User Tools

Site Tools


services:challenges

Challenges for Service Developers

When offering services in the different Community Networks that are part of CONFINE, the service developers have to take into consideration the characteristics of the type of services that they want to deploy and also the specific features and limitations of CNs. Once this assessment is performed, they will be able to find a trade-off between the variety and features of the services offered and the Quality of Service (QoS) that they can guarantee to end users. Therefore, a service developer has to answer the following questions about the services that he intends to run in Community-Lab:

  1. Is it a high-bandwidth demanding service?: What are the requirements in terms of network bandwidth? How will network topology and fluctuations affect the QoS?
  2. Is it a hardware-intensive service?: What resources are needed? How much CPU, memory or disk space?
  3. Is it a time-sensitive service?: What are the necessary conditions in terms of RTT, latency and jitter? Is it a time sensitive service such as video streaming or VoIP? Or is it a non-time-constrained service such as distributed storage?

In addition, the developer has to consider the special characteristics of CNs and how they can affect the quality, usefulness or even viability of the service. The main obstacles observed in Community-Lab for service deployment are the following:

  1. Network bandwidth: It is common to have network fluctuations and some nodes with limited bandwidth.
  2. TincVPN: It affects the nodes' connectivity. In some experiments performed in Community-Lab we experienced some difficulty when packets were exchanged from one node to another, having packet loss or no connectivity at all. In addition, we observed that maintaining the TincVPN involved a high CPU consumption in the nodes.
  3. Geographical location of nodes (broadcast domain): We observed in some tests performed in AWMN that the network nodes perceive the services with bad quality. A possible solution to this problem would be to replicate services in different Community Networks. For this specific case, we could locate the service publisher within the AWMN network. Another possibility is to use the CONFINE monitoring system to define a strategy for choosing nodes with best connectivity and lower latency. It would also be necessary to check TincVPN configuration in order to determine the causes of the increment in CPU consumption in some nodes.
  4. Hardware and software diversity: We have several networks, each one of them with an heterogeneity of devices, routing protocols and software components.
  5. Hardware limitations: In Community Networks, nodes are usually hardware-limited devices, with low processing power, memory and disk space.
  6. Low availability of nodes: It is also common to find that a high number of nodes are offline or not working properly. This fact makes more difficult the selection of appropriate nodes for service deployment.
  7. Unknown characteristics of nodes: Due to the network heterogeneity and node diversity it is difficult to select nodes with the appropriate hardware and software characteristics to deploy specific services. In order to solve this problem, we are currently working on a mechanism to help identify some hardware characteristics of the nodes and therefore be able to select nodes according to such characteristics.
  8. Quality of Service (QoS): It is difficult to guarantee a QoS in CONFINE. We need to find a solution to optimize the QoS offered to service users taking into account the network fluctuation.
services/challenges.txt · Last modified: 2016/12/19 11:23 by ivilata