Allocation of shared reserve memory to queues in a network device

Information

  • Patent Grant
  • 11973696
  • Patent Number
    11,973,696
  • Date Filed
    Monday, January 31, 2022
    2 years ago
  • Date Issued
    Tuesday, April 30, 2024
    7 months ago
Abstract
A network device includes one or more ports, a packet processor, and a memory management circuit. The one or more ports are to communicate packets over a network. The packet processor is to process the packets using a plurality of queues. The memory management circuit is to maintain a shared buffer in a memory and adaptively allocate memory resources from the shared buffer to the queues, to maintain in the memory, in addition to the shared buffer, a shared-reserve memory pool for use by a defined subset of the queues, to identify in the subset a queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pool, and to allocate memory resources to the identified queue from the shared-reserve memory pool.
Description
FIELD OF THE INVENTION

The present invention relates generally to network communication, and particularly to management of memory resources in network devices.


BACKGROUND OF THE INVENTION

Some network devices, such as packet switches, use shared memory schemes for efficient storage of packets and other data. For example, U.S. Pat. No. 10,250,530 describes techniques for flexible buffer allocation in a network switch. The patent describes a communication apparatus that includes multiple interfaces configured to be connected to a packet data network for receiving and forwarding of data packets of multiple types. A memory is coupled to the interfaces and configured as a buffer to contain packets received through the ingress interfaces while awaiting transmission to the network via the egress interfaces. Packet processing logic is configured to maintain multiple transmit queues, which are associated with respective ones of the egress interfaces, and to place both first and second queue entries, corresponding to first and second data packets of the first and second types, respectively, in a common transmit queue for transmission through a given egress interface, while allocating respective spaces in the buffer to store the first and second data packets against separate, first and second buffer allocations, which are respectively assigned to the first and second types of the data packets.


SUMMARY OF THE INVENTION

An embodiment of the present invention that is described herein provides a network device including one or more ports, a packet processor, and a memory management circuit. The one or more ports are to communicate packets over a network. The packet processor is to process the packets using a plurality of queues. The memory management circuit is to maintain a shared buffer in a memory and adaptively allocate memory resources from the shared buffer to the queues, to maintain in the memory, in addition to the shared buffer, a shared-reserve memory pool for use by a defined subset of the queues, to identify in the subset a queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pool, and to allocate memory resources to the identified queue from the shared-reserve memory pool.


In an embodiment, the memory management circuit is to identify that the queue requires additional memory resources by receiving a request from the queue. In some embodiments, the memory management circuit is to identify that the queue is not eligible for additional allocation from the shared buffer, by identifying that the occupancy of the queue exceeds a MAX threshold set for the queue. In an example embodiment, the memory management circuit is to verify that the occupancy of the queue meets the eligibility condition for the shared-reserve memory pool, by verifying that the occupancy is no more than a defined margin above the MAX threshold.


In a disclosed embodiment, the memory management circuit is to pre-allocate to the queue a private-reserve memory resource, irrespective of the shared buffer and the shared-reserve memory pool. In an embodiment, the queues in the subset are associated with a same port of the network device.


There is additionally provided, in accordance with an embodiment that is described herein, a memory management circuit in a network device. The memory management circuit includes one or more interfaces and circuitry. The one or more interfaces are to communicate with a plurality of queues of the network device that process packets, and with a memory. The circuitry is to maintain a shared buffer in the memory and adaptively allocate memory resources from the shared buffer to the queues, to maintain in the memory, in addition to the shared buffer, a shared-reserve memory pool for use by a defined subset of the queues, to identify in the subset a queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pool, and to allocate memory resources to the identified queue from the shared-reserve memory pool.


There is also provided, in accordance with an embodiment that is described herein, a method for communication in a network device. The method includes communicating packets over a network, and processing the packets using a plurality of queues. A shared buffer is maintained in a memory, and memory resources are allocated adaptively from the shared buffer to the queues. In addition to the shared buffer, a shared-reserve memory pool is maintained in the memory for use by a defined subset of the queues. A queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pool, is identified in the subset. Memory resources are allocated to the identified queue from the shared-reserve memory pool.


There is further provided, in accordance with an embodiment that is described herein, a method for memory management in a network device. The method includes communicating with a plurality of queues of the network device that process packets, and with a memory. A shared buffer is maintained in a memory, and memory resources are allocated adaptively from the shared buffer to the queues. In addition to the shared buffer, a shared-reserve memory pool is maintained in the memory for use by a defined subset of the queues. A queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pool, is identified in the subset. Memory resources are allocated to the identified queue from the shared-reserve memory pool.


There is additionally provided, in accordance with an embodiment that is described herein, a method for memory management in a network device that processes packets using queues. The method includes maintaining in a memory (i) a shared buffer and (ii) shared-reserve memory pool. A queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) is eligible for allocation from the shared-reserve memory pool, is identified. Memory resources are allocated to the identified queue from the shared-reserve memory pool.


The present invention will be more fully understood from the following detailed description of the embodiments thereof, taken together with the drawings in which:





BRIEF DESCRIPTION OF THE DRAWINGS


FIG. 1 is a block diagram that schematically illustrates a network device comprising a shared buffer, a private reserve memory and a shared reserve memory pool, in accordance with an embodiment of the present invention;



FIG. 2 is a diagram that schematically illustrates occupancy thresholds and ranges used for allocating memory resources to a queue in the network device of FIG. 1, in accordance with an embodiment of the present invention; and



FIG. 3 is a flow chart that schematically illustrates a method for allocating memory resources to queues in the network device of FIG. 1, in accordance with an embodiment of the present invention.





DETAILED DESCRIPTION OF EMBODIMENTS
Overview

Embodiments of the present invention that are described herein provide improved methods and apparatus for allocating memory resources to queues in network devices, e.g., in packet switches.


In some embodiments, a network device comprises one or more ports for sending and receiving packets over a network, and a packet processor that processes the packets. As part of the packet processing, the packet processor queues the packets in a plurality of queues. In a typical implementation, a given port is assigned separate receive and transmit queues. The packets of a given direction (receive or transmit) are queued in separate queues according to their priority group. Typically, the packet data is stored in a memory, and the queues store descriptors or pointers that point to the data.


In real-life network devices, traffic load may vary significantly from one port to another and/or over time. Consequently, the memory requirements of the queues are typically highly variable, as well. Since the overall memory resources of the network device are limited, memory allocation to the queues has to be fast, flexible and efficient in order to maintain high throughput and avoiding congestion. In some embodiments of the present invention, the network device comprises a Memory Management Circuit (MMC) that manages the memory resources of the network device, including allocating suitable memory resources to the various queues.


