodiments provide a method, apparatus, and computer program product for managing a number of cache lines in a cache. In one illustrative embodiment, it is determined whether activity on a memory bus in communication with the cache exceeds a threshold activity level. A least important cache line is located in the cache responsive to a determination that the threshold activity level is exceeded, wherein the least important cache line is located using a cache replacement scheme. It is determined whether the least important cache line is clean responsive to the determination that the threshold activity level is exceeded. The least important cache line is selected for replacement in the cache responsive to a determination that the least important cache line is clean.

Abstract: An adaptive prediction threshold scheme for dynamically adjusting prediction thresholds of entries in a Pattern History Table (PHT) by observing global tendencies of the branch or branches that index into the PHT entries. A count value of a prediction state counter representing a prediction state of a prediction state machine for a PHT entry is obtained. Count values in a set of counters allocated to the entry in the PHT are changed based on the count value of the entry's prediction state counter. The prediction threshold of the prediction state machine for the entry may then be adjusted based on the changed count values in the set of counters, wherein the prediction threshold is adjusted by changing a count value in a prediction threshold counter in the entry, and wherein adjusting the prediction threshold redefines predictions provided by the prediction state machine.

Abstract: A method, system and computer program product for detecting denial-of-service attacks. The randomness in the Internet Protocol (IP) source addresses of transmitted IP packets may be detected by performing a hash function on the IP source addresses thereby generating one or more different hash values. If a high number of different hash values were generated for a small number of IP packets evaluated, then random IP source addresses may be detected. By detecting random source IP addresses, a denial-of-service attack may be detected.

Abstract: A method and system within a telecommunications network for allocating available bandwidth among a plurality of sessions that share a common data link. First, a repeating unit time interval is designated for conducting the plurality of sessions. Next, the input flow rate from the source node of each of the sessions into the common data link is monitored. Thereafter, a target flow rate is computed and assigned to each of the sessions. The sum of the target flow rates is equal to the available bandwidth of the common data link. Finally, for each of the sessions in which the monitored input flow rate exceeds the assigned target flow rate, the source node is dynamically paused during each repeating unit time interval, such that the monitored input flow rate conforms to the assigned target flow rate for each of the sessions.

Abstract: A data transfer flow control system for a packet ATM communications system includes a plurality of nodes interconnected by transmission links. Available bandwidth allocation for data packets or cells is improved by defining a Nrm parameter to be the product of an allowable cell rate and a sampling period. Nrm is related to the number of data cells being transmitted between resource management cells in a transmission session, and is set to a number that is independent of round trip times between source and destination nodes, but varying with the allowable cell rate of the source. The allocation fairness problem is solved by normalizing the Nrm value to ensure that the inter-RM cell times for all sessions are fixed at a predetermined time interval. The selection of the predetermined time interval value is a trade-off between network overhead and the responsiveness of the traffic source to network congestion.

Abstract: A data transfer flow control system for a packet communications system includes a plurality of nodes interconnected by transmission links. The rate at which a sender node transmits information to a destination node in a network is modified in accordance with congestion information returned to the sender node from nodes along the path of the transmission or from the information receiver. The rate change for information being sent from the sender node is modified based upon the amount of elapsed time occurring since the last rate change of the same type. In first and second examples, the rate change is implemented in accordance with exponential and linear relationships, respectively, between the modified flow rate and the elapsed time since the last rate change.

Abstract: A method of flow control for Available Bit Rate (ABR) sources in an Asynchronous Transfer Mode (ATM) network is implemented. An effective rate for the source is determined by an ATM switch based on a critically damped second order system. The effective rate is damped toward a share value that is based on the source queue length, a target queue length, and the actual cell rates of the connected ABR sources sending traffic through the ATM switch. The resulting feedback loop ensures that the source queue length will not exceed the target queue length.

Abstract: Cache management in a computer system is achieved by using a cache with dual stacks with storage locations for labels associated with data units stored either in the stack or in a separate cache memory unit. The first or preliminary stack is always refreshed following each label request using a conventional Least Recently Used (LRU) algorithm. If the requested label exists in either stack, the conventional LRU algorithm is also used to update the second stack. If the requested label does not exist in either stack and the bottom position in the second stack is empty, the label is written into the top position of the second stack and existing stack entries are pushed down one position. If the bottom position in the second stack is not empty, stack entries at and below a predetermined insertion point (which may be set statically or dynamically) are pushed down, the entry in the bottom position in the stack is discarded and the label is written into the predetermined insertion point in the second stack.

