Real-time compressive data collection for cloud monitoring

1. A server configured to employ compressive sensing based data collection in cloud monitoring, the server comprising:

a memory configured to store instructions;

a processor coupled to the memory, the processor configured to execute a cloud monitoring service, wherein the cloud monitoring service is configured to:

receive multi-dimensional data associated with a performance of a cloud infrastructure collected by a probe within the cloud infrastructure; and

compress the multi-dimensional data through a process to:

vectorize the multi-dimensional data, and

concatenate the vectorized multi-dimensional data into a single-dimensional data vector, wherein the multi-dimensional data is reconstructable with the single-dimensional data vector.

2. The server of claim 1 , wherein the cloud monitoring service is further configured to:

apply compressive sampling to the multi-dimensional data through a multiplication of the single-dimensional data vector with a random construction matrix, wherein each column of the random construction matrix includes a series of random numbers for the probe and the probe corresponds to elements of the single-dimensional data vector.

3. The server of claim 1 , wherein a state data of the cloud infrastructure is associated with a property of the probe.

4. The server of claim 3 , wherein the property of the probe includes one or more of: a processor utilization, a memory utilization, a disk input/output (I/O), and adverse event associated with an application, a running time associated with the application, a resource allocation, a context information, a network usage, and a disk temperature.

5. The server of claim 1 , wherein the cloud monitoring service is further configured to:

in response to receiving the multi-dimensional data from the probe, initialize a produce thread of the probe and initialize a consumer thread of the probe;

send the multi-dimensional data through the producer thread to allow the producer thread to collect the multi-dimensional data from the probe; and

send the multi-dimensional data through the consumer thread to allow the consumer thread to read the multi-dimensional data from the probe.

6. The server of claim 1 , wherein the cloud monitoring service is further configured to:

send the single-dimensional data vector through a load balancer on a network switch to allow the load balancer to balance a transmission of the single-dimensional data vector, wherein the load balancer employs a pre-defined load balancing mechanism computed from global control parameters and is configured to assign the load balancing mechanism based on a network context.

7. The server of claim 1 , wherein the cloud monitoring service is further configured to:

steer the single-dimensional data vector through multiple levels of network switches to a service manager.

8. The server of claim 1 , wherein the cloud monitoring service is further configured to:

reconstruct the single-dimensional data vector, wherein the single-dimensional data vector includes a linear programming optimization;

extract a sparse domain information from the single-dimensional data vector; and

transform the single-dimensional data vector to an original state domain by an application of a reverse sparse domain transform operation generated from the sparse domain information.

9. The server of claim 1 , wherein the cloud monitoring service is further configured to:

monitor a transmission rate of the single-dimensional data vector within the cloud infrastructure; and

adjust the transmission rate of the single-dimensional data vector according to one or more real-time monitoring service specifications.

10. The server of claim 1 , wherein the cloud monitoring service is further configured to:

broadcast a random global seed from a central node of the cloud infrastructure to allow the probe to generate a seed using the random global seed and a probe identification to avoid a transmission of a random construction matrix throughout the cloud infrastructure.

11. A cloud-based datacenter configured to employ compressive sensing based data collection in cloud monitoring, the datacenter comprising:

sets of servers within a cloud based infrastructure, wherein the sets of servers are monitored by a probe to capture a performance of the cloud based infrastructure within a multi-dimensional data;

a cloud monitoring service executed on a server, the cloud monitoring service configured to provide control parameters to network switches, wherein the network switches include one or more of a first switch level, a second switch level, and a third switch level; and

the network switches connected to the sets of servers, wherein the network switches are configured to:

receive the control parameters from the cloud monitoring service at the first switch level of the network switches;

receive the multi-dimensional data at the first switch level of the network switches;

encode the multi-dimensional data at the first switch level of the network switches by compressive sampling;

determine a sparsity feature of the multi-dimensional data; and

apply compressive sensing to compress the multi-dimensional data into a single-dimensional data vector using the sparsity feature, wherein the multi-dimensional data is reconstructable with the single-dimensional data vector at the third switch level of the network switches.

12. The cloud-based datacenter of claim 11 , wherein the cloud monitoring service is configured to:

transmit an instruction to the probe to allow the probe to control a collection of the multi-dimensional data.

13. The cloud-based datacenter of claim 11 , wherein the cloud monitoring service is configured to:

compress the multi-dimensional data through vectorization of the multi-dimensional data and concatenating the multi-dimensional data into the single-dimensional data vector.

14. The cloud-based datacenter of claim 11 , wherein the network switches is further configured to:

compress the multi-dimensional data through a sparse transformation, wherein the sparse transformation includes one of: a discrete cosine transform and a wavelet transform.

15. The cloud-based datacenter of claim 11 , wherein the network switches are further configured to:

employ a first communication link between the first switch level of the network switches and the second switch level of the network switches and employ a second communication link between the second switch level of the network switches and the third switch level of the network switches, wherein the first communication link and the second communication link allow summation operations of the multi-dimensional data.

16. The cloud-based datacenter of claim 11 , wherein the network switches are further configured to:

filter the multi-dimensional data with a data filtering module to generate global control parameters, wherein the global control parameters are utilized to correct measurements by the probe based on the control parameters.