Brocade Communications Systems, Inc.

Avago Technologies International Sales Pte. Limited

Badrinath Kollu

Sathish Kumar Gnanasekaran

Daniel Ji Yong Park Chung

David Aaron Skirmont

Ganesh Kumar Murugesan

Saumitra Buragohain

Balakrishna Wusirika

Nhan Trong Pham

Kyuh Kim

Kung-Ling Ko

Vineet M Abraham

Dennis Hideo Makishima

Subramanian Lakshmanan

Rishi Amit Sinha

Vinodh Ravindran

Howard Lynn Johnson

Chao Zhang

Vidya Sagar Ravipati

Srikara Subramanyan

Nadeem Kattangere

Vidya Renganarayanan Vineet

Blayne Rawsky

Richard L Hammons

Daniel Y Chung

Ramkumar Vadivelu

Ponpandiaraj Rajarathinam

Dan Norbert Retter

Varghese Kallarackal

Li Zhao

Xupei Liang

Bei Wang

Hung Quoc Nguyen

Prasanta Kumar David Bhuyan

Amit Kumar Tyagi

Sharanappa Ijeri

Pushpanathan Chidambaram

Sathish K Gnanasekaran

Prasanta Kumar David Bhuya

Jianqiang Zhou

Tarun Bhatia

Leifang Hu

A system and method for non-disruptive route change. The method includes receiving a first data transmission frame associated with a first exchange at a switch, transmitting the first data transmission frame through a first route based on a route lookup, determining, by a traffic optimization circuit, an improved second route, determining if the first exchange includes any other data transmission frames, and transmitting a first data transmission frame associated with a second exchange through the second route in response to determining that the first exchange does not include any other data transmission frames.

Non-disruptive route change in fibre channel network

A Layer 2 network switch is partitionable into a plurality of switch fabrics. The single-chassis switch is partitionable into a plurality of logical switches, each associated with one of the virtual fabrics. The logical switches behave as complete and self-contained switches. A logical switch fabric can span multiple single-chassis switch chassis. Logical switches are connected by inter-switch links that can be either dedicated single-chassis links or logical links. An extended inter-switch link can be used to transport traffic for one or more logical inter-switch links. Physical ports of the chassis are assigned to logical switches and are managed by the logical switch. Legacy switches that are not partitionable into logical switches can serve as transit switches between two logical switches.

Partitioning of switches and fabrics into logical switches and fabrics

A hypervisor preferably provides VM (virtual machine) identification, priority and LUN/LBA range information to the HBA (hoist bus adapter) when a VM is created. Alternatively, the HBA can determine that a LUN/LBA range is new and request VM identity, priority and LUN/LBA range from the hypervisor. The HBA creates a table containing the VM identification, priority and LUN/LBA range. The HBA then detects operations directed to the LUN/LBA range and does a lookup to determine VM identification and priority. VM identification and priority are then mapped into a field in a frame using a unique identifier. The unique identifier can be placed using reserved bits on the existing Fiber Channel (FC) header or can use bits in an additional header, such as a modified IFR header or an optional device header. The VM identification aware HBAs register with the NS.

Method and apparatus for registering and storing virtual machine unique information capabilities

A switch detecting a slow drain situation and providing a slow drain primitive to the slow draining device, such as a storage unit. The slow draining device detects the slow drain primitive and provides a throttling message to the relevant sources of frames being received by the slow draining device. The use of a primitive instead of a frame allows the slow condition notification to be provided even when there is no available credit for sending a frame.

Use of primitives to notify of slow drain condition

A hypervisor preferably provides VM (virtual machine) identification, priority and LUN/LBA range information to the HBA (host bus adapter) when a VM is created. Alternatively, the HBA can determine that a LUN/LBA range is new and request VM identity, priority and LUN/LBA range from the hypervisor. The HBA creates a table containing the VM identification, priority and LUN/LBA range. The HBA then detects operations directed to the LUN/LBA range and does a lookup to determine VM identification and priority. VM identification and priority are then mapped into a field in a frame using a unique identifier. The unique identifier can be placed using reserved bits on the existing Fiber Channel (FC) header or can use bits in an additional header, such as a modified IFR header or an optional device header. The VM identification aware HBAs register with the NS.

Method and apparatus for providing virtual machine information to a network interface

An analytics and diagnostic node according to the present disclosure monitors oversubscription and determines flow metrics by receiving mirror command frames from one or more switching nodes. The mirror command frames could correspond to a multiple flows traversing over a connection within a network. The analytics and diagnostic node collects at least one latency metric for each of the flows using timestamps found within the mirror command frames. Based on the latency metrics and timestamps, the analytic diagnostic node determines an average data rate for each of the flows. The analytics and diagnostic node also computes the cumulative data rates corresponding to different bucket intervals based on the average data rates. To detect oversubscription, the analytics and diagnostic node compares the cumulative data rates with one or more oversubscription rules.