In some embodiments, the MMC maintains three different types of memory regions in the memory, referred to as a shared buffer, a private reserve memory and a shared reserve memory pool. The MMC allocates memory resources from the three regions to the queues according to criteria that are described herein.


Resources of the private reserve memory are typically allocated a-priori to the queues, and remain allocated to the same queues regardless of actual queue occupancies. Such a-priori, static allocation enables the queues to respond immediately to fast-developing traffic bursts.


Memory resources of the shared buffer are typically allocated to the queues adaptively, depending on individual queue occupancy. Typically, a highly active and full queue will be allocated more memory from shared buffer 44, whereas an idle queue will receive little or no shared-buffer allocation. A queue will typically begin requesting allocation from the shared buffer after it has exhausted its pre-allocated private-reserve allocation. The shared buffer allocation mechanism maintains fairness among the active queues.


In some embodiments, the MMC defines a maximal occupancy (referred to herein as a “MAX-STATE” threshold), above which a queue is not eligible to receive additional allocations from the shared buffer. The MAX-STATE threshold is adaptive and depends on the overall demand for shared-buffer resources. When the overall demand for shared-buffer resources is high, e.g., when some queues experience congestion, the MMC will reduce the MAX-STATE threshold. When the overall demand for shared-buffer resources is low, the MMC will increase the MAX-STATE threshold. This mechanism introduces a certain degree of fairness among the queues in competing for shared buffer resources.


In practice, the above-described allocation scheme may be problematic in some scenarios. Consider, for example, a scenario in which one queue has a relatively low activity level and small occupancy, while other queues are highly active and full. With the above-described allocation schemes, the large and busy queues will receive large allocations from the shared buffer, and will cause the MMC to reduce the MAX-STATE threshold. As a result, the low-activity queue may become ineligible for shared-buffer allocation. In such a situation, the low-activity queue may not cope well with bursts of packets, e.g., may become congested or drop packets.


As seen in this example, a low-activity queue may suffer from the fact that higher-activity queues dominate the shared-buffer allocations. This scenario is especially problematic because a low-activity queue will also typically receive a small private-reserve allocation to begin with.


In some embodiments the MMC avoids the above-described scenario, and possibly other problematic scenarios, using the resources of the shared reserve memory pool. In these embodiments, the MMC defines an additional range of occupancy levels, above the MAX-STATE threshold, in which a queue is eligible for allocation from the shared reserve memory pool. In an embodiment, the MMC identifies a queue that requires additional memory but is too full to be eligible for allocation from the shared buffer (i.e., a queue whose occupancy exceeds the MAX-STATE threshold). If the occupancy level of the queue is not too high above the MAX-STATE threshold (e.g., up to a predefined margin above the MAX-SATATE threshold), the MMC allocates to the queue additional memory from the shared reserve memory pool. The shared-reserve mechanism thus ensures that a small and relatively idle queue will not be starved of memory resources due to high demand from other queues.


In some embodiments, the MMC maintains multiple separate shared reserve memory pools for multiple respective subsets of queues. For example, the MMC may maintain a shared reserve memory pool for the queues of a respective port.


Several example implementations of the disclosed memory allocation scheme are described herein.


System Description


FIG. 1 is a block diagram that schematically illustrates a network device 20, in accordance with an embodiment of the present invention. Network device 20 may comprise, for example, a switch, a router, a network adapter such as a Network Interface Controller (NIC), or any other suitable network device that processes packets. The description that follows will refer mainly to a packet switch, by way of example.


Network device 20 comprises one or more ports 24 for receiving packets from a packet network 28 and/or for transmitting packets to network 28. A switch or router typically comprises a plurality of ports 24. A network adapter may comprise a single port 24 or multiple ports 24. Network 28 and network device 20 may operate in accordance with any suitable network communication protocol, such as, for example, Ethernet, InfiniBand™ or the Nvidia forwarding protocol (“NVL”).


Network device 20 further comprises a packet processor 32, which processes the packets that are received and transmitted via ports 24. In a switch, for example, packet processor 32 forwards each incoming packet to a suitable port for transmission.


As part of the processing of packets, packet processor 32 queues the packets in multiple queues 36. In practice, queues 36 typically do not store the actual packet data, but rather descriptors, pointers or other metadata that points to the packets. For clarity, queuing of this sort is also referred to herein as “queuing the packets.” In an example embodiment, for a given port 24, separate queues are designated for reception (“ingress”) and for transmission (“egress”). For a given direction (transmission or reception), separate queues are designated for different priority groups being used.


Network device 20 further comprises a memory 40 that stores, inter alia, the packet data including headers and/or payloads. Resources of memory 40 can be allocated to the various queues 36 for storage of packets. Three separate memory regions are defined in memory 40, namely a shared buffer 44, a private reserve memory 48 and a shared reserve memory pool 52. In alternative embodiments, at least some of the regions of memory 40 (e.g., shared buffer 44) may reside externally to the network device.


Memory 40 is managed by a Memory Management Circuit (MMC) 56. Among other tasks, MMC 56 defines shared buffer 44, private reserve memory 48 and shared reserve memory pool 52, and allocates memory resources from the various memory regions to queues 36. Example allocation schemes are described in detail below. MMC 56 typically comprises one or more interfaces for communicating with queues 36 and with memory 40, and circuitry for performing the memory allocation tasks described herein.


Network device 20 further comprises a controller 60, which performs various control-plane and management tasks.


The configurations of network device 20 and its various components, such as packet processor 32, MMC 56 and memory 40, are example configurations that are chosen purely for the sake of conceptual clarity. Any other suitable configurations can be used in alternative embodiments. In various embodiments, network device 20 and its various components can be implemented using hardware, e.g., using one or more Application-Specific Integrated Circuits (ASIC) and/or Field-Programmable Gate Arrays (FPGA), using software, or using a combination of hardware and software components. Memory 40 typically comprises a suitable Random-Access Memory (RAM).


In some embodiments, certain components of network device 20, e.g., controller 60 and possibly some of functions of MMC 56, may be implemented using a general-purpose processor, which is programmed in software to carry out the functions described herein. The software may be downloaded to the processor in electronic form, over a network, for example, or it may, alternatively or additionally, be provided and/or stored on non-transitory tangible media, such as magnetic, optical, or electronic memory.


Memory Allocation Using Shared Buffer, Private Reserve Memory and Shared Reserve Memory

In some embodiments of the present invention, Memory Management Circuit (MMC) 56 allocates memory resources to the various queues 36 from (i) shared buffer 44, (ii) private reserve memory 48 and (iii) shared reserve memory pool 52. Each of these memory regions has its own characteristics. MMC 53 allocates the resources of each region in accordance with different criteria.


