Browse Prior Art Database

Optimized Cache Update

IP.com Disclosure Number: IPCOM000029683D
Original Publication Date: 2004-Jul-08
Included in the Prior Art Database: 2004-Jul-08
Document File: 2 page(s) / 32K

Publishing Venue

IBM

Abstract

Optimized Cache Update

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Optimized Cache Update

Policy Agent (Pagent) is the center piece of the z/OS Policy Based Networking. It provides a variety of services, such as retrieving and managing policies for different disciplines (Quality of Service; Intrusion Detection Service; IPSec VPN etc.); and QoS performance monitoring to effect how load is distributed within Sysplex using Sysplex Distributor. Another important service provided by Pagent is maintaining a performance cache for users that exploit Policy API to access network QoS policy performance. The optimized cache information update algorithm described here is used to manage the cache content in such as way as to minimize system overhead and still provide clients acceptable up-to-date data.

Performance data for policies are collected in the kernel (stack). Management applications obtain the performance data, analyze it and make decisions using the data obtained. Different management applications can retrieve the management data for different purposes. For example, the SNMP subagent might retrieve the performance data to support the MIB requests from managers. A policy management application might retrieve the performance data to see if any new resources need to be provisioned or to determine if the SLAs are being met. Obtaining performance data from the kernel is an expensive operation and if multiple applications retrieve the same data from the kernel, performance would be affected adversely. Minimizing the number of times the performance data is retrieved from the kernel would help the performance of the system.

In the case of the policy system, the performance data that is collected in the kernel is retrieved and cached by a server (Policy Agent) that runs in the application space. The management applications that need the performance data obtain it from the Policy Agent's cache. This prevents the load on the kernel to supply the performance data to all the management applications. The Server (Policy Agent) has total control of when and how often it goes down to the kernel to obtain the data.

In determining the time at which the Policy Agent needs to go down to the kernel to obtain the performance data the following criteria have to be taken into consideration :

The entries in the cache should not be stale for any client request

The client can pass in an acceptableCacheTime. AcceptableCacheTime is the maximum age of the data that the client is willing to accept, while the cache time is the maximum amount of time that has elapsed since the cache was updated from the kernel. If the cache time is greater than the acceptableCacheTime specified by the client, the cache needs to be refreshed (data needs to be obtained from the kernel and stored).

Multiple clients can request the performance data from the Server.

The server itself is configured with a minimum samplin...