The same data is added to the corresponding TVAs in each time period but each period stores the data differently, based on the resolution of that period. A TVA can be visualised as being one of a number of graphs of traffic that records the data relevant for its own combination of key details. Figure 3-6 Web representation of data in Figure 3.5 22 <strong>IBM</strong> <strong>Tivoli</strong> Netcool Performance Flow Analyzer: <strong>Installation</strong> <strong>and</strong> <strong>User</strong> <strong>Guide</strong>
The memory footprint for each TVA is dependent on how many buckets it has to hold. Period Coverage Number of Buckets Resolution (seconds) Resolution TVA Memory Requirement Hour 1 hour 360 10 10 seconds 12296 Bytes Day 25 hours 300 300 5 minutes 10376 Bytes Week 7 days 1 hour 338 1800 30 minutes 11592 Bytes Month 31 days 2 373 7200 2 hours 12712 Bytes hours Quarter 93 days 4 hours 26 minutes 40 seconds 272 29600 8 hours 13 minutes 20 seconds 9480 Bytes Year 366 days 366 86400 24 hours 12488 Bytes Month 31 days 8940 300 5 minutes 286856 Bytes. (HiRes) 1 hour Figure 3-7 Memory requirements The number of TVAs that must be stored depends on the traffic that is received. If all the flows are for one IP Address, Application Name pair then there is one TVA. If only a single octet is sent to another port on that same IP address then a second TVA is created for that IP Address, Application Name pair to hold that information. The memory footprint is the same for any aspect that records octets, flows, or packets. As traffic information is analyzed by <strong>Tivoli</strong> Netcool Performance Flow Analyzer, the number of TVAs increases as the flow analyzer recognises that existing TVAs do not cover the data presented. If the traffic is quite widely spread across IP address <strong>and</strong> ports, the memory required to store all the octets broken down by IP Address, Application Name increases by TVA, until the available RAM is used up by <strong>Tivoli</strong> Netcool Performance Flow Analyzer. To avoid RAM being consumed without restrictions, there are two attributes of each aspect- Upper Limit <strong>and</strong> Maximal Number. By default, these values are 4000 <strong>and</strong> 2000 respectively. If the number of TVAs grows to reach the value of Upper Limit, TVAs are deleted to bring the number of TVAs down to the Maximal Number. The TVAs that are removed are deleted, <strong>and</strong> their data is lost. The total amounts are still maintained. No guarantee is made regarding which TVAs are removed, but an effort is made to keep the same distribution of TVAs based on amount of traffic. For the data for one hour of octets, 4000 separate IP Address <strong>and</strong> Application Name combinations, the amount of RAM required is 4000 * 12,296 Bytes = 49,184,000 Bytes or 46.9 MB. If we were to record an hour, day, week, month, quarter, <strong>and</strong> year of octets, packets, <strong>and</strong> flows for Host Application that hit 4000 TVAs, then the calculation would be: (12,296 + 10,376 + 11,592 + 12,712 + 9,480 + 12,488) * 3 units * 4,000 TVAs = 68,944 * 3 * 4,000 = 827,328,000 Bytes or 789 MB. The same default values for Upper Limit <strong>and</strong> Maximal Number operate for each aspect type even though certain aspects could not reach 4000 TVAs. For example, there are likely to be only a certain number of flow emitters in a network. © Copyright <strong>IBM</strong> Corp. 2004, 2010 23