Memory resources of private reserve memory 48 are typically allocated by MMC 56 a-priori to queues 36 (or to at least some of queues 36). The private-reserve memory typically remains allocated to the same queues regardless of actual queue occupancies. This allocation allows queues 36 to respond immediately to traffic bursts and maintain forward progress, even in a congested device. The size of the private-reserve allocation may differ from one queue to another. Further aspects of using a private reserve memory are addressed, for example, in U.S. patent application Ser. No. 17/503,383, entitled “Dynamic Reserve Allocation on Shared-Buffer,” filed Oct. 18, 2021.


Memory resources of shared buffer 44 are typically allocated by MMC 56 adaptively, depending on individual queue occupancy. Typically, a highly active and full queue 36 will be allocated more memory from shared buffer 44, whereas an idle queue will receive little or no shared-buffer allocation. Aspects of shared-buffer allocation are addressed, for example, in U.S. Pat. No. 10,250,530, cited above. A queue will typically begin requesting allocation from shared buffer 44 after it has exhausted its private-reserve allocation (that was pre-allocated from private reserve memory 48).


In some embodiments, MMC 56 defines a maximal occupancy (referred to herein as a “MAX-STATE” threshold), above which a queue 36 is not eligible to receive additional allocations from shared buffer 44. Unless assisted by the disclosed technique, a queue that is ineligible to receive additional allocations from the shared buffer may drop packets (in lossy traffic implementations) or make excessive use of flow control (e.g., credits, in lossless traffic implementations).


The MAX-STATE threshold is adaptive and depends on the overall demand for shared-buffer resources. When the overall demand for shared-buffer resources is high, e.g., when some queues experience congestion, MMC 56 will typically reduce the MAX-STATE threshold. When the overall demand for shared-buffer resources is low, MMC 56 will typically increase the MAX-STATE threshold. This mechanism introduces a certain degree of fairness among the queues in competing for the resources of shared buffer 44.


As explained above, the allocation schemes that use the shared buffer and private reserve memory may be problematic in some scenarios. For example, due to the adaptation of the MAX-STATE threshold, a low-activity queue may become ineligible for shared-buffer allocation when high-activity queues have high demand for memory.


In some embodiments, MMC 56 uses the resources of shared reserve memory pool 52 to avoid such problematic scenarios. In these embodiments, MMC 56 defines an additional range of occupancy levels, above the MAX-STATE threshold, in which a queue is eligible for allocation from shared reserve memory pool 52. The various occupancy thresholds and ranges are depicted in FIG. 2 below.


In an embodiment, MMC 56 identifies a queue 36 that requires additional memory but is too full to be eligible for allocation from shared buffer 44 (i.e., a queue whose occupancy exceeds the MAX-STATE threshold). If the occupancy level of the queue is not too high above the MAX-STATE threshold (e.g., up to a predefined margin above the MAX-SATATE threshold), MMC 52 allocates to the queue memory resources from shared reserve memory pool 52.



FIG. 2 is a diagram that schematically illustrates occupancy thresholds and ranges used for allocating memory resources to a queue 36, in accordance with an embodiment of the present invention. The queue in the figure may be any of queues 36 in network device 20 of FIG. 1 above.


The occupancy of queue 36 is variable and depends on its rate of filling and rate of emptying. At a given time, queue 36 has a current occupancy level denoted 110. The private reserve allocation assigned to queue 36 is marked by a private reserve threshold 114. This level is static, i.e., does not vary depending on the actual occupancy level of the queue.


The current level of the MAX-STATE threshold is denoted 118 in the figure. As explained above, MAX-STATE threshold 118 is varied by MMC 56 depending on the overall demand for allocations from shared buffer 44. When demand is high, MMC 56 decreases MAX-STATE threshold 118, and vice versa. In a typical implementation, the MMC calculates the MAX-STATE threshold for a certain queue based on (i) the total free memory space remaining in the shared buffer and (ii) the number of queues assigned to the shared buffer pool. Thus, as the free memory space in the shared buffer decreases, the potential additional memory that can be allocated to a given queue decreases, as well.


When occupancy level 110 of queue 36 is in the range between private reserve threshold 114 and MAX-STATE threshold 118, queue 36 is eligible to receive additional allocations from shared buffer 44. (When occupancy level 110 is below this range, the queue will still have available resources from its private-reserve allocation. When occupancy level 110 is above this range, the shared-buffer management scheme will prevent allocating additional memory from the shared buffer to the queue.)


A predefined margin 122 is set by MMC 56 as an upper limit for eligibility for allocation from shared reserve memory pool 52. In other words, the range between levels 118 and 122 is predefined, and this constant-size range moves up and down following the variations of MAX-STATE threshold 118. When occupancy level 110 of queue 36 is in this range (between MAX-STATE threshold 118 and margin 122), queue 36 is eligible to receive additional allocations from shared reserve memory pool 52. (When occupancy level 110 is below this range, the queue may request allocations from shared buffer 44. When occupancy level 110 is above this range, the queue is not eligible to any additional allocations, from any source.)



FIG. 3 is a flow chart that schematically illustrates a method, carried out by MMC 56 in network device 20, for allocating memory resources to queues 36, in accordance with an embodiment of the present invention.


The method begins with MMC 56 pre-allocating memory resources of private reserve memory 48 to queues 36, at a private reserve allocation stage 80. At a packet processing stage 84, packet processor 32 receives packets from network 28 via ports 24, processes the packets including queuing them in queues 36, and sends the processed packets back via ports 24 to network 28. At this stage, MMC 56 also performs accounting on the various queues with respect to the shared buffer.


At a requirement checking stage 88, MMC 56 checks whether any of queues 36 requires additional memory resources. In an embodiment, a queue that requires additional memory will send a suitable request to MMC 56. Thus, MMC may identify that a queue requires additional memory by identifying a request from the queue. Alternatively, MMC 56 may proactively determine that a queue requires additional memory, using any suitable method. If no queue requires additional memory (i.e., if all queues have sufficient free memory in their pre-allocated private-reserve allocations), the method loops back to packet processing stage 84.


In response to identifying that a certain queue requires additional memory, MMC 56 checks whether the queue is eligible for allocation from shared buffer 44, at a shared-buffer eligibility checking stage 92. In an embodiment, MMC checks whether the current occupancy (110) of the queue is (i) higher than private-reserve threshold 114 and (ii) lower than MAX-STATE threshold 118. If so, MMC allocates memory resources from shared buffer 44 to the queue, at a shared-buffer allocation stage 96. The method then loops back to packet processing stage 84.


If the outcome of stage 92 is that the queue is not eligible for allocation from shared buffer 44, MMC 56 checks whether the queue is eligible for allocation from shared reserve memory pool 52, at a shared-reserve eligibility checking stage 100. In an embodiment, MMC checks whether the current occupancy (110) of the queue is (i) higher than MAX-STATE threshold 118 and (ii) lower than margin 122. If so, MMC allocates memory from shared reserve memory pool 52 to the queue, at a shared-reserve allocation stage 104. The method then loops back to packet processing stage 84.


