News?

[Rotten Luck]

Sounds like a plan for us.  If we ever get that far.

[Victoria Victrix]

Thanks for that.  I guess Zwill must have been referring to them being in a different building.  I know he mentioned something about not having direct access to the machines physically, and that they login to them remotely, much as we do.

I'm often surprised that so few games have emulated Google's model - a cluster network that allows the servers to be always on, with no downtime.  They can literally update machines from the cluster on the fly, and at the halfway point, switch from the half running the old version to the half running the new, and then finish updating the other half.  The idea of single, large-scale server machines just seems so terribly outdated by comparison.

Zwill and the rest of the Paragon Studios team were just outside San Francisco (in a community MUCH cheaper to live in and rent office space in than the Bay Area proper).  When we paid them a visit it was something of an epic drive to get to them, in fact.  There was a lot of "you can't get there from here."  The servers all live in the NCSoft(core) building in Austin Texas.

[Valjean]

Then having them here and under ownership.  It does seem like the plan is to drop CoH so they can use those Servers for Brest and as... I mean Blade and Soul.

I don't think that's the case. My understanding is that the server architecture for COH is very dissimilar to the rest of NCsoft's games. They couldn't be easily repurposed for Aion or B&S.

[Victoria Victrix]

I don't think that's the case. My understanding is that the server architecture for COH is very dissimilar to the rest of NCsoft's games. They couldn't be easily repurposed for Aion or B&S.

You are confusing hardware and software.  A server is just a computing box.  The software for the servers to run CoH is very different from Aion or Boobs and Shame.  However, all they have to do is archive the CoH server architecture to a hard drive to save it, wipe the server hard drive, and load new software, and bingo, it has become a Boobs and Shame server.

[Valjean]

You are confusing hardware and software.  A server is just a computing box.  The software for the servers to run CoH is very different from Aion or Boobs and Shame.  However, all they have to do is archive the CoH server architecture to a hard drive to save it, wipe the server hard drive, and load new software, and bingo, it has become a Boobs and Shame server.

No, I'm not confused. The non-COH games run on a dissimilar physical infrastructure than the COH games. All of the NCsoft games are run on architecture unique to each game.

Additionally, NCsoft's migrated most of their non-COH game servers to third parties anyway in the West (Aion/Lineage).

[Sleepy Wonder]

No, I'm not confused. The non-COH games run on a dissimilar physical infrastructure than the COH games. All of the NCsoft games are run on architecture unique to each game.

Additionally, NCsoft's migrated most of their non-COH game servers to third parties anyway in the West (Aion/Lineage).

You are in fact, confused. If not, perhaps misinformed is a better word. The architecture you speak of has nothing to do with the servers.*

* The lower level code in the server's code base modules may however have architecture dependencies with regard to the platform it was designed for, but for major MMO's in general, as a programmer, this layer is often obscured from the rest of the code in order to prevent dependencies.

For example, MySQL and Oracle are two different database technologies. If I was designing interfaces for other programmers, they shouldn't necessarily have to worry about which database is being used. My interfaces I create are all the same (store(x,y). All that would be required should a database change occur is a switch for those interfaces to use a different method behind the scenes.

With regard to CoH and other MMO's, the server doesn't (and shouldn't) care what type of CPU's are being used, how much memory there is, or what sort of network infrastructure there is. This is a general statement, of course, in reality if you want to get technical the server would throw a fit or take other courses of action if the OS told it it was out of memory.

Basically, should you care what type of vehical you are in so long as it gets you to point A & B to do your job?

[Valjean]

We'll have to agree to disagree on this, particularly since it's not really germane to the overall discussion.

[downix]

You are confusing hardware and software.  A server is just a computing box.  The software for the servers to run CoH is very different from Aion or Boobs and Shame.  However, all they have to do is archive the CoH server architecture to a hard drive to save it, wipe the server hard drive, and load new software, and bingo, it has become a Boobs and Shame server.

This is not exactly true. Servers are not uniform. While low end units use similar hardware to desktops there are ones with radically different hardware that makes the desktop compatible units look like tinker toys. POWER, SPARC, PA-RISC, CELL, you find archetectures here that are completely incompatible with each other.

[TonyV]

Comparatively speaking (to other expenses), servers are dirt cheap, especially since the infrastructure runs on virtual machines.  They are, for the most part, a one-off cost that is a fraction of what companies pay in support salaries and such.  I really, really believe that repurposing hardware had absolutely zero impact on the decision that NCsoft made.  It would be like making a decision to get rid of your car based on your air freshener getting a little stale.  We're talking about a few thousands of dollars in projects whose budgets are millions.  I'm not saying that it's completely unimportant, but I can pretty much guarantee you that at no time did someone use "...and we can use the infrastructure for City of Heroes on Blade and Soul instead!" as a bullet point to justify killing off a game that was making orders of magnitude more money than that infrastructure costs.

[Codewalker]

This is not exactly true. Servers are not uniform. While low end units use similar hardware to desktops there are ones with radically different hardware that makes the desktop compatible units look like tinker toys. POWER, SPARC, PA-RISC, CELL, you find archetectures here that are completely incompatible with each other.

Since most of those are being desupported by just about every major vendor, they're the ones that look like tinker toys.

