Lucy, Brian, On Jan 11, 2011, at 08:41 MST, Yong Lucy wrote: >> In fact no solution works well for short flows and the problem isn't >> important >> for long flows with few packets. So I think it's OK to discuss a solution >> that works for long flows with many packets, since that covers most >> of the load. > > [LY] Hash works well for short flows. That was developed when Internet is > primarily used for data transport. As more and more video data on Internet, > traffic become mixed with long flows with many packets and short flow with > few packets. Hash no longer works well in this traffic pattern.
Let's be careful here and not cause a panic. :-) In reality, hash works quite well for the bulk of Video traffic currently being distributed over the public Internet, specifically in the form of streaming TV or movies such as Youtube, Hulu, Netflix, iTunes, etc. where individual streams/flows tend to be on the order of a few Mbps -- obviously because individual streams need to fit on "small" access circuits (relative to the size of Backbone links) to subscriber's homes that are primarily using either DSL or Cable (DOCSIS) -- at least from the PoV of here in North America. It's probably better to say that a hash algorithm works well where individual, long-lived flows (regardless of traffic type) are a small-ish fraction of the physical BW of any individual component-link in a LAG or ECMP group. It's when those long-lived flows are a substantial portion of component-link's physical BW that a hash algorithm is ineffective and you either need higher capacity component-links or more "creative" techniques. >>> [LY] the server load balance and the network path load balance have >>> different criteria. >> >> Very true. It isn't clear to me that balancing the network load >> without knowledge of the server load is a good idea. Brian: It is from the perspective that it's stateless, which is a very desirable property when you're talking about Core/Edge devices with millions or billions of flows ... -shane > [LY] Agree. However, today network performs load balance without considering > server load balance; server load performs load balance without considering > network load balance. Two operations are independent. > > Lucy >> >> Brian > > -------------------------------------------------------------------- > IETF IPv6 working group mailing list > ipv6@ietf.org > Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 > -------------------------------------------------------------------- -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------