If the outcome of stage 100 is that the queue is not eligible for allocation from shared reserve memory pool 52, the method loops back to packet processing stage 84 without allocating any additional memory to the queue.


The method flow of FIG. 3 is an example flow that is depicted purely for the sake of conceptual clarity. In alternative embodiments, the disclosed techniques can be implemented using any other suitable flow.


For example, the method can be divided into two separate processes that are executed in parallel. One process is a shared-buffer management process, which allocates memory from shared buffer 44 and adapts MAX-STATE threshold 118. The other process is a shared-reserve management process, which allocates memory from shared reserve memory pool 52. The variations in MAX-STATE threshold 118 affect both processes, since they affect both eligibility ranges (eligibility for shared-buffer allocation, and eligibility for shared-reserve allocation). As another example, the method can be used with a shared-buffer mechanism that uses a static (i.e., non-variable) MAX-STATE threshold.


As yet another example, MMC 56 may define a hierarchy of shared-reserve memory pools 52. The hierarchy may comprise, for example, a “global” shared-reserve pool, plus shard-reserve pools for receive (ingress) queues, for multicast (MC) queues, for transmit (egress)_queues, and the like. When a certain shared-reserve pool is exhausted, the MMC may allocate additional memory to this pool from the global shared-reserve pool.


The description up to this point referred to a single shared reserve memory pool 52, whose resources may be allocated to any of queues 36 in network device 20. In some embodiments, MMC 56 maintains multiple separate shared reserve memory pools 52 in memory 40. Each shared reserve memory pool 52 is assigned to a respective subset of queues 36. In other words, in these embodiments MMC 56 allocates to a given queue only shared-reserve allocations from the pool 52 of the subset of queues to which the given queue belongs. For example, MMC 52 may maintain a separate pool 52 for the receive queues of each port 48, and/or to the transmit queues of each port. In an example embodiment, a given port 24 is assigned a set of queues 36 corresponding to respective priority groups, and MMC 56 maintains a shared reserve memory pool 52 for this set of queues. As another example, MMC 52 may maintain a separate pool 52 for any other group of (receive or transmit) queues.


In an example embodiment, MMC 56 carries out the disclosed technique by executing the pseudo-code sections below. The present example considers a receive queue Rq[ ] of a certain port 24. Several priority groups pg[ ] are defined for this receive queue, and each priority group is assigned a separate queue 36. These queues are thus denoted Rq[ ],pg[ ].


Attributes denoted Rq[ ].SharedReserve and Rq[ ].SharedReserveTH are defined for Rq[ ]. Rq[ ].SharedReserve is the overall size of the allocation from shared reserve memory pool 52, allocated to the various queues 36 belonging to Rq[ ]. Rq[ ].SharedReserveTH denotes the size of the shared reserve pool. In addition, attributes denoted Rq.Pg[ ].PrivateReserve, Rq.Pg[ ].PrivateReserveTH and Rq.Pg[ ].ReserveEligibleTH are defined for each priority group Pg[ ] of receive queue Rq[ ]. Rq.Pg[ ].PrivateReserve is the size of the private-reserve allocation that was allocated to queue Rq.Pg[ ]. Rq.Pg[ ].PrivateReserveTH denotes the upper threshold for allocating memory to the queue from the private reserve pool. Rq.Pg[ ].ReserveEligibleTH denotes the margin (range) above the MAX-TH threshold within which the queue is eligible for allocation from the shared reserve pool.


In this example, the allocation process is performed in accordance with the following pseudo-code:

    • 1. If Inc. & Rq.pg[ ].Occupancy<PrivateReserve[ ]→Done, HaveReserve for Rg.Pg[ ]
    • 2. If Inc. & (Rq.Pg[ ].PrivateReserveTH[ ]<(Rq.pg[ ].Occupancy)<(Rq.Pg[ ].PrivateReserveTH[ ]+Rq.Pg[ ].ReserveEligibleTH)) & (Published Rq.pg[ ]. AboveMax)
      • a. If Rq[ ]. SharedReserve<Rq[ ]. SharedReserveTH
        • i. Increase Rq[ ].SharedReserve by Delta
        • ii. Increase Rq.Pg[ ].PrivateReserve[ ] by Delta
      • b. If Rq[ ].SharedReserve>Rq[ ].SharedReserveTH→Done, NoReserve for Rg.Pg[ ]
    • 3. If Dec. & Rq.pg[ ] & (Rq.Pg[ ].PrivateReserve[ ]>Rq.Pg[ ].PrivateReserveTH[ ])
      • a. Decrease Rq.Pp[ ].PrivateReserve[ ] by Delta
      • b. Decrease Rq[ ].SharedReserve by Delta


In parallel, an additional “crawler process” is performed according to the following pseudo-code:

    • 4. If (˜Rq.pg[ ].AboveMax) & (Rq.Pg[ ].PrivateReserve[ ]>Rq.Pg[ ].PrivateReserveTH[ ])
      • a. Decrease Rq.Pp[ ].PrivateReserve[ ] by Delta
      • b. Decrease Rq[ ].SharedReserve by Delta


In the pseudo-code above, Rq.pg[ ].Occupancy denotes the current occupancy (110) of the queue Rq.pg[ ], Inc. denotes a detected increase in the current occupancy, and Dec. denotes a detected decrease in the current queue occupancy. HaveReserve for Rg.Pg[ ] is an outcome indicating that the queue currently has sufficient reserve memory, whereas NoReserve for Rg.Pg[ ] is an outcome indicating that the queue does not have sufficient reserve memory.


Published Rq.pg[ ].AboveMax is a Boolean attribute that indicates whether the current occupancy of queue Rq.pg[ ] is above MAX-STATE threshold 118. Delta denotes the increment by which MMC 56 increases or decreases the reserve allocations.


It will be appreciated that the embodiments described above are cited by way of example, and that the present invention is not limited to what has been particularly shown and described hereinabove. Rather, the scope of the present invention includes both combinations and sub-combinations of the various features described hereinabove, as well as variations and modifications thereof which would occur to persons skilled in the art upon reading the foregoing description and which are not disclosed in the prior art. Documents incorporated by reference in the present patent application are to be considered an integral part of the application except that to the extent any terms are defined in these incorporated documents in a manner that conflicts with the definitions made explicitly or implicitly in the present specification, only the definitions in the present specification should be considered.

