[chord] chord simulator scalability?

[Michael Walfish]

There's a brand new from-scratch simulator; not sure if it's been released
yet (someone should jump in and answer). Depending on timing, you might
want to use the old one, in which case, please see below.

| I have a question about the Chord Simulator scalability. We are planning
| on trying some experiments on large scale data sets will 2-million,
| 5-million documents and higher, with of the order of 1000s of nodes. Do
| you have experience with simulations of this order with the Chord
| simulator? 

I've done simulations that are of this order (9*10^6 documents, 1000 nodes,
e.g.), but I needed to modify the simulator before doing so. Will soon 
submit these changes to Ion.

| Also we are expecting to simulate the system for the order of
| 10 to 20 days (the current def.h seems to have a max of about 2 hours).

I redefined MAX_TIME to MAX_INT.

| Do you suspect any scalability problems on the simulator for these kinds
| of configurations?

Believe that, with a few modifications, you won't have a problem. The 
simulator won't run quickly, but if you've got a relatively fast machine, 
you should be able to make progress.

When do you need to run these simulations?