Eleri, if I may, you are right. But getting to the multiple shard situation (like UU) is almost turnkey compared to getting to the single distributed assets scenario hosting...everybody. The later will undoubtedly require a great deal of dedicated time, effort, and programing.

I realize I just posted, but I thought about this, and I can't imagine this working. I don't think the issues are technical - it's the whole business side. By "business", I include the fan run stuff. Also, without some paid humans to manage this, I can't imagine it working.

To me, the issue isn't the code (though the change management part is tricky enough, if you ask me) - it's the whole distributed system, distributed network that no one is managing. UU worked because everyone managed their own piece, their own server/shard. It seemed like just the right amount of ownership and control. I think the UU separate shards was a good model for fan run games. If a shard broke, that's the only thing that was affected. If there was vault corruption, it was only on one shard.

I tell you - if this works, someone can write it up as interesting business case, in terms of people, process and technology.

_________________
Retired forum mod, mszv, amarez in Uru (KI 89257), other online games
Blog - http://www.amarez.com, Twitter - http://www.twitter.com/amareze

I think in other threads it has been asked to the affect, what does Cyan get out of this by going Open Source?

Of course one, their URU and now the communities continues to have life. But beyond that would be something truly remarkable.

Imagine a game platform with truly global presence. If the distributed architecture for a game platform could be fully realized, imagine then a renewed Cyan, with a new epic game, or any other game company for that matter, it will be Open Source code, being able to place scalable servers in data centers in Europe, Asia, Africa, Australia, North and South America. All acting as and working as one super massive single entity shard. (Think more of a URU SkyNet)

This is implemented now with network databases where the GUI applications are local to the local user group and the database is on a central high-availability server platform (This is essentially what was in place with UU with separate shards/servers with Cyans central authentication server). If this could go beyond that to a fully and truly homogeneous and cohesive architecture for a game platform serving high quality graphics to hundreds and thousands of user at the same time (without lag), that would I think, be something.

And as far as change controls, no issue really, if one server goes down, a momentary pause as all the users running in the logical host on that server, which would contain the age or ages that are within that logical host, are shunted over as the logical host moves (Something that happens with clustered high-availability databases and applications now where the host server goes down and the logical host containing the database or application is moved to another server in the cluster).

Edit:

My point is, if individual servers in a distributed architecture could be configured to host one or two specific ages (not ALL of URU with every age), depending on the servers size, network bandwidth, etc., with other servers hosting other ages of URU, larger scalable server with A'egura(sp?), and these ages within logical hosts able to fail-over from one server to another should server hardware fail, and then be failed back, a truly solid distributed scalable super shard should then be possible.

Well, since this a game, with failover it won't be critical to preserve the transactions. If you lose some stuff, that will be OK.

You are talking about a kind of interconnection that is certainly possible, but is done (the way I've seen it) with network and system management, by people who get paid to do that sort of thing, using a decent (often not cheap) set of tools. This is a little different.

I'm also a bit concerned about the central database, the vault. Is that fan run? it's not that our fans aren't wonderfully smart - they are - but something you do in your spare time tends to be a little different from something they pay you to do. That's been my experience - maybe I'm wrong.

You are talking about a more tightly coupled system than I envisioned. It's interesting, but it's seems like a big leap from the UU days, when the only thing that was centralized, as far as I can tell, was the authorization server.

I don't know - I guess maybe the ages are separate enough that there's not that much interconnection. It still seems too experimental to me, but hey - I'm not doing it! I'm just along for the goodness of Uru, along for the ride.

_________________
Retired forum mod, mszv, amarez in Uru (KI 89257), other online games
Blog - http://www.amarez.com, Twitter - http://www.twitter.com/amareze

Last edited by mszv on Tue Dec 16, 2008 3:35 am; edited 1 time in total

Mszv, at this point you are correct, concerning who does the work, how it is done, and the tools. I know as well, I work with and support those kind of systems I mentioned. And as people now simply want back into URU, initially, what comes up is I think going to look very much like what the UU configuration was.

What I am stating would no doubt not happen easily or quickly. How easily or quickly cannot be determined until Cyan releases the Open Source and it can be investigated as to how pliable and modifiable it is and what they have already attempted to build in.

I suspect they have already considered and perhaps even worked into the code some elements of what I am speaking of. Without going back to check, I believe Chogon did mention scalability in regards to the code they are reworking for Open Source release.

I hope no one feels excluded here, since RW and I seem to have been in a discussion, but I really enjoyed the conversation.

You know, sometimes I forget - it a game - it's not a gazillion credit card charges done for millions of people. Perhaps this is a good venue to experiment!

_________________
Retired forum mod, mszv, amarez in Uru (KI 89257), other online games
Blog - http://www.amarez.com, Twitter - http://www.twitter.com/amareze

Well, I can't speak for the rest, but I've learned a ton by reading the exchange between mszv and RW.

Thank you both.

_________________
Stop hitting that key... there is no Escape.

Wow, 7 pages read, almost.

Jkla wrote:

COSMOS

D'mala was the cannon if I remember from UU, so I think that COSMOS
COMMUNITY
OPEN
SOURCE
MYST
ONLINE
SERVERS
Works well

MOOSE can also be suggested as the cannon server net name, just my thoughts

Should the MOUL forums continue to be the central location for all guilds?
If so, would Cyan still hold onto the domain name and not pas it onto one of the Admins here, if in the worst case scenario Cyan closed up shop?