Claims
  • 1. A network device, comprising: a plurality of ports, to communicate packets over a network;a packet processor, to process the packets using a plurality of queues, which belong to subsets of queues; anda memory management circuit, to: maintain a shared buffer in a memory, and adaptively allocate memory resources from the shared buffer to the queues;maintain in the memory, in addition to the shared buffer, multiple separate shared-reserve memory pools, each pool for use by a respective subset from the subsets of the queues;identify a queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pools; andallocate memory resources to the identified queue from the shared-reserve memory pool corresponding to a subset from the subsets of queues, including the identified queue.
  • 2. The network device according to claim 1, wherein the memory management circuit is to identify that the queue requires additional memory resources by receiving a request from the queue.
  • 3. The network device according to claim 1, wherein the memory management circuit is to identify that the queue is not eligible for additional allocation from the shared buffer, by identifying that the occupancy of the queue exceeds a MAX threshold set for the queue.
  • 4. The network device according to claim 3, wherein the memory management circuit is to verify that the occupancy of the queue meets the eligibility condition for the shared-reserve memory pools, by verifying that the occupancy is no more than a defined margin above the MAX threshold.
  • 5. The network device according to claim 1, wherein the memory management circuit is to pre-allocate to each of the plurality of queues a private-reserve memory resource, irrespective of the shared buffer and the shared-reserve memory pools.
  • 6. The network device according to claim 1, wherein the queues in each subset are associated with a same port of the network device.
  • 7. The network device according to claim 1, wherein the multiple separate shared-reserve memory pools include separate pools for the receive queues of each port and for the transmit queues of each port.
  • 8. A memory management circuit in a network device, the memory management circuit comprising: one or more interfaces, to communicate with a plurality of queues of the network device that process packets, and with a memory; andcircuitry, to: maintain a shared buffer in the memory, and adaptively allocate memory resources from the shared buffer to the queues;maintain in the memory, in addition to the shared buffer, multiple separate shared-reserve memory pools, each pool for use by a respective subset from the subsets of the queues;identify a queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pools; andallocate memory resources to the identified queue from the shared-reserve memory pool corresponding to a subset from the subsets of queues, including the identified queue.
  • 9. A method for communication in a network device, the method comprising: communicating packets over a network, and processing the packets using a plurality of queues;maintaining a shared buffer and multiple separate shared-reserve memory pools in a memory,wherein the shared-reserve memory pools are assigned to respective subsets of the queues; andrepeatedly performing: identifying a queue that (i) requires additional memory resources;check whether the identified queue (ii) is eligible for additional allocation from the shared buffer,if the identified queue is eligible for additional allocation from the shared buffer, allocating additional memory resources to the identified queue from the shared buffer;if the identified queue is not eligible for additional allocation from the shared buffer, check whether the identified queue (iii) meets an eligibility condition for the shared-reserve memory pool corresponding to a subset to which the identified queue belongs; andif the eligibility condition for the shared-reserve memory pool is met, allocating memory resources to the identified queue from the shared-reserve memory pool corresponding to the subset to which the identified queue belongs.
  • 10. The method according to claim 9, wherein identifying that the queue requires additional memory resources comprises receiving a request from the queue.
  • 11. The method according to claim 9, wherein identifying that the queue is not eligible for additional allocation from the shared buffer comprises identifying that the occupancy of the queue exceeds a MAX threshold set for the queue.
  • 12. The method according to claim 11, wherein identifying that the occupancy of the queue meets the eligibility condition for the shared-reserve memory peel pools comprises verifying that the occupancy is no more than a defined margin above the MAX threshold.
  • 13. The method according to claim 9, wherein further comprising pre-allocate to each of the plurality of queues a private-reserve memory resource, irrespective of the shared buffer and the shared-reserve memory pools.
  • 14. The method according to claim 9, wherein the queues in each subset are associated with a same port of the network device.
  • 15. A method for memory management in a network device, the method comprising: communicating with a plurality of queues of the network device that process packets, and with a memory;maintaining a shared buffer in the memory, and adaptively allocating memory resources from the shared buffer to the queues;maintaining in the memory, in addition to the shared buffer, multiple separate shared-reserve memory pools, each pool for use by a respective subset from the subsets of the queues;identifying a queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) meets an eligibility condition for the shared-reserve memory pools; andallocating memory resources to the identified queue from the shared-reserve memory pool corresponding to a subset from the subsets of queues, including the identified queue.