Command response and completion determination

A host connected to a switch using a PCI Express (PCIe) link. At the switch, the packets are received and routed as appropriate and provided to a conventional switch network port for egress. The conventional networking hardware on the host is substantially moved to the port at the switch, with various software portions retained as a driver on the host. This saves cost and space and reduces latency significantly. As networking protocols have multiple threads or flows, these flows can correlate to PCIe queues, easing QoS handling. The data provided over the PCIe link is essentially just the payload of the packet, so sending the packet from the switch as a different protocol just requires doing the protocol specific wrapping. In some embodiments, this use of different protocols can be done dynamically, allowing the bandwidth of the PCIe link to be shared between various protocols.

PCI express connected network switch

A diagnostic testing utility is used to perform single link diagnostics tests including an electrical loopback test, an optical loopback test, a link traffic test, and a link distance measurement test. To perform the diagnostic tests, two ports at each end of a link are identified and then statically configured by a user. The ports will be configured as D_Ports and as such will be isolated from the fabric with no data traffic flowing through them. The ports will then be used to send test frames to perform the diagnostic tests.

Diagnostic port for inter-switch link testing in electrical, optical and remote loopback modes

Handling of ELS REQ and RSP packets that contain addresses in the payload is shifted to the edge fabric switches connected to the node devices issuing and receiving the ELS REQ packet, the ingress and egress switches. This allows the ELS REQ and RSP packet payload address modification operations to be removed from the tasks handled by the router processor. As this removes a processing burden from the router processors, those router processors are free to handle other normal operations, thus allowing more processor bandwidth to be provided to those other operations, which allows further growth of the network as one limitation has been removed. The need to replicate or provide commands between switches or routers is avoided as there are no redundant paths at that point.

Increased network scalability by router aware switches

Is in a Fibre Channel storage area network (SAN) and its included routers, the routers are placed in groups or pods. Each router only contains router port and fabric access data for routers in the same group or pod. In this manner the size of the relevant tables are reduced, which allows for greater expansion of the SAN as a whole. Each router may be programmed by an administrator with a pod value, indicating the pod containing the router. This value may be provided to the Name Server during router registration and may be requested from the Name Server when developing the router's inter-fabric router (IFR) topology or the pod value can be exchanged in the IF_ILS_HLO messages. Router port database information is only stored from routers in the same pod or provided to routers in the same pod, thus reducing the entries in the router port database.

Increased fabric scalability by designating router groups

A diagnostic testing utility is used to perform single link diagnostics tests including an electrical loopback test, an optical loopback test, a link traffic test, and a link distance measurement test. These diagnostics tests can be performed on trunked links and virtual channels and can be performed while QoS is enabled. Additionally the tests can be performed non-intrusively by using a dedicated VC such that regular traffic is not affected by the diagnostics testing.

Non-intrusive diagnostic port for inter-switch and node link testing

Back pressure is mapped within a network, and primary bottlenecks are distinguished from dependent bottlenecks. Further, the presently disclosed technology is capable of performing network healing operations designed to reduce the data load on primary bottlenecks while ignoring dependent bottlenecks. Still further, the presently disclosed technology teaches identifying and/or suggesting a switch port for adding a node to the network. More specifically, various implementations analyze traffic load and back pressure in a network, identify primary and dependent bottlenecks, resolve the primary bottlenecks, collect new node parameters, and/or select a switch port for the new node. Further, a command can be sent to a selected switch to activate an indicator on the selected port. New node parameters may include new node type, maximum load, minimum load, time of maximum load, time of minimum load and type of data associated with the new node.

Back pressure remediation

Specially configured host bus adaptors (HBAs) are located in the switches in the fabric or added to a management server. The HBAs are programmed to perform higher level protocol analysis, such as monitoring exchanges, LUNs, and VM/LUN/LBA ranges and developing SCSI statistics on the flows. The HBA can detect protocol exceptions. When located in the switches, mirroring of the desired frames is simplified as all routing is handled internally to the switch, with the HBA/switch combination providing data over the management Ethernet port to a management server. To maintain performance, preferably only the protocol relevant frames are mirrored to the HBAs. With the data provided from the HBAs, in conjunction with the normal diagnostic and statistics data available from the switch itself, the management server can correlate events end-to-end, correlate I/O error statistics collected from the fabric with server statistics, and correlate performance statistics collected from the fabric with server statistics.

Pervasive network diagnostics using HBAs

A Layernetwork switch is partitionable into a plurality of switch fabrics. The single-chassis switch is partitionable into a plurality of logical switches, each associated with one of the virtual fabrics. The logical switches behave as complete and self-contained switches. A logical switch fabric can span multiple single-chassis switch chassis. Logical switches are connected by inter-switch links that can be either dedicated single-chassis links or logical links. An extended inter-switch link can be used to transport traffic for one or more logical inter-switch links. Physical ports of the chassis are assigned to logical switches and are managed by the logical switch. Legacy switches that are not partitionable into logical switches can serve as transit switches between two logical switches.

Creation and deletion of logical ports in a logical switch

Growing community of inventors

San Jose, CA, United States of America

Badrinath Kollu