Apparatuses, methods, and systems directed to efficient cooling of data centers. Some embodiments of the invention allow encapsulation of cold rows through an enclosure and allow server fans to draw cold air from the cold row encapsulation structure to cool servers installed on the server racks. In other particular embodiments, the systems disclosed can be used to mix outside cool air into the cold row encapsulation structure to cool the servers. In some embodiments, the present invention involves using multiple cold row encapsulation structures to cool the servers installed on the racks.

Cold row encapsulation for server farm cooling system

A distributed load balancer in which a router receives packets from at least one client and routes packet flows to multiple load balancer (LB) nodes according to a per-flow hashed multipath routing technique. For a given packet flow, the LB nodes randomly select a server node as a target for the packet flow from among multiple server nodes and send a connection request to the server node. A load balancer module on the server node makes the decision on whether to accept or reject the connection based on one or more metrics indicating a respective server's current load. If the module accepts the connection request, a connection is established between the server and the client. Otherwise, the load balancer nodes may select another server node and try again. The connections established between clients and servers pass through the load balancer nodes, but are not terminated at the load balancer nodes.

Distributed load balancer

A distributed load balancer in which a router receives packets from at least one client and routes packet flows to multiple load balancer nodes. The router exposes a public IP address and the load balancer nodes all advertise the same public IP address to the router. The router may implement a per-flow hashed multipath routing technique, for example an equal-cost multipath (ECMP) routing technique, to distribute the flows across the load balancer nodes. Thus, the multiple load balancer nodes may service a single public endpoint. The load balancer nodes may advertise to the router according to the Border Gateway Protocol (BGP). Rather than advertising itself, however, a load balancer node may be advertised to the router by one or more neighbor load balancer nodes; the neighbor nodes may terminate the BGP sessions with the router in response to determining that the load balancer node has failed.

Multipath routing in a distributed load balancer

Data that is collected and disseminated by mobile devices typically has to be processed, correlated with other data, aggregated, and then transmitted back to the mobile device users before the information becomes stale or otherwise irrelevant. These operations may be performed in a cloud-based solution that manages dataflow. The cloud-based solutions may be scalable and implemented in a fault-tolerant distributed system to support user-facing continuous sensing and processing services in the cloud-computing system. A system may monitor execution of data and shift workloads (i.e., balancing) in response to spatial and temporal load imbalances that occur in a continuous computing environment. A failure recovery protocol may be implemented that uses a checkpoint-based partial rollback recovery mechanism with selective re-execution, which may allow recovery of the continuous processing after an error while avoiding large amounts of downtime and re-execution.

Platform for Continuous Mobile-Cloud Services

A distributed load balancer in which a router receives packets from at least one client and routes packet flows to multiple load balancer (LB) nodes, which in turn distribute the packet flows among multiple server nodes. Each LB node may serve in ingress, egress, and/or flow tracker roles. Each LB node may include a first network interface controller (NIC) that faces the router and a second NIC that faces the server nodes. Each LB node may implement a core packet processing architecture in which packets received at the NICs are distributed among non-blocking input queues of a set of worker cores by receiver cores for the NICs, processed from the input queues by the worker cores, and placed on non-blocking input queues of transmit cores for the NICs.

Distributed load balancer node architecture

A load balancer allocates requests to a pool of web servers configured to have low queue capacities. If the queue capacity of a web server is reached, the web server responds to an additional request with a rejection notification to the load balancer, which enables the load balancer to quickly send the rejected request to another web server. Each web server self-monitors its rejection rate. If the rejection rate exceeds a threshold, the number of processes concurrently running on the web server is increased. If the rejection rate falls below a threshold, the number of processes concurrently running on the web server is decreased.

Robert M. English

Papers

Load balancing web service by rejecting connections

Receiving at a secondary port of a first web-server a request that has been rejected a predetermined amount of time by multiple web-servers

Load balancing by circling back to a secondary port of a first web server after a request is rejected a threshold number of times