Traffic analyzer for isolated virtual networks

A method, comprising:

• collecting, at a virtualization management component of a first host of a virtualized computing service, a first metric set associated with a network packet generated at a guest virtual machine of a first isolated virtual network of the virtualized computing service, wherein the first isolated virtual network comprises a particular configuration of resources, pathways, and connections, wherein the guest virtual machine is implemented at the first host, and wherein the first metric set includes one or more of:
• (a) a source network address, (b) a destination network address, or (c) a packet size;
• identifying, at a control plane component executing on one or more computing devices of the virtualized computing service, a plurality of endpoint categories with respect to network traffic of the first isolated virtual network, including a first endpoint category which includes a second isolated virtual network of the virtualized computing service, and a second endpoint category does not include either the first isolated virtual network or the second isolated virtual network;
• determining, at the control plane component based at least in part on an analysis of a plurality of metric sets including the first metric set, and based at least in part on the identified plurality of endpoint categories, (a) a first amount of data transmitted between the first isolated virtual network and the first endpoint category during a selected time interval, wherein the first endpoint category does not include the first isolated virtual network, and (b) a second amount of data transmitted between the first isolated virtual network and the second endpoint category;
• storing, by the control plane component, a result of a predictive model to which the first amount of data is provided as input, wherein the result comprises an indication of an expected future level of traffic between the first isolated virtual network and the first endpoint category;
• and
• generating a change recommendation for the particular configuration of resources, pathways, and connections of the first isolated virtual network based at least in part on the result of the predictive model.

A non-transitory computer-accessible storage medium storing program instructions that when executed on one or more processors implement a traffic analyzer of a provider network, wherein the traffic analyzer is configured to:

• identify a plurality of endpoint categories with respect to network traffic associated with a first isolated virtual network of the provider network, including a first endpoint category which includes an endpoint configured within a second isolated virtual network of the virtualized computing service, and a second endpoint category does not include either the first isolated virtual network or the second isolated virtual network, and wherein the first isolated virtual network comprises a particular configuration of resources, pathways, and connections;
• determine, based at least in part on an analysis of a plurality of metric sets, including a particular metric set collected at a first virtualization host of the provider network, and based at least in part on the identified plurality of endpoint categories, (a) a first amount of data transmitted between the first isolated virtual network and the first endpoint category during a selected time interval, wherein the first endpoint category does not include the first isolated virtual network, and (b) a second amount of data transmitted between the first isolated virtual network and the second endpoint category, wherein the particular metric set is associated with a particular network packet transmitted between the first virtualization host and a particular endpoint, and wherein the particular metric set includes one or more of:
• (a) a source network address, (b) a destination network address, or (c) a packet size;
• store a result of a predictive model to which the first amount of data is provided as input, wherein the result comprises an indication of an expected future level of traffic between the first isolated virtual network and the first endpoint category;
• and
• generate a change recommendation for the particular configuration of resources, pathways, and connections of the first isolated virtual network based at least in part on the result of the predictive model.