  • 16. A method for memory management in a network device that processes packets using queues, the method comprising: maintaining in a memory (i) a shared buffer and (ii) multiple separate shared-reserve memory pools each pool for use by a different respective subset of the queues;identifying a queue that (i) requires additional memory resources, (ii) is not eligible for additional allocation from the shared buffer, and (iii) is eligible for allocation from the shared-reserve memory pools; andallocating memory resources to the identified queue from the shared-reserve memory pool corresponding to a subset from the subsets of queues, including the identified queue.
US Referenced Citations (299)
Number Name Date Kind
5014265 Hahne et al. May 1991 A
5235642 Wobber et al. Oct 1993 A
5367520 Cordell Nov 1994 A
5541912 Choudhury et al. Jul 1996 A
5574885 Denzel et al. Nov 1996 A
5790522 Fichou et al. Aug 1998 A
5917947 Ishida et al. Jun 1999 A
6108713 Sambamurthy et al. Aug 2000 A
6154446 Kadambi et al. Nov 2000 A
6160814 Ren et al. Dec 2000 A
6169748 Barbas et al. Jan 2001 B1
6178448 Gray et al. Jan 2001 B1
6219728 Yin Apr 2001 B1
6324165 Fan et al. Nov 2001 B1
6347337 Shah et al. Feb 2002 B1
6456590 Ren et al. Nov 2002 B1
6490248 Shimojo Dec 2002 B1
6535963 Rivers Mar 2003 B1
6539024 Janoska et al. Mar 2003 B1
6594263 Martinsson et al. Jul 2003 B1
6606666 Bell et al. Aug 2003 B1
6633395 Tuchitoi et al. Oct 2003 B1
6678277 Wils et al. Jan 2004 B1
6687254 Ho et al. Feb 2004 B1
6771654 Sang et al. Aug 2004 B1
6859435 Lee et al. Feb 2005 B1
6895015 Chiang et al. May 2005 B1
6922408 Bloch et al. Jul 2005 B2
6973555 Fujiwara et al. Dec 2005 B2
6993032 Dammann et al. Jan 2006 B1
7027457 Chiussi et al. Apr 2006 B1
7068822 Scott Jun 2006 B2
7088713 Battle et al. Aug 2006 B2
7131125 Modelski et al. Oct 2006 B2
7136381 Battle et al. Nov 2006 B2
7190667 Susnow et al. Mar 2007 B2
7215641 Bechtolsheim et al. May 2007 B1
7321553 Prasad et al. Jan 2008 B2
7324442 Pan et al. Jan 2008 B1
7327749 Mott Feb 2008 B1
7330426 Berzosa et al. Feb 2008 B2
7334065 Rose et al. Feb 2008 B1
7346059 Gamer et al. Mar 2008 B1
7529245 Muller et al. May 2009 B1
7590058 Cherchali et al. Sep 2009 B1
7609636 Mott Oct 2009 B1
7630306 Chuang Dec 2009 B2
7680139 Jones et al. Mar 2010 B1
7738454 Panwar et al. Jun 2010 B1
7747086 Hobbs et al. Jun 2010 B1
7773621 Jensen Aug 2010 B2
7773622 Schmidt et al. Aug 2010 B2
7778168 Rodgers et al. Aug 2010 B1
7813348 Gupta et al. Oct 2010 B1
7821939 Decusatis et al. Oct 2010 B2
7835380 Aloni et al. Nov 2010 B1
7853738 Pothireddy et al. Dec 2010 B2
7856026 Finan et al. Dec 2010 B1
7872973 Sterne et al. Jan 2011 B2
7894343 Chao et al. Feb 2011 B2
7924708 Spink Apr 2011 B2
7936770 Frattura et al. May 2011 B1
8060644 Michailidis et al. Nov 2011 B1
8078743 Sharp et al. Dec 2011 B2
8125902 Rochon et al. Feb 2012 B2
8149710 Bergamasco et al. Apr 2012 B2
8270295 Kendall et al. Sep 2012 B2
8274971 Battle et al. Sep 2012 B2
8352648 Puranik Jan 2013 B1
8345548 Gusat et al. Mar 2013 B2
8473693 Muppalaneni et al. Jun 2013 B1
8478811 Garg et al. Jul 2013 B2
8565092 Arumilli et al. Oct 2013 B2
8576715 Bloch et al. Nov 2013 B2
8589587 Michailidis et al. Nov 2013 B1
8599868 Sindhu et al. Dec 2013 B2
8630294 Keen et al. Jan 2014 B1
8635386 Takahashi Jan 2014 B2
8656188 Goodwill et al. Feb 2014 B2
8699491 Koren et al. Apr 2014 B2
8705349 Bloch et al. Apr 2014 B2
8730982 Wu et al. May 2014 B2
8738860 Griffin et al. May 2014 B1
8762595 Muller et al. Jun 2014 B1
8767561 Gnanasekaran et al. Jul 2014 B2
8811183 Anand et al. Aug 2014 B1
8838882 Quong Sep 2014 B1
8879396 Guay et al. Nov 2014 B2
8923337 Singh Dec 2014 B2
8989017 Naouri et al. Feb 2015 B2
8995265 Basso et al. Mar 2015 B2
9014006 Haramaty et al. Apr 2015 B2
9130885 Bukspan et al. Sep 2015 B1
9210095 Basso et al. Dec 2015 B2
9325619 Guay et al. Apr 2016 B2
9325641 Haramaty et al. Apr 2016 B2
9356868 Tabatabaee et al. May 2016 B2
9426085 Anand et al. Aug 2016 B1
9497125 Raindel et al. Nov 2016 B2
9563361 Asiot et al. Feb 2017 B1
9584429 Haramaty et al. Feb 2017 B2
9648148 Rimmer et al. May 2017 B2
9699095 Elias et al. Jul 2017 B2
9742683 Vanini Aug 2017 B1
9742702 Bukspan et al. Aug 2017 B1
9762491 Gafni et al. Sep 2017 B2
9807024 Shpiner et al. Oct 2017 B2
10050896 Yang et al. Aug 2018 B2
10069701 Elias Sep 2018 B2
10069748 Shpiner et al. Sep 2018 B2
10084716 Gafni Sep 2018 B2
10205683 Elias et al. Feb 2019 B2
10237376 Crupnicoff et al. Mar 2019 B2
10250530 Aibester et al. Apr 2019 B2
10387074 Kriss Aug 2019 B2
10530846 Jung Jan 2020 B2
10728156 Kwan et al. Jul 2020 B2
10778568 Levy et al. Sep 2020 B2
10951549 Elias et al. Mar 2021 B2
10999221 Gafni et al. May 2021 B2
11099891 Baldocchi Aug 2021 B2
11218413 Shpigelman et al. Jan 2022 B2
11296988 Shpigelman et al. Apr 2022 B2
11470010 Mula et al. Oct 2022 B2
11563695 Mitulal Jan 2023 B2
20020012340 Kalkunte et al. Jan 2002 A1
20020019916 Henrion Feb 2002 A1
20020027908 Kalkunte et al. Mar 2002 A1
20020039350 Wang et al. Apr 2002 A1
20020055993 Shah et al. May 2002 A1
20020067695 Skarpness et al. Jun 2002 A1
20020087723 Williams et al. Jul 2002 A1
20020176430 Sangha et al. Nov 2002 A1
20020176432 Courtney et al. Nov 2002 A1
20020191559 Chen et al. Dec 2002 A1
20030016628 Kadambi et al. Jan 2003 A1
