Managing idle and active servers in cloud data centers

Patent metadata

Field	Value
Publication number	US-10834182-B2
Application number	US-201715472374-A
Country	US
Kind code	B2
Filing date	Mar 29, 2017
Priority date	Mar 29, 2017
Publication date	Nov 10, 2020
Grant date	Nov 10, 2020

Abstract

A method, product, and apparatus for treating idle servers in a cloud system provide for extrapolating a purpose of each of a plurality of servers by comparing a list of processes active on the server to a plurality of lists of processes associated with a plurality of purposes; selecting vectors of idle/active features corresponding to the extrapolated purposes of each of the plurality of servers; classifying as idle or active each of the plurality of servers, by assessing the specified feature vectors using a linear support vector machine; validating as idle or active each server classified as idle, by assessing the connectivity of the server with all servers classified as active; and implementing at least one treatment option on servers that have been validated as idle. The treatment options may include terminating, terminating with snapshot, and stopping a virtual machine.

First claim

What is claimed is:

1. A non-transitory computer readable medium comprising computer executable instructions which when executed by a computer cause the computer to perform the method of:
extrapolating, by an analyzer executing on at least one hardware processor, a purpose of each of a plurality of servers by comparing a categorized list of processes active on each of the plurality of servers to a plurality of lists of processes associated with a plurality of purposes, wherein a list of processes active on each server is collected by at least one data collector agent running on the plurality of servers and the plurality of purposes comprise an analytics server, a DevOps server, and a development server;
selecting, by the analyzer executing on at least one hardware processor, vectors of idle or active features corresponding to the extrapolated purposes of each of the plurality of servers, wherein:
a first vector of idle or active features corresponding to the analytics server comprises a first plurality of predetermined parameters, which are specific to the operation of the analytics server;
a second vector of idle or active features corresponding to the DevOps server comprises a second plurality of predetermined parameters, at least one of which is different from the first plurality of parameters, which are specific to the operation of the DevOps server; and
a third vector of idle or active features corresponding to the development server comprises a third plurality of predetermined parameters, at least one of which is different from the first and second pluralities of parameters, which are specific to the operation of the development server;
wherein the first plurality of predetermined parameters, the second plurality of predetermined parameters and the third plurality of predetermined parameters are extracted from a dataset collected by the plurality of data collector agents;

classifying, by the analyzer executing on at least one hardware processor, as idle or active each of the plurality of servers, by assessing the specified idle or active feature vectors corresponding to the analytics server, the DevOps server, and the development server using a linear support vector machine, wherein the linear support vector machine finds a plurality of data points that define a hyperplane that separates all data points of the idle or active feature vectors, and determines whether each server is idle or active based on coordinates of the hyperplane;
validating, by the analyzer executing on at least one hardware processor, as idle or active each server classified as idle, by assessing the connectivity of each server classified as idle with all servers classified as active;
developing, by an action plan generator executing on at least one other hardware processor, treatment options for idle servers; and
implementing, by a cloud data center manager executing on at least one other hardware processor, at least one treatment option on servers that have been validated as idle, wherein the treatment option is selected from the list consisting of: terminating a virtual machine, terminating the virtual machine with snapshot, and stopping the virtual machine.

2. The non-transitory computer readable medium of claim 1 further comprising performing rule-based detection of idle servers, wherein the rules detect as idle servers that are long running, have no observable user processes, have no user login activities, or have no established network connections.

3. The non-transitory computer readable medium of claim 2 wherein the rules detect as active a server that has been running less than six months.

4. The non-transitory computer readable medium of claim 2 wherein the rules detect as active a server that has had a user login within the past three months.

5. The non-transitory computer readable medium of claim 2 wherein the rules detect as active a server that has had a network connection within the past month.

6. The non-transitory computer readable medium of claim 1 wherein the computer executable instructions comprise distinct software modules, each of the distinct software modules being embodied on the computer-readable storage medium, and wherein the distinct software modules comprise an analyzer, an action plan generator, and a cloud data center manager;
wherein:
the extrapolating, the selecting, the classifying, and the validating are carried out by the analyzer executing on at least one hardware processor;
the action plan generator executes on at least one other hardware processor to develop the at least one treatment options for idle servers; and
the implementing of the at least one treatment option is carried out by the cloud data center manager executing on at least one other hardware processor.

7. An apparatus comprising:
a memory; and
at least one hardware processor, coupled to said memory, and operative to perform a method comprising:
extrapolating, by an analyzer executing on at least one hardware processor, a purpose of each of a plurality of servers by comparing a categorized list of processes active on each of the plurality of servers to a plurality of lists of processes associated with a plurality of purposes, wherein a list of processes active on each server is collected by at least one data collector agent running on the plurality of servers and the plurality of purposes comprise an analytics server, a DevOps server, and a development server;
selecting, by the analyzer executing on at least one hardware processor, vectors of idle or active features corresponding to the extrapolated purposes of each of the plurality of servers, wherein:
a first vector of idle or active features corresponding to the analytics server comprises a first plurality of predetermined parameters, which are specific to the operation of the analytics server;
a second vector of idle or active features corresponding to the DevOps server comprises a second plurality of predetermined parameters, at least one of which is different from the first plurality of parameters, which are specific to the operation of the DevOps server; and
a third vector of idle or active features corresponding to the development server comprises a third plurality of predetermined parameters, at least one of which is different from the first and second pluralities of parameters, which are specific to the operation of the development server;
wherein the first plurality of predetermined parameters, the second plurality of predetermined parameters and the third plurality of predetermined parameters are extracted from a dataset collected by the plurality of data collector agents;

classifying, by the analyzer executing on at least one hardware processor, as idle or active each of the plurality of servers, by assessing the specified idle or active feature vectors corresponding to the analytics server, the DevOps server, and the development server using a linear support vector machine, wherein the linear support vector machine finds a plurality of data points that define a hyperplane that separates all data points of the idle or active feature vectors, and determines whether each server is idle or active based on coordinates of the hyperplane;
validating, by the analyzer executing on at least one hardware processor, as idle or active each server classified as idle, by assessing the connectivity of each server classified as idle with all servers classified as active;
developing, by an action plan generator executing on at least one other hardware processor, treatment options for idle servers; and
implementing, by a cloud data center manager executing on at least one other hardware processor, at least one treatment option on servers that have been validated as idle, wherein the treatment option is selected from the list consis

Assignees

• IBM

Inventors

• Hwang Jinho
• Kim In Kee
• Young Christopher C
• Zeng Sai

Classifications

• H04L43/20

  the monitoring system or the monitored elements being virtualised, abstracted or software-defined entities, e.g. SDN or NFV · CPC title

• H04L41/40

  using virtualisation of network functions or resources, e.g. SDN or NFV entities · CPC title

• H04L41/5019

  Ensuring fulfilment of SLA · CPC title

• H04L67/1031 Primary

  Controlling of the operation of servers by a load balancer, e.g. adding or removing servers that serve requests · CPC title

• H04L43/0811

  by checking connectivity · CPC title