Virtual machine managing system using snapshot

1 . A method comprising:

hosting, by a virtualization computing platform, a plurality of virtual machines comprising a first virtual machine, wherein the first virtual machine is connected with a first user device associated with a user account;

detecting that the first user device is disconnected from the first virtual machine;

determining, using a machine learning model and based on computing resource usage, of the first virtual machine, during a time period while the user device is disconnected, that the first virtual machine is in an idle state;

obtaining, based on the determination that the first virtual machine is in the idle state, a snapshot of the first virtual machine;

ceasing hosting the first virtual machine;

establishing, based on a request from a second user device, a second virtual machine, wherein the second user device is associated with the user account; and

loading, to memory of the second virtual machine, information associated with the snapshot;

training, using training data comprising a plurality of input items each comprising information associated with computing resource usage of a virtual machine, the machine learning model to output a determination, corresponding to each input item, of whether the respective virtual machine is in the idle state or not;

receiving a plurality of user responses each indicating whether the determined idle state of a corresponding virtual machine is correct or not; and

updating, based on the plurality of user responses, the machine learning model; wherein the training of the machine learning model is according to the following formula:

Loss

N

(

θ

)

=

1

N

⁢

∑

i

{

-

[

y

i

*

log

⁡

(

p

i

θ

)

+

(

1

-

y

i

)

+

log

⁡

(

1

-

p

i

θ

)

]

×

t

i

λ

+

t

i

}

wherein:

N=a total number of samples;

yi=expected idle/busy state of the ith sample;

θ=parameters of the machine learning;

pi_θ=prediction result made by the machine learning with parameters θ;

ti=disconnection time period of the ith sample;

λ=a non-trainable hyperparameter; and

wherein a training goal is to adjust the parameters θ to minimize the Loss N (θ) by leveraging Stochastic Gradient Descent optimizer

D

ˇ

=

arg

min

θ

Loss

N

.

2 . The method of claim 1 , wherein the computing resource usage comprises at least one of:

processor usage;

memory usage; or

network usage.

3 . The method of claim 1 , further comprises:

sending, after the determination that the first virtual machine is in the idle state, before the ceasing hosting of the first virtual machine, and to a third user device associated with the user account, a notification that the hosting of the first virtual machine is to be ceased; and

wherein the ceasing hosting the first virtual machine is further based on a determination that a declination to ceasing hosting the first virtual machine is not received during a second time period.

4 . The method of claim 1 , wherein the plurality of virtual machines further comprises a third virtual machine connected with a third user device, and wherein the method further comprises:

determining that the third virtual machine is in the idle state;

sending, after the determination that the third virtual machine is in the idle state, a notification that the hosting of the third virtual machine is to be ceased;

receiving a response, to the notification, that declines to cease hosting the third virtual machine; and

determining, based on the response, that the third virtual machine is in a busy state.

5 . The method of claim 1 , wherein the first virtual machine is managed by a virtual delivery agent (VDA).

6 . The method of claim 1 , further comprises:

determining that utilization of one or more computing resources associated with the virtual computing platform exceeds a threshold, and wherein the ceasing of hosting the first virtual machine is further based on the determination.

7 . An apparatus, comprising:

one or more processors; and

memory storing instructions that, when executed by the one or more processors, cause the apparatus to:

host a plurality of virtual machines comprising a first virtual machine, wherein the first virtual machine is connected with a first user device associated with a user account;

detect that the first user device is disconnected from the first virtual machine;

determine, using a machine learning model and based on computing resource usage, of the first virtual machine, during a time period while the user device is disconnected, that the first virtual machine is in an idle state;

obtain, based on the determination that the first virtual machine is in the idle state, a snapshot of the first virtual machine;

cease hosting the first virtual machine;

establish, based on a request from a second user device, a second virtual machine, wherein the second user device is associated with the user account; and

load, to memory of the second virtual machine, information associated with the snapshot;

train, using training data comprising a plurality of input items each comprising information associated with computing resource usage of a virtual machine, the machine learning model to output a determination, corresponding to each input item, of whether the respective virtual machine is in the idle state or not:

receive a plurality of user responses each indicating whether the determined idle state of a corresponding virtual machine is correct or not; and

update, based on the plurality of user responses, the machine learning model;

wherein training the machine learning model comprises training the machine learning model according to the following formula:

Loss

N

(

θ

)