20030016697 Jordan Jan 2003 A1
20030043828 Wang et al. Mar 2003 A1
20030048792 Xu et al. Mar 2003 A1
20030065812 Beier et al. Apr 2003 A1
20030076849 Morgan et al. Apr 2003 A1
20030095560 Arita et al. May 2003 A1
20030108010 Kim et al. Jun 2003 A1
20030112817 Woo et al. Jun 2003 A1
20030117958 Nation et al. Jun 2003 A1
20030118016 Kalkunte et al. Jun 2003 A1
20030120894 Wang et al. Jun 2003 A1
20030123392 Ruutu et al. Jul 2003 A1
20030137939 Dunning et al. Jul 2003 A1
20030179765 Goetzinger et al. Sep 2003 A1
20030198231 Kalkunte et al. Oct 2003 A1
20030198241 Putcha et al. Oct 2003 A1
20030200330 Oelke et al. Oct 2003 A1
20030217184 Nair Nov 2003 A1
20030222860 Yamaura Dec 2003 A1
20030223368 Allen et al. Dec 2003 A1
20030223435 Gil Dec 2003 A1
20040008714 Jones Jan 2004 A1
20040027989 Martin et al. Feb 2004 A1
20040037558 Beshasi Feb 2004 A1
20040066785 He et al. Apr 2004 A1
20040081090 Hara et al. Apr 2004 A1
20040202169 Mukouyama et al. Oct 2004 A1
20050053077 Blanc et al. Mar 2005 A1
20050063370 Beshai et al. Mar 2005 A1
20050076138 Sterne Apr 2005 A1
20050094643 Wang et al. May 2005 A1
20050125424 Herriott et al. Jun 2005 A1
20050129033 Gordy et al. Jun 2005 A1
20050135356 Muthukrishnan et al. Jun 2005 A1
20050169172 Wang et al. Aug 2005 A1
20050204103 Dennison Sep 2005 A1
20050216822 Kyusojin et al. Sep 2005 A1
20050223139 Wagh et al. Oct 2005 A1
20050226156 Keating et al. Oct 2005 A1
20050228900 Stuart et al. Oct 2005 A1
20050259574 Figueira et al. Nov 2005 A1
20060008803 Brunner et al. Jan 2006 A1
20060034172 Morton Feb 2006 A1
20060087989 Gai et al. Apr 2006 A1
20060088036 De Prezzo Apr 2006 A1
20060092837 Kwan et al. May 2006 A1
20060092842 Beukema et al. May 2006 A1
20060092845 Kwan et al. May 2006 A1
20060095609 Radhakrishnan et al. May 2006 A1
20060155938 Cummings et al. Jul 2006 A1
20060182112 Battle et al. Aug 2006 A1
20060251109 Muller et al. Nov 2006 A1
20070015525 Beming et al. Jan 2007 A1
20070019553 Sagfors et al. Jan 2007 A1
20070025242 Tsang Feb 2007 A1
20070041385 Sali et al. Feb 2007 A1
20070053350 Spink et al. Mar 2007 A1
20070097257 El-Maleh et al. May 2007 A1
20070104102 Opsasnick May 2007 A1
20070104211 Opsasnick May 2007 A1
20070147292 Van Ewijk et al. Jun 2007 A1
20070201499 Kapoor et al. Aug 2007 A1
20070274215 Gusat et al. Nov 2007 A1
20070291644 Roberts et al. Dec 2007 A1
20070297453 Niinomi Dec 2007 A1
20080037420 Tang et al. Feb 2008 A1
20080144670 Gooossens et al. Jun 2008 A1
20080175146 Van Leekwuck et al. Jul 2008 A1
20080192764 Arefi et al. Aug 2008 A1
20080259936 Hussain et al. Oct 2008 A1
20090003212 Kwan et al. Jan 2009 A1
20090010162 Bergamasco et al. Jan 2009 A1
20090073881 Cui Mar 2009 A1
20090122702 Chew et al. May 2009 A1
20090161684 Voruganti et al. Jun 2009 A1
20090182944 Comparan et al. Jul 2009 A1
20090207848 Kwan et al. Aug 2009 A1
20100057953 Kim et al. Mar 2010 A1
20100061238 Godbole et al. Mar 2010 A1
20100061390 Godbole et al. Mar 2010 A1
20100067539 Lin et al. Mar 2010 A1
20100088756 Balakrishnan et al. Apr 2010 A1
20100100670 Jeddeloh Apr 2010 A1
20100165842 Wang et al. Jul 2010 A1
20100220742 Brewer et al. Sep 2010 A1
20100322076 Goel Dec 2010 A1
20100325318 Desoli et al. Dec 2010 A1
20110058571 Bloch et al. Mar 2011 A1
20110286468 Tomonaga et al. Nov 2011 A1
20120072635 Yoshida et al. Mar 2012 A1
20120106562 Laor et al. May 2012 A1
20120106567 Hanes et al. May 2012 A1
20120144064 Parker et al. Jun 2012 A1
20120155264 Sharma et al. Jun 2012 A1
20130014118 Jones Jan 2013 A1
20130021354 Van Eerd et al. Jan 2013 A1
20130039178 Chen et al. Feb 2013 A1
20130077489 Bloch et al. Mar 2013 A1
20130212296 Goel et al. Aug 2013 A1
20130239119 Garg et al. Sep 2013 A1
20130250757 Tabatabaee et al. Sep 2013 A1
20130250762 Assarpour Sep 2013 A1
20130275631 Magro et al. Oct 2013 A1
20130286834 Lee Oct 2013 A1
20130305250 Durant Nov 2013 A1
20140036930 Lih et al. Feb 2014 A1
20140095745 Kawahara Apr 2014 A1
20140133314 Mathews et al. May 2014 A1
20140140206 Hendel et al. May 2014 A1
20140155043 Gell et al. Jun 2014 A1
20140192646 Mir et al. Jul 2014 A1
20140204742 Pandit Jul 2014 A1
20140269274 Banavalikar et al. Sep 2014 A1
20140269324 Tietz et al. Sep 2014 A1
20140269711 Ravid et al. Sep 2014 A1
20140286349 Kitada Sep 2014 A1
20140289568 Koyoma et al. Sep 2014 A1
20140310354 Fountain et al. Oct 2014 A1
20150026309 Radcliffe et al. Jan 2015 A1
20150026361 Matthews et al. Jan 2015 A1
20150058857 Sandstorm Feb 2015 A1
20150095568 Lamb et al. Apr 2015 A1
20150103667 Elias Apr 2015 A1
20150124611 Attar et al. May 2015 A1
20150127797 Attar et al. May 2015 A1
20150180782 Rimmer et al. Jun 2015 A1
20150180790 Rimmer et al. Jun 2015 A1
20150200866 Pope et al. Jul 2015 A1
20150371607 Holland et al. Dec 2015 A1
20150381505 Sundararaman et al. Dec 2015 A1
20160135076 Grinshpun et al. May 2016 A1
20160173383 Liu et al. Jun 2016 A1
20160191392 Liu Jun 2016 A1
20160226951 Talpey et al. Aug 2016 A1
20160294715 Raindel et al. Oct 2016 A1
20160337257 Yifrach et al. Nov 2016 A1
20170118108 Avci et al. Apr 2017 A1
20170142020 Sundararaman et al. May 2017 A1
20170180261 Ma et al. Jun 2017 A1
20170187641 Lundqvist et al. Jun 2017 A1
20170201469 Elias Jul 2017 A1
20170295112 Cheng et al. Oct 2017 A1
20170337010 Kriss Nov 2017 A1
20170373989 Gafni et al. Dec 2017 A1
20180063030 Mitulal Mar 2018 A1
20180063038 Ghanwani et al. Mar 2018 A1
20180091388 Levy et al. Mar 2018 A1
20180115498 Sundaram et al. Apr 2018 A1
20180205653 Wang et al. Jul 2018 A1
20180241677 Srebro et al. Aug 2018 A1
20180278550 Rosen et al. Sep 2018 A1
20190042443 Maciejewski et al. Feb 2019 A1