PA-RISC has been dead for years. It was slated to be replaced by Itanium (with HP being the only vendor still using that even), but that's more or less dead in the water due Oracle's announcement that they're pulling Itanium support for all their software. There's a few die-hards still using it, mostly in the US Government, but eventually those systems will be replaced.

SPARC is floundering, despite Oracle being desperate to justify their purchase of Sun and pushing it hard. It's clear they're losing the war to cheaper systems that can be clustered to provide greater performance and comparable reliability.

POWER is probably the only of those still seeing serious development, but you won't find those outside of an extremely high-end OLTP setup in a bank's mainframe, or in Cell/Xenon CPUs on consoles. Even Apple, once the champion of the architecture, has abandoned it.

I'm in the process of migrating a large financial system off PA-RISC onto x86-64 for my day job. Each of our database cluster nodes has 32 cores and 512GB of RAM, hardly what you'd call low-end.

It's possible that B&S or other games use a different architecture for their server farm, but given how commoditized the server market has become, a developer would be foolish to design such a system.

Also, since we have it on pretty good authority that the COH servers are virtualized, they could very likely share the same hardware with Blade & Soul without an issue. The only question is that of available CPU/memory resources, and IOPS on their storage units.

[downix]

Since most of those are being desupported by just about every major vendor, they're the ones that look like tinker toys.

PA-RISC has been dead for years. It was slated to be replaced by Itanium (with HP being the only vendor still using that even), but that's more or less dead in the water due Oracle's announcement that they're pulling Itanium support for all their software. There's a few die-hards still using it, mostly in the US Government, but eventually those systems will be replaced.

PA-RISC should have been axed instead of Alpha with the HP/Compaq merger, IMHO.

SPARC is floundering, despite Oracle being desperate to justify their purchase of Sun and pushing it hard. It's clear they're losing the war to cheaper systems that can be clustered to provide greater performance and comparable reliability.

Contrary to popular belief, Oracle is not the largest supplier of SPARC machines. Several European and Japanese firms produce them, with the largest vendor of SPARC hardware Fujitsu. Being an open-source platform (several versions of the chips are available under the GPL in fact), there are now more SPARC vendors today than even 5 years ago.

Sun had not been top dog in SPARC since the 90's.

POWER is probably the only of those still seeing serious development, but you won't find those outside of an extremely high-end OLTP setup in a bank's mainframe, or in Cell/Xenon CPUs on consoles. Even Apple, once the champion of the architecture, has abandoned it.

But Microsoft embraced it, and POWER is now at the core of their future development. Ironic. (And Apple never used POWER, they used PowerPC, which was based on a very old form of POWER)  Sony has also been migrating to POWER as well, with their CELL architecture.

I'm in the process of migrating a large financial system off PA-RISC onto x86-64 for my day job. Each of our database cluster nodes has 32 cores and 512GB of RAM, hardly what you'd call low-end.

It's possible that B&S or other games use a different architecture for their server farm, but given how commoditized the server market has become, a developer would be foolish to design such a system.

Also, since we have it on pretty good authority that the COH servers are virtualized, they could very likely share the same hardware with Blade & Soul without an issue. The only question is that of available CPU/memory resources, and IOPS on their storage units.

This is quite probable. I was just pointing out that it was not guaranteed to be the same hardware is all.

[Codewalker]

PA-RISC should have been axed instead of Alpha with the HP/Compaq merger, IMHO.

I can't say that the end result was necessarily a bad one, though (for everyone except HP anyway). Some of the ex-Alpha team ended up going to AMD and designed the K7, which pioneered many of the fundamentals used by CPUs today. It gave Intel much-needed kick in their complacent butts and the resulting CPU war led to technology improvements coming much faster than they would have otherwise IMO.

[downix]

I can't say that the end result was necessarily a bad one, though (for everyone except HP anyway). Some of the ex-Alpha team ended up going to AMD and designed the K7, which pioneered many of the fundamentals used by CPUs today. It gave Intel much-needed kick in their complacent butts and the resulting CPU war led to technology improvements coming much faster than they would have otherwise IMO.

The ex-Alpha team which went to AMD were from the Digital-Compaq merger, not from the Compaq-HP merger. Compaq hired a new team to replace them, and that team eventually went to Intel with the HP-Compaq merger, and the i7 is their baby.

My father worked for Digital, I knew most of the Alpha team.

[Codewalker]

Oh yeah, that's right, I forgot about the Compaq acquisition. That was a long time ago...

[downix]

Oh yeah, that's right, I forgot about the Compaq acquisition. That was a long time ago...

I could talk for hours about the Alpha, and how Compaq messed it up badly.....

[Codewalker]

What didn't Compaq mess up?

I still remember when HP announced they were buying Compaq, thinking what a terrible idea it was.

[downix]

Did you know that one of the tricks Digital used to get the Alpha's performance so high is that they tweaked it around the fab machines they used. It was to the point that *only* those fab machines could produce the chip without performance loss.

One of the first thing Compaq did when they took over was sell off those machines.

That's why the Alpha dropped in performance suddenly.

[dwturducken]

Do we need to give you two some time alone?   

[Victoria Victrix]

GET A ROOM!

 

[downix]

Hello, I'm Nate, and I'm a Geek.....