Scaling Network Management Systems (NMS)
Tuesday, 18 August 2009 14:54.
Scaling Network Management Systems (NMS)
One of the big problems with scaling NMS is IO. With so many device resources being updated at once traditional hard drives just can’t keep up and cause massive wait issues. You can of course use RAM disks but then you have a persistence issue, if the box dies you lose all your data.
The solution is to copy the data at set intervals but that just seems like a hack. We have been looking at SSD on PCIe cards, they don’t have traditional disks issues with massive random write and keep the data when they have lost power. There are quite a few players out there, Fusion-io (http://www.fusionio.com/), OCZ (http://www.ocztechnology.com/) etc. Fusion-io has the best performance but you pay for it (like most things in life), the OCZ products seem to have about half the performance for half the price. We will be testing these products using a great piece of software from SimpleSoft (http://www.smplsft.com/) that allows us to simulate 1000s of SNMP nodes. We will also be using another piece of great software, OpenNMS (http://www.opennms.org/index.php/Main_Page). OpenNMS is an excellent open source NMS, well designed, very flexible and of course free, I like free :)
