In tc, each hashtable can have a maximum of 256 buckets and each bucket contains actual filters. Besides, there is an option of 'hashkey' which when used on traffic, gives a 8 bit number using which one can achieve Hashing filters. i.e. we dont have to search linearly in filters and can directly reach a bucket. If there are say, 1024, linearly increasing IPs then I need to search only 4 filters at max. however, if there are more IPs, and more importantly not linearly increasing and can fall in anywhere in a class A subnet 10.0.0.0/8, then I will have to do is multi-level hashing and achieve similar behaviour. But that would require atleast 3 times hashing calculation on every pkt and can be a performance intensive task. If this 8 bit can be increased to 24 bit and no. of buckets become 4M then it will be possible to achieve the required behaviour and that too with a nice performance. I agree that overhead could be memory in this but that is acceptable. Whats your view on this? Is it possible by some parameter editing or variable modification and recompiling tc or some other way? -- To unsubscribe from this list: send the line "unsubscribe lartc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
