Accedian is now part of Cisco  |

Avatar photo
By Boris Rogier

Performance troubleshooting in a virtual or cloud data center: 3 guidelines

Cloud,

Cisco’s latest Global Cloud Index shows that 76% of network traffic today never even leaves the data center. This makes performance troubleshooting in a virtual data center or in the cloud extremely challenging.

76% of network traffic remains INSIDE the data center

According to Cisco, this high degree of intra-data center traffic can be attributed to functional separation of application servers, storage, and databases, which generates replication, backup, and read/write traffic traversing the data center. Contrast this with an older, simpler model where monitoring was focused on “the Internet connection” or at least a few identifiable choke points that all traffic went through, and you can see how the issue of visibility is growing in every network.

Performance troubleshooting in a virtual data center or cloud environment

Performance troubleshooting requires visibility in the virtual, cloud, and SD layers

You cannot conduct any performance troubleshooting or diagnostic, without having a straightforward and permanent means of capturing traffic inside your virtual, cloud, or software-defined environment.

3 rules for performance troubleshooting in a virtual datacenter

Here are 3 important criteria that you should consider before you get started with a performance troubleshooting/monitoring solution:

  1. Native integration: your solution should rely on existing / native switching devices to capture traffic and imply no change in your virtualization/cloud infrastructure
  2. Scalability and distribution: you should make sure that you can analyze the traffic inside your virtual or cloud infrastructure and that your solution can easily be distributed throughout your virtual data center (e.g., avoid extracting traffic from your virtual data center to physical devices as this will not be easy to extend to the entire data center)
  3. No side effects: pay attention to the resource requirements of your performance troubleshooting tool. They should be minimal for CPU and RAM consumption, as well as for the bandwidth required to centralize the data.

SkyLIGHT PVX is an example of a network and application performance monitoring and management tool that provides full visibility across virtualized data centers and cloud environments.