Abstract: A data transfer flow control system includes a plurality of nodes or switches in a network, which are interconnected by transmission links. Resource management cells are transmitted along with data cells in information transfer sessions. As the amount of session traffic at any node increases, the level of occupancy of the buffers for that node correspondingly increases to temporarily store overflow data. At every network node, a “fair share” cell rate is determined for each output transmission link. Whenever the current cell rate (CCR) of any session exceeds the fair share cell rate for the output link at a node, the buffer occupancy threshold values for setting congestion indication fields in the resource management cells are adjusted downwardly such that a lower threshold buffer occupancy will effect a traffic congestion setting for that session.

Abstract: A data transfer flow control system for a packet communications system includes a plurality of nodes interconnected by transmission links. The rate at which a sender node transmits information to a destination node in a network is modified in accordance with congestion information returned to the sender node from nodes along the path of the transmission or from the information receiver. The rate change for information being sent from the sender node is modified based upon the amount of elapsed time occurring since the last rate change of the same type. In first and second examples, the rate change is implemented in accordance with exponential and linear relationships, respectively, between the modified flow rate and the elapsed time since the last rate change.

Abstract: A bin packing algorithm is employed to schedule computer network activities, such as pause times required for operation of an Ethernet network which implements existing IEEE 802.3x standards. In such a network, any node in the network can control the flow of traffic from upstream stations in order to avoid congestion at the flow-controlling node. Upon sensing congestion, the flow-controlling node determines how long each upstream node contributing to the congestion should pause transmission over the next control interval. In accordance with the invention, the pause times are scheduled or staggered by using the bin packing algorithm to sort the sources into one or more bins. One required bin property is that there is no overlap in pause times for the sources within a particular bin. Another required bin property is that the sum of the pause times within a bin can be no greater than the length of the control interval.

Abstract: An enhanced adaptive rate-based congestion control system for packet transmission networks uses the absolute rather than the relative network queuing delay measure of congestion in the network. Other features of the congestion control system include test transmissions only after a predetermined minimum time, after the receipt of an acknowledgment from the previous test, or transmission of a minimum data burst, whichever takes longest. The congestion control system also provides a small reduction in rate at low rates and a large reduction in rates at high rates. A logarithmic rate control function provides this capability. Rate damping is provided by changing all of the values in a rate look-up tables in response to excessive rate variations. Finally, the fair share of the available bandwidth is used as the starting point for rates at start-up or when a predefined rate damping region is exited.

Abstract: An explicit rate algorithm is disclosed for use in an end-to-end closed loop flow control algorithm for an ATM network which carries at least constant bit rate (CBR) traffic, variable bit rate (VBR) traffic and adjustable bit rate (ABR) traffic. The algorithm determines how much additional bandwidth is available for ABR traffic on an output link from a node and produces an explicit rate variable that can be forwarded to a source for the ABR traffic. Both the maximum and minimum bandwidths already reserved for all connections on the output link are determined. A single reserved bandwidth value is chosen within the range defined by the maximum and minimum reserved bandwidth values. The current utilization of the ABR input buffer for the node is also determined. The explicit rate variable is generated as a function of the link capacity, the reserved bandwidth value and the current utilization of the ABR input buffer.

Abstract: Access agents in nodes at the LAN/WAN interface are formed into a group of access agents so that the access agents may be managed by the WAN as a group. The group must maintain group operation integrity in that if communications between agents in the group are broken, the access agents will coalesce into subgroups and continue performing communication jobs as a group activity. Each of the access agents contains a finite state machine to perform the tasks of group formation and maintenance. The formation of interconnected access agents into a group is accomplished by one access agent being identified as a group leader. All other access agents communicating with the group leader within the LAN may then join the group. The maintenance of group activity integrity is accomplished by detecting a break in group communication integrity and thereafter reforming the group into multiple smaller groups.