[PROXYPER] many questions about blocks with pproxy

Jeff Lawson jlawson at bovine.net
Sat Jan 23 20:19:55 EST 1999

Jeff Lawson   http://www.cs.hmc.edu/~jlawson/   http://www.bovine.net/
Jeffrey_Lawson at hmc.edu   jlawson at bovine.net    bovine at distributed.net
Programmer, Developer, Mascot, Founder of the largest computer on earth!
Don't waste those cycles!  Put them to use!  http://www.distributed.net/

On Tue, 19 Jan 1999, patriiiiiiiiiick wrote:

> Hi!
> Here are some affirmations (that I ask you to tell me if they are true) and 
> questions. I put them on separate lines for easier answer. I haven't found 
> any FAQ with these. Maybe I didn't search well. Otherwise, I think some of 
> them should be in a FAQ.
> There is no way to tell the pproxy what size of blocks it should fetch.

Currently the proxies will try to request blocks of up to 64*2^28 keys in
size.  If the number of additional blocks needed is below 64, then it will
request fewer.  You will frequently receive blocks that are smaller than
64 since the full servers may not always have a full-sized block available
on hand.

> When the client asks the pproxy a block of a certain size, the pproxy 
> "makes" a block of the appropriate size.


> When the client flushes its blocks, the pproxy reassembles (when
> implemented ==> not 303 yet) them if possible (same participant, ...)
> into bigger blocks.

Correct when implemented.  280 did this only when the email and platform
and os of all of the adjascent blocks were identical.

> What is the size of blocks the pproxy exchanges with the outside?

Whatever size the blocks could be consolidated to in the above step.

> When does the activity of a pproxy becomes perceptible on a normal
> modem line (e.g. 33.6 kbps)? How many machines or what total speed?

I doubt you should have problems with full utilization because of proxy
even on a modem, unless you are running hundreds of machines.  The most
significant factor that will probably limit speed is latency of the modem.

> What is the interest then of asking big blocks from the clients, apart from 
> decreasing the disk and LAN traffic (negligible, I presume)?

It is much more efficient, particularly higher up in the network, such as
at the keymaster and the statsbox.

To unsubscribe, send 'unsubscribe proxyper' to majordomo at lists.distributed.net

More information about the proxyper mailing list