20190042513 Fleming, Jr. et al. Feb 2019 A1
20190044874 Zhang et al. Feb 2019 A1
20200177521 Blumrich et al. Jun 2020 A1
20200280518 Lee et al. Sep 2020 A1
20210006502 Zhou et al. Jan 2021 A1
20210344600 Urman et al. Nov 2021 A1
20210377074 Yoshida et al. Dec 2021 A1
20220045972 Aibester et al. Feb 2022 A1
20220231933 Mannor et al. Jul 2022 A1
20230068902 Seely Mar 2023 A1
20230137826 Ajami et al. May 2023 A1
Foreign Referenced Citations (11)
Number Date Country
1379569 Nov 2002 CN
0706298 Apr 1996 EP
1698976 Sep 2006 EP
1720295 Nov 2006 EP
2466476 Jun 2012 EP
03024033 Mar 2003 WO
2009107089 Sep 2009 WO
2013136355 Sep 2013 WO
2013180691 Dec 2013 WO
2018044426 Mar 2018 WO
2018106868 Jun 2018 WO
Non-Patent Literature Citations (36)
Entry
U.S. Appl. No. 17/516,731 Office Action dated Jun. 16, 2023.
U.S. Appl. No. 17/503,383 Office Action dated Jun. 8, 2023.
U.S. Appl. No. 17/503,383 Office Action dated Aug. 17, 2023.
Gran et al., “Congestion Management in Lossless Interconnection Networks”, Submitted to the Faculty of Mathematics and Natural Sciences at the University of Oslo in partial fulfillment of the requirements for the degree Philosophiae Doctor, pp. 1-156, Sep. 2013.
Pfister et al., “Hot Spot Contention and Combining in Multistage Interconnect Networks”, IEEE Transactions on Computers, vol. C-34, pp. 943-948, Oct. 1985.
Zhu et al.,“Congestion control for large-scale RDMA deployments”, SIGCOMM, ACM, pp. 523-536, Aug. 17-21, 2015.
CISCO Systems, Inc., “Advantage Series White Paper Smart Buffering”, pp. 1-16, year 2016.
Hoeiland-Joergensen et al., “The FlowQueue-CoDel Packet Scheduler and Active Queue Management Algorithm”, Internet Engineering Task Force (IETF) as draft-ietf-aqm-fq-codel-06, pp. 1-23, Mar. 18, 2016.
Hahne et al., “Dynamic Queue Length Thresholds for Multiple Loss Priorities”, IEEE/ACM Transactions on Networking, vol. 10, No. 3, pp. 368-380, Jun. 2002.
Choudhury et al., “Dynamic Queue Length Thresholds for Shared-Memory Packet Switches”, IEEE/ACM Transactions Networking, vol. 6, issue 2, pp. 130-140, Apr. 1998.
HUAWEI Technologies Co., Ltd., Baidu, Inc., “The Lossless Network for Data Centers,” Revision 1.0, pp. 1-15, Nov. 7, 2017.
IEEE Standard 802.1Q™-2005, “IEEE Standard for Local and metropolitan area networks Virtual Bridged Local Area Networks”, pp. 1-303, May 19, 2006.
Infiniband™ Architecture, Specification vol. 1, Release 1.2.1, Chapter 12, pp. 657-716, Nov. 2007.
IEEE Std 802.3, Standard for Information Technology—Telecommunications and information exchange between systems—Local and metropolitan area networks—Specific requirements; Part 3: Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications Corrigendum 1: Timing Considerations for PAUSE Operation, Annex 31B (MAC Control PAUSE operation), pp. 763-772, year 2005.
IEEE Std 802.1Qbb., IEEE Standard for Local and metropolitan area networks—“Media Access Control (MAC) Bridges and Virtual Bridged Local Area Networks—Amendment 17: Priority-based Flow Control”, pp. 1-40, Sep. 30, 2011.
CISCO Systems, Inc., “Priority Flow Control: Build Reliable Layer 2 Infrastructure”, pp. 1-8 pages, year 2015.
Aibester et al., U.S. Appl. No. 17/955,591, filed Sep. 29, 2022.
Mula et al., U.S. Appl. No. 17/902,936, filed Sep. 5, 2022.
U.S. Appl. No. 17/503,383 Office Action dated Nov. 4, 2022.
INFINIBAND Architecture Release 1.2.1, vol. 1, General specifications, Chapter 7.9, pp. 212-216, Nov. 2007.
Raatikainen, P., “ATM Switches—Switching Technology S38.3165”, Switching Technology, L8-1, 34 pages, year 2006 (http://www.netlab.hut.fi/opetus/s383165).
Fahmy, S., “A Survey of ATM Switching Techniques”, Department of Computer and Information Science, The Ohio State University, USA, pp. 1-22m Aug. 21, 1995 downloaded from http://www.cs.purdue.edu/homes/fahmy/cis788.08Q/atmswitch.html.
Minkenberg et al., “Multistage Interconnection Networks for Data Centers; Bidirectional Fat Tree Construction and Routing for IEEE 802.1au”, IBM Research GmbH, Zurich, Switzerland, pp. 1-9, Jul. 2, 2007.
Bwalya et al., “Performance Evaluation of Buffer Size for Access Networks in First Generation Optical Networks”, International Journal of Internet of Things, vol. 6, issue 3, pp. 98-105, year 2017.
Ramakrishnan et al., “The Addition of Explicit Congestion Notification (ECN) to IP,” Request for Comments (RFC) 3168, The Internet Society, pp. 1-63, Sep. 2001.
Spitulski, U.S. Appl. No. 17/840,371, filed Jun. 14, 2022.
Kadosh et al., U.S. Appl. No. 17/348,801, filed Jun. 16, 2021.
U.S. Appl. No. 17/175,716 Office Action dated Jun. 22, 2022.
“Understanding CoS Priority Group and Queue Guaranteed Minimum Bandwidth,” Traffic Management User Guide (QFX Series and EX4600 Switches), juniper.net, pp. 1-4, Feb. 1, 2021, as downloaded from https://www.juniper.net/documentation/us/en/software/junos/traffic-mgmt-qfx/topics/concept/cos-qfx-series-guaranteed-bandwidth-understanding.html.
“QoS Modular Qos Command—Line Interface Configuration Guide, Cisco IOS Xe Fuji 16.8x,” Cisco, pp. 1-25, Apr. 20, 2018, as downloaded from https://www.cisco.com/c/en/us/td/docs/ios-XML/ios/qos_mqc/configuration/xe-16-8/qos-mqc-xe-16-8-book/qos-limits-wred.html.
Mula et al., U.S. Appl. No. 17/175,716, filed Feb. 15, 2021.
Aibester et al., U.S. Appl. No. 17/516,731, filed Nov. 2, 2021.
Aibester et al., U.S. Appl. No. 17/503,383, filed Oct. 18, 2021.
CN Application # 2021108967450 Office Action dated Sep. 8, 2023.
U.S. Appl. No. 17/516,731 Office Action dated Oct. 6, 2023.
U.S. Appl. No. 17/902,936 Office Action dated Oct. 25, 2023.
Related Publications (1)
Number Date Country
20230283575 A1 Sep 2023 US