The 'P' in FCAPS is a very strong Quality Of Service(QOS) measure for any Networking environment and means a lot from the Service Level Agreement(SLA) perspective since it involves money. Lets see in this post what does comprise parameters for “P” and what would an NMS require from this perspective.
Network bandwidth or the performance depends upon the traffic or the payload that is being carried in the network. There are many factors those are responsible for the performance. Each of the device or Network Element has certain set of characteristics which effect the bandwidth and these characteristics depend upon some input criteria. The mechanism of optimizing the input parameters for the device so that the specified output criteria is met is called Throttling. This can mean different things for different devices for different parameters, either hardware or software. Lets look at different kinds of throttling:
Bandwidth throttling is a method of ensuring a bandwidth intensive device, such as a server, will limit ("throttle") the quantity of data it transmits and/or accepts within a specified period of time. Bandwidth throttling helps provide quality of service (QoS) by limiting network congestion and server crashes
A server, such as a web server, is a host computer connected to a network, such as the Internet, which provides data in response to requests by client computers. Understandably, there are periods where client requests may peak (certain hours of the day, for example). Such peaks may cause congestion of data ( bottlenecks) across the connection or cause the server to crash, resulting in downtime. In order to prevent such issues, a server administrator may implement bandwidth throttling to control the number of requests a server responds to within a specified period of time.
When a server using bandwidth throttling has reached the allowed bandwidth set by the administrator, it will block further read attempts, usually moving them into a queue to be processed once the bandwidth use reaches an acceptable level. Bandwidth throttling will usually continue to allow write requests (such as a user submitting a form) and transmission requests, unless the bandwidth continues to fail to return to an acceptable level.
Likewise, some software, such as peer-to-peer (P2P) network programs, have similar bandwidth throttling features, which allow a user to set desired maximum upload and download rates, so as not to consume the entire available bandwidth of his or her Internet connection.
CPU throttling refers to a series of methods for reducing power consumption in computers by lowering the clock frequency. Other methods include reducing the supply voltage and the capacitance
I/O throttling: input /output throttling, a technique used to more efficiently handle memory processing. During low-memory conditions, a system will slow down the processing of I/O memory requests, typically processing one sequence at a time in the order the request was received. I/O throttling slows down a system but typically will prevent the system from crashing.
When the NMS wants to tackle throttling issue it should serve the following purpose:
Determine the problems if any prior to their affecting the services
Maximize network utilization
Preempt the occurrence of the congestion
Demonstrate the compliance with the agreed SLAs (Service Level Agreement)
Indicate when extra network investment is needed.
Requirements for such system:
Receives asynchronously to the generated NE data
Pro actively retrieve the data from the NE the data such as Bandwidth consumption with respect to interfaces
Configuring intelligent NE to give out this data either as Alarm/Trap
Mediation being introduced for producing sanitizing details for the downstream use by third party applications
Aggregation of separate performance data records of Nodes / Interfaces / Links / LSPs / Multi service cross connections – Ethernet to MPLS, FR over ATM etc.
Policies can be built with predefined format and the device/network views can be polled
Correlation of the aggregated data with the associated managed objects like no of IP packets by an LSP End to End, No of Ethernet frames forwarded by LSP End to End, No of Cells carried or dropped by an ATM switch etc.
Reports can be generated from such data with respect to the utilization of managed objects like interfaces, nodes etc., difference between actual and planned loads, real time and historical SLA conformance etc.
Topology issues like detection of a congested link, traffic flow in the circuit which would have serious implications with respect to the performance of the network.
Maintaining Database tables for this effect at the NE level and at network level is essential for producing such kind of data and reports.
In my next post I would delve further on the strategies from NMS perspective.