Archived

This topic is now archived and is closed to further replies.

ButterFly.Net - The Real Deal or Hogwash?

This topic is 4979 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I read the ButterFly.Net news article and even browsed their website and they make some pretty outrageous claims - namely that grid-enabled MMOGs will "scale infinitely", have low latency, and have "absolutely no cheating possible". I personally think it''s hogwash. I''ve interned at IBM for the past two summers working on their OGSA (Open Grid Services Architecture) toolkit, so I know something about this grid computing idea. While I will allow that there may be some sophisticated security scheme one could implement that will prevent the worst kinds of cheating, the idea of using the grid to support a MMOG is ludricris. In my personal experience the data transfer latencies on the grid are incredibly high - not suited to anything remotely realtime. This fact also makes me question the "infinitely scalable" claim. My intent here is not to start a flame war, but rather to educate myself. Has anyone given the technology a serious look and determined that it might deliver on these crazy promises? ---------------------------------------- Let be be finale of seem, seems to me. ---------------------------------------- Coding: http://www.stanford.edu/~jjshed/coding Miscellany: http://www.stanford.edu/~jjshed

Share this post


Link to post
Share on other sites
They allegedely experimented with making Quake II into a massively multiplayer first person shooter, and it died a horrible death. I don''t have a reference for that, though.

When you say "incredibly high" latencies, does this mean 10 ms? 100 ms? 1,000 ms? A 100 ms latency server-to-server might be a little much, but a typical modem will add more than that, and you have to live with that. If it''s 1,000 ms, then forget it :-)

I also think the big draw of the Grid is that the data center, hardware, and management infrastructure is already there; you just buy some capacity and go. When you need more capacity, just call ''em up, and it''s online. Having seen the process of building solid data center practices and management from scratch, I know that takes a while.

But, if the price of that data center availability is using technology that doesn''t fit your project, it''s probably not worth it.

Share this post


Link to post
Share on other sites
I dont know about butterfly, but there is Second Life, another MMORPG which uses the grid concept. There is a write up about it on Gamastura. The concept is sound enough, there isn''t much traffic between grids so higher latency shouldn''t be too much of an issue. Its equvalient to jumping around zones, a small load dealy and your in a new grid node.

-ddn

Share this post


Link to post
Share on other sites
There''s a difference between "the Grid" and "a grid concept" though.

Modern games (Asheron''s Call, Asheron''s Call II, There, etc) allow arbitrary travel between the different cells without any loading delay. You absolutely need fast communication between neighboring servers for this hand-off to work well. I''m pretty sure these games don''t claim to be "on the Grid" but they''re using data centers and machine clusters to serve a single world anyway.

Share this post


Link to post
Share on other sites
Looking at the "technical" docs, it looks like a classic 3-tier architecture running (a bit specialised) application servers.

There''s nothing revolutionary there, apart from the fact that they seem to plan on hosting several games on their existing server farm (a sort of akamai-like strategy)

Share this post


Link to post
Share on other sites
Read the gamastura article I mentioned. It specifcally talks about the concept of having server nodes in a hardware grid configuration, with each node controlling a well defined NxN sector of the map. I know some MMORPG do this in software and dynamcially distrbute the load across serverlets, but that requires enourmous amount of cross networking capacity between serverlets which is usually wasted most of the time, as how often can a server reconfigure itself without introducing horendous lag?

Here is the article, you''ll need a Gamastura account to access it through i think.

http://www.gamasutra.com/resource_guide/20030916/rosedale_02.shtml

Good Luck!

-ddn

Share this post


Link to post
Share on other sites
In my personal experience Grid latencies can be in the range of 1000 - 10000 ms, depending, obviously on a large number of factors.

I was under the impression, from the ButterFly.Net article and website that ButterFly.Net was a toolkit for The Grid, not some in-house network called a Grid. If it''s actually just a regular server farm, their claims aren''t so crazy - but neither are they doing anything particularly impressive. Maybe they are just cashing in on the "Grid Computing" buzz word for free publicity?

----------------------------------------
Let be be finale of seem, seems to me.
----------------------------------------

Coding:
http://www.stanford.edu/~jjshed/coding

Miscellany:
http://www.stanford.edu/~jjshed

Share this post


Link to post
Share on other sites
I also believe Butterfly.net is trying to run on IBM''s Grid. What I said was perhaps Grid-like was the mention of Second Life. I''m pretty sure SL does NOT use the IBM grid.

1,000-10,000 ms latency? You have to work HARD to make it that bad! That can''t be true... can it?

Share this post


Link to post
Share on other sites
quote:
Original post by hplus0603
1,000-10,000 ms latency? You have to work HARD to make it that bad! That can''t be true... can it?



Not really, I''ve worked a short time on Websphere/IBM platforms for an insurance company, and it really ran like a dog.

Share this post


Link to post
Share on other sites
Hmmm... I''ve been thinking about the problem a little bit more and it seems that a Grid-enabled network middle ware would be kind of like using p2p for your network communication, only the clients are not directly connected and thus the latency is always higher (or almost always). So if p2p games thus far have not worked well, there''s no way that a Grid-enabled approach will. Am I wrong?

----------------------------------------
Let be be finale of seem, seems to me.
----------------------------------------

Coding:
http://www.stanford.edu/~jjshed/coding

Miscellany:
http://www.stanford.edu/~jjshed

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
i''m the chief software architect at butterfly.net: saw this
thread an thought i''d weigh in ;-)

we''ve applied distributed computing principles to a mesh
of servers mediated by gateway proxies that route your
game packets to the server you''re currently "on". using
low level proprietary protocols for "in-band" game traffic
to minimise latencies end-to-end, we allow you to stand
on a server in one region and (via boundary proxies called
sentinels) receive packets from other nearby players
even across server-boundaries.

overall the engineering effort has been to reduce the overhead
involved from roughly n-squared to n * log(n) for large numbers
of concurrent users distributed over the server grid.

application of higher level standard Grid protocols is reserved
for configuration, commissioning, administration, logging, billing, and monitoring of game systems (including reconfiguration if necessary).

there is more "secret sauce" to the Butterfly Grid than
mentioned in this summary... per packet strong cryptographic
authentication and security... server side rules enforcement...
remote script invocation... distributed artificial intelligence
to control non-player characters. but all the extras are layered
on top of an efficient working protocol.

bart whitebook

Share this post


Link to post
Share on other sites
Hi Bart, could you provide some performance metrics in terms of latency. Say an estimaition of the transactional overhead of reciving a packet, processing it within the main executable loop(s) and bounching back the packet to the client, a simple ping. Would it be possible to create a FPS using Butterfly.Net?

-ddn

Share this post


Link to post
Share on other sites
Hey Bart, ddn3''s post is extremely relevant and I''m interested as well in your performance metrics. ;-) BTW, you should update your "Games" page to remove some of the projects that were canned/put on hold since you put that page up last year...

Sylvain Beaudry
Quazal Technologies Inc.

Share this post


Link to post
Share on other sites
I spent some time reading the Butterfly net site + technical docs some time ago. Its all very impressive, but my technical knowledge of Networking and Grid technology isnt up to the standard to be able to make any judgements about it.

Since then, I''ve been waiting with baited breath for a Game release using the technology to see if it really delivers what it promises. I''ll be damned impressed if it does what it says it can, and I''ll be seriously considering some start-up ideas of mine that would not be possible without such technology.

Jon

Share this post


Link to post
Share on other sites