Raza Speed Got Suddenly Up!

[jackc]

Well, this is weird: I was used to get Raza speeds of no more than 10-12 KB/s (if there were Raza clients in my d/l sources), or 5-6 KB/s (if there were only ed2k ones). This was the reason that as I explained in a previous post, Raza had become my fourth choice (after BT, IRC, and FT).

Imagine my amazement then, when I saw today file ("Requiem for a Dream" video), running at 35-40 KB/s in Raza! Not only that, but the source was an emule client!!! :huh: Considering that I was also d/ling two other files in Raza (with usual speeds of 5-6 KB/s, AND an IRC file at a 10 KB/s), my 512/128 connection was maxed out. WHich simply means, that maybe this 35-40 KB/s might have an even higher potential.

I tried to see what could be the reason. The only thing I noticed was that the generous client providing me this speed was an emule v.0.61, whereas the others who were queuing me were v.0.3 or v.0.4. Maybe this was a coincidence, maybe not.

Does anyone have a clue about this? Does this newer version of emule work better as far transfer speeds are concerned?

[Switeck]

Probably just a statistical fluke.

Maybe the user was one of the 'bleeding edge' types that HAD to have 'latest-and-greatest' tech, be it software or hardware. So he might have a really fast connection that wouldn't even notice a 'measly' 40 KB/sec upload.

There's quite a few broadband connections now that have 384k (k here is kilobit/sec) upload bandwidth. So they could reach upload speeds of 40 KB/sec if only uploading to 1 person. Maybe the user didn't set upload limits -- or put them near the limit of his connection -- then allowed only 1 upload at a time?

Lastly, there could be a major change in the E-mule program on how many uploads to allow at once BY DEFAULT. For broadband users, a reduction in default max uploads at once would translate into what seems to be a download speed increase at the price of downloads becomming slightly more rare per search. Queues might grow in length, but tick down more rapidly because uploads would complete faster.

If someone has 10 KB/sec upload speed devoted in E-mule, and uploads are set to 1, then much of the time they're wasting upload bandwidth uploading to 56k users and overloaded 'broadband' users. If that same someone had uploads set to 10, then their uploads will fail occassionally and average LESS than 1 KB/sec in speed. NEITHER setting helps the network (or the user!) very much.

[jackc]

Anyhow, the issue is that if the newer version of emule DOES make d/ls faster (a big if, as you also mention correctly), then one of edonkey's (and Raza's consequently) great flaws diminishes. Raza is a great client, but the slow d/l speeds wrecks all its advantages; if this is successfully overcome, then other clients (i.e. KLite) will start looking a bit unnecessary on my PC...

[Switeck]

"Speed" is a relative thing. If you don't MIND, having 100 sources each going at 1 KB/sec makes for a great TOTAL speed but they're individually pretty poor.

And due to a client's overall percent-size of a network, for instance if Morpheus (which connects to Gnutella 1 and possibly Gnutella 2 and E-mule as well) decides to use absolutely horrible default upload settings -- like 20 upload slots and 'unlimited' upload bandwidth, then if Morpheus makes up ENOUGH of the network then it will seem that EVERYONE connected to that network suddenly sees lots of very slow downloads per source.

On the other hand, if E-mule's latest version uses FEWER upload slots by default, then if it was only a modest percentage of the network (or rather the networks Shareaza connects to) it may seem to disappear because overall it is offering few upload slots to the network relative to everything else. So you will have few/er downloads from them, but the few you do get will be better than in the past.

Assume there's 2 different users on otherwise identical broadband connections (with ~25 KB/sec upload bandwidth):
1 on Morpheus and allowing 100 upload slots
1 on E-mule and allowing only 5 upload slots

The Morpheus user will seem 20 times more prevalent as far as active downloads are concerned and offering average download speeds of possibly 0.25 KB/sec, thus giving the impression that 'Shareaza sucks horribly' despite the much better E-mule user. There would need to be 20 TIMES as many good users just to make up for that 1 BAD one.

In short, it is FAR FAR easier to sabotage a large network than it is to improve it using default settings for a program on the network.

The secret is finding the fast sources. Many of them are often firewalled and on corporate/university networks, so that makes them considerably harder to connect to.

[jackc]

Points well made, Switeck! B)