Load Balancing & Hashing by Hand: Round Robin, Weighted & Least-Connections Rules, Hash Mod N, Consistent Hashing Rings & Capacity Headroom

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Hand out 12 requests round robin to three servers with MOD, count what each server receives, work out how long each one is busy and see why equal shares are unfair when one server is slower.
• Give three servers the weights 5, 3 and 2, calculate their expected shares, assign 15 requests with a weighted cycle and cumulative weights, and measure how far the actual shares are from the plan.
• Step through ten arrivals under the least-connections rule: open connections change as requests finish and arrive, you pick the server with the fewest open connections, count where the requests went and check how balanced the pool ends up.
• Place twelve keys on four servers with hash mod N, look up the server names, count the keys per server and measure how uneven the split is.
• Add a fifth server to a hash-mod-N pool, recalculate the server of every key, mark the keys that move and compare the share that moves with the minimum that would have been necessary.
• Place three servers and twelve keys on a ring from 0 to 999, give every key to the next server clockwise with MINIFS and a wrap-around, let a fourth server join and count how few keys move.
• Give each of three servers three positions on the ring, calculate the arc every position owns, add the arcs up per server, measure the imbalance and compare the spread of the shares for different numbers of virtual nodes.
• Size Wrenfield's pool for its peak with a target utilisation, add one spare server, calculate the utilisation per hour with all servers up and after one failure, the headroom that is left and chart both curves.