Current trends toward virtualization and convergence tightly link networking and system performance and blur the line between the network and the servers. For example, virtualization places switching and routing functions on servers. Monitoring the network in this environment requires that the servers are also monitored. Similarly anyone concerned with application and server performance must now also be concerned about network performance since application features (such as virtual machine migration) can significantly affect and be affected by network performance.
Currently, performance monitoring of servers and applications is highly fragmented. Each server vendor, operating system vendor and application developer creates agents and software for performance monitoring, none of which interoperate. Monitoring performance of a server might require the same (or similar but incompatible metrics) be monitored more than once since each management application requires its own agent. In addition, there may be other agents monitoring different hardware and software elements within the server. The development of performance and application monitoring in sFlow requires progress on two parallel tracks: 1. specifying the additional structures to be used when exporting performance metrics in sFlow 2. implementing sFlow agents that export the metrics in order to allow application developers and system administrators to deploy agents and gain experience. The draft Host sFlow specification describes some basic sFlow structures for monitoring host performance and lays a foundation for future work to extend sFlow to monitor virtual machines and application services (e.g. Xen, Hyper-V, VMware, HTTP, NFS, memcached, Hadoop etc): http://www.sflow.org/sflow_host-draft1.txt The Host sFlow sourceforge project is building an open source Host sFlow agent that can easily be ported to a wide variety of operating systems: http://host-sflow.sourceforge.net/ Initial prototypes of the Host sFlow agent have been running on over 1000 Windows and Linux servers. For server and application management, the combination of sFlow in the top of rack switch and Host sFlow is particularly compelling. Servers maintain many useful performance counters, but don't typically have hardware support for packet sampling so have a very limited view their network I/O. The top of rack switch has hardware packet sampling support and sees all the traffic in and out of the servers. Host sFlow exports the counters that are currently stranded on the server, allowing a performance management system to put together the network and system performance for each server by combining data from the switch and the servers.