=

1

N

⁢

∑

i

{

-

[

y

i

*

log

⁡

(

p

i

θ

)

+

(

1

-

y

i

)

+

log

⁡

(

1

-

p

i

θ

)

]

×

t

i

λ

+

t

i

}

wherein

N=a total number of samples;

yi=expected idle/busy state of the ith sample;

θ=parameters of the machine learning;

pi_θ=prediction result made by the machine learning with parameters θ;

ti=disconnection time period of the ith sample;

λ=a non-trainable hyperparameter; and

wherein a training goal is to adjust the parameters θ to minimize the Loss N (θ) by leveraging Stochastic Gradient Descent optimizer

D

ˇ

=

arg

min

θ

Loss

N

.

8 . The apparatus of claim 7 , wherein the computing resource usage comprises at least one of:

processor usage;

memory usage; or

network usage.

9 . The apparatus of claim 7 , wherein the instructions, when executed by the one or more processors, cause the apparatus further to:

send, after the determination that the first virtual machine is in the idle state, before the ceasing hosting of the first virtual machine, and to a third user device associated with the user account, a notification that the hosting of the first virtual machine is to be ceased; and

wherein the ceasing hosting the first virtual machine is further based on a determination that a declination to ceasing hosting the first virtual machine is not received during a second time period.

10 . The apparatus of claim 7 , wherein the plurality of virtual machines further comprises a third virtual machine connected with a third user device, and wherein the instructions, when executed by the one or more processors, cause the apparatus further to:

determine that the third virtual machine is in the idle state;

send, after the determination that the third virtual machine is in the idle state, a notification that the hosting of the third virtual machine is to be ceased;

receive a response, to the notification, that declines to cease hosting the third virtual machine; and

determine, based on the response, that the third virtual machine is in a busy state.

11 . The apparatus of claim 7 , wherein the first virtual machine is managed by a virtual delivery agent (VDA).

12 . The apparatus of claim 7 , wherein the instructions, when executed by the one or more processors, cause the apparatus further to:

determine that utilization of one or more computing resources associated with the virtual computing platform exceeds a threshold, and wherein the ceasing of hosting the first virtual machine is further based on the determination.

13 . A non-transitory computer-readable medium storing computer instruction that, when executed by one or more processors of a computing device, cause:

hosting a plurality of virtual machines comprising a first virtual machine, wherein the first virtual machine is connected with a first user device associated with a user account;

detecting that the first user device is disconnected from the first virtual machine;

determining, using a machine learning model and based on computing resource usage, of the first virtual machine, during a time period while the user device is disconnected, that the first virtual machine is in an idle state;

obtaining, based on the determination that the first virtual machine is in the idle state, a snapshot of the first virtual machine;

ceasing hosting the first virtual machine;

establishing, based on a request from a second user device, a second virtual machine, wherein the second user device is associated with the user account; and

loading, to memory of the second virtual machine, information associated with the snapshot;

wherein the instructions, when executed by the one or more processors, cause the computing device to train the machine learning model according to the following formula:

Loss

N

(

θ

)

=

1

N

⁢

∑

i

{

-

[

y

i

*

log

⁡

(

p

i

θ

)

+

(

1

-

y

i

)

+

log

⁡

(

1

-

p

i

θ

)

]

×

t

i

λ

+

t

i

}

wherein:

N=a total number of samples;

yi=expected idle/busy state of the ith sample;

θ=parameters of the machine learning;

pi_θ=prediction result made by the machine learning with parameters θ;

ti=disconnection time period of the ith sample;

λ=a non-trainable hyperparameter; and

wherein a training goal is to adjust the parameters θ to minimize the Loss N (θ) by leveraging Stochastic Gradient Descent optimizer

D

ˇ

=

arg

min

θ

Loss

N

.

14 . The non-transitory computer-readable medium of claim 13 , wherein the computing resource usage comprises at least one of:

processor usage;

memory usage; or

network usage.

15 . The non-transitory computer-readable medium of claim 13 , wherein the instructions, when executed by the one or more processors, cause the computing device further to perform actions comprising:

training, using training data comprising a plurality of input items each comprising information associated with computing resource usage of a virtual machine, the machine learning model to output a determination, corresponding to each input item, of whether the respective virtual machine is in the idle state or not;

receiving a plurality of user responses each indicating whether the determined idle state of a corresponding virtual machine is correct or not; and

updating, based on the plurality of user responses, the machine learning model.