One thing I like to make clear, that even though the code may become Open Source, I am estimating that it will be a few months before we believe that the server code is stable enough after modifications for you guys to play on it.

Does the open sourcing lead to the actual client code as well?
Reason for that is that, is that when we have Ages that are playable for the general populous, we would need a mechanism that would allow you to visit these Ages, can't use the Relto books shelf as that is part of "Yeesha's Power" to spawn books on the bookshelf "magically".

I'd personally like to see the Nexus used for managing player created Ages

_________________
Lian-Li PC-V1000 plus II case, 600W powersupply, Striker Extreme motherboard, Intel Core2 Quad Q6600 CPU, nVidia 8800GTS 512, 2x1GB Corsair memory @ 800Mhz, 5.1 audio, wireless mouse, 200GB main hard disk.
Grogyan KI# 00793984 -> www.guildofwriters.com

Eleri wrote:

Rusty_Russell wrote:

Eleri wrote:

But we could all still gather by the tent in Tokotah for dancing if we felt like it.

You're still seeing this as one implementation, Eleri.

And you're still seeing it as multiple. I have yet to see anything that insists it must be one way or the other.

Ok...

Quote:

We realize that this could turn UruLive into the "wild west" and lead to many fractured and diverse MystOnline servers.

That could support either view. (Not necessarily what I wanted ) However, I think technology is against your vision at present. In a word - lag. 35 people at most in Ae'gura in Prologue, 100 in MOUL... I think that "all" is a long way off.

Last edited by Rusty_Russell on Tue Dec 16, 2008 11:16 am; edited 1 time in total

Grogyan wrote:

I'd personally like to see the Nexus used for managing player created Ages

Me trying to imagine having to scroll down a long list in Nexus, click by click, and having
to remember the exact name and/or spelling or having to remember age descriptions

I remember looking for neighborhoods (limited to 50) and
how bad it was.

Other possibility:
A video library where one could see a small trailer about an age and
being able to click on a book for access to a "video trailer" or to link to an age ?

Last edited by Artic_Wagon on Tue Dec 16, 2008 9:43 am; edited 1 time in total

There is all kinds of stuff like this that we are going to have to face but until we have the code and decisions are pointless.

Speculation does keep the forums ticking over so while we are waiting speculation is not at issue however calls for decisions other than things like a name are moot.

Lets restrict ourselves to arranging the deckchairs till Cyan comes up with the code.

_________________

mszv wrote:

I'm also a bit concerned about the central database, the vault.

I have often wondered, pulling from an environment I work in, if some of the elements of large scale network management could not be applied here. For instance, in the network management world, the world of networks that underpin the WWW, there are multiple network element vendors. These are the folks that supply the large scale switches and routers that move the voice and data traffic around the world. The vendors (Nortel, Cisco, etcetera) provide their own element management application with its own database to manage their specific proprietary elements, this is known as an Element Manager (EMS). This works fine if all the elements in your network are from the same vendor. In the real world, they are not. So to present a whole world view of what is happening in the overall network to the folks who set in the control centers, another layer is added. A Network Management System (NMS). This system aggregates/collects the feed from multiple vendor EMS databases, that is, only the key information that is needed to give an overall view of the whole network state, not the full details about what is occurring in each vendors specific network element.

Okay, apply this to URU. Rather then a central vault/database, containing all the information about all the possible states from all the ages, each age would have associated with it its own specific vault/database, tracking as is performed now by the single central vault, the progress, location of objects, within that environment, this information would not be stored or transferred to the central vault/database. Only flags that are needed to express states in other ages/environments would be transferred to the central vault/database and much of that would occur only when you 'link' from one age to the next (this is needed for instance, so Bahro poles, books, etc., can appear and disappear at the Relto and in other ages). An exception example would be KI traffic which would always communicate with the central vault/DB.

I would be interested to know if Cyan (or others) have investigated along this line.

Again though, if mods are possible and doable, none of this is going to happen quickly, I do realize that.

RW, I think you have yourself a job, designing database(s) for a super-shard.

Crawl - walk - run - fly - ? [Eleri's vision for URU ]

Rusty_Russell wrote:

RW, I think you have yourself a job, designing database(s) for a super-shard.

Crawl - walk - run - fly - ? [Eleri's vision for URU ]

For me, it will be a crawl. But I am excited about getting a chance to look at openly (pun intended), what is under the hood of URU. I am sure others feel the same way and that with other items that have been discussed on this site in related threads and on other sites, could speed things up.

In UntilUru I had experimented a bit with distributed servers where instances of ServerAgent, GameServer and LobbyServer were run on different machines. Basically the ServerAgents registered with the LookupServer, AuthServer and VaultServer and spawned GameServers (Age Instances) and a LobbyServer where clients connected to. I remember someone from Cyan saying that would require fast connections and would probably be only possible within a LAN. My experience was that our setup worked for a bit but when anything went wrong because some of the UDP packets went missing, zombie servers were not automatically removed properly. And things got more complicated with NATed servers. I guess with the source available the servers could be modified to provide a better recovery from lost connections. Of course I don't know how the server code has progressed during the Myst Online phase.

Given the server structure is still the same this would mean a "Guild" shard maintainer would have to run the Lookup, Auth and Vault servers (and probably a data/web server) and Server Agents could be distributed on suitable machines, probably dedicated servers with at least 100Mbit Internet connection.

_________________
272 924