[J HOW]

RE: WHY EVERY SENSIBLE STAR CITIZEN BACKER SHOULD BE UP IN ARMS RIGHT NOW

Reading some of those comments are troubling, some are logical and rational criticism – then immediately attacked.  Reminds me of the cold war paranoia that there was a Commie under every bed, except instead of a Commie it’s a Derek Smart post.  Not saying that for comedic value either, some of the responses to criticism on the official forums sound non-compus mentis and very paranoid on any criticism of CIG or Chris Roberts.

The twisted thing is here; they are driving away a large portion of their own community by allowing, showing the attacks of forum members raising criticism.  Effectively it’s getting out of hand and the forum thread doesn’t appear to be getting moderated.  A self fulfilling prophecy.

To be absolutely honest, I was skeptical on “the white knight” parts of your argument originally (always look a gift horse in the mouth, verify sources, critical thinking etc) , however based on those forum responses I am struggling to disagree with you on many of your points.  The more I am seeing, the more troubling it gets, it’s beginning to look probable that this project will fail.  Also I get the feeling for every thing we see that is visible, there could be three or four things happening internally which are not visible.

[J HOW]

On the server side it would prove very difficult to scale up.  There are several problems when running instances on nodes – whilst a node may have many instances, an instance may only have one node. It’s a trade-off between the amount of instances on a node and the memory available to the instances.

i.e.
If an instance requires 2GB
6 instances, 10GB plus OS overhead would be 12GB per node
Data is not truly shared between instances or nodes (no replication) therefore it’s a architectural limitation

Without making architectural changes it would be very difficult to increase unless, you scale up the nodes (being virtual machines it is possible) but it would be prohibitively expensive.  If they require changes it will come at cost of changing the architecture server side, it’s not very easy to do and it may require code changes on both the server and the clients.

Looking at the problem what you’d do to overcome this would be to use a reverse proxy (if you’re pushed on time, money or resources) to interface between the server (the active node you’re connecting to) and the client. This way you could potentially create a pool of instances that would act as one – although data wouldn’t really be shared between instances. To overcome this a method would be to allow information to be shared via a service bus between instances, although that would be difficult to implement and potentially cause sync issues between instances (which one is master or are they all acting as master?).

The ideal solution is to re-engineer from the ground up to allow one instance across nodes – budget and time permitting.  At the time I was not aware of the 64bit implementation, depending on how it’s implemented it will probably add more than a couple of months to development time to re-engineer although I’ve never worked on 64bit specifically to comment on the impact.

At the time they were using Google Compute, there are methods of making it “appear” to be one large instance where information is saved between nodes by connecting at the backend with a permanent SQL cluster (where information is saved) and then loaded between instances.  Effectively making the SQL Server the single source of truth for both account information and character (ship, player identity, cache, items), from notes they only use SQL Server for account information and instances refer to flat files or SQL lite for configuration.  Now this being said using SQL in this way does have overhead, and would slow down calls from instances, too many instances then from the player perspective it would get slower and slower – unless it was clustered or over many SQL servers in a cluster however this massively increases OPEX costs.

It’s a tough one to solve, usually MMOs wouldn’t have this issue as this would generally be thought about and planned before implementation.  The risk here is when you are changing architecture mid development, it has risks and costs. Cost of redoing work, code refactoring, multiple levels of complexity, it rather compounds their current problems and amplifies them.

Personally interested to see how they handle it, no doubt they do have some smart people working for them so it’s an interesting one to watch.

[J HOW]

You’re right beginning to think I did luck out after reading through these threads.

I applied around earlier in the year, then got a reply so around March – May which is when they were recruiting. I applied to the UK but some of the roles were in the US hence the conference call – Ops was UK according to notes but servers were managed in the US. Yes it was a woman manager who was having issues with the way I spoke. Sorry can’t remember the name and don’t have it in my notes either.

Can’t comment but I have read on Sandi, don’t know if it was the same person.

If it helps on my notes it mentioned the problem they were having was server instancing specifically load balancing and fault domains. Basically in the virtual space – if a Virtual Machine died then it would take out that instance or node – or how many nodes were on the virtual machine. This was the technical conversation around the conference call. I suggested a DNS round robin approach and then a load balancer for each node group – all connected to a SQL cluster backend so it did not matter what node it was connected to then all instances would be available as long as they were all acting on the software layer. The point was to make all nodes mutli-session aware in each node group – as in able to talk to each other and the SQL DBs then get around the instancing problem as all nodes could share data freely in a mesh configuration for that node group. Which is how it was done in a few other smaller MMOs.

From the notes it’s not what they were looking for as cost factor was raised as an issue and the fact the server side software did not support certain required functions to make it possible.

As I understand it, currently it’s still an outstanding issue with their Persistent Universe.

[J HOW]

Have been following Star Citizen for a very long while, admittedly not backing the project but rather looking with interest at how it was progressing.  Normally wouldn’t post here however it does look in this instance that Derek Smart does have several valid points.  For clarity would like to explain why he may be on the mark about employment at CIG.

Will caveat now that this is my personal opinion based on my experiences. They could be completely off the mark.  No malice intended or indeed making any waves.  It may be that my experience was a one off, however the more I think about it, the more it seems to make sense based on what Derek Smart has said.

I had applied for a role within CIG before the media storm, the escapist article and before a lot of what we see now.  Being in the gaming industry at large albeit in a very small way compared to some developers.  Applied online for a role in CIG, which frankly did not expect a response, at the time I was looking for work so thought it would be worth a shot.  Applied and got a response from HR, then received a call asking for follow up questions.  Initially they said they would wait for interview if they were interested, great I thought.  I then received a second call on the more technical aspects, things were going quite well!  About two weeks later is when things went a bit pair shaped when I spoke to one of the managers who seemed to imply on how oddly I spoke English.  At the time it was “sorry”, “pardon”, “could you repeat that”, now my English may be accented but normally everyone can understand me.  Had a conference call based on my technical capabilities, now it was the choice of words which seemed to imply heavily on following instructions and if I could communicate with the team.  I asked a lot of questions on their server setup, their architecture and their support services as you do, in that call it seemed to be focusing on what I could do from my technical understanding until some follow up questions on my language skills.  After that I heard nothing and of course did not get any position.  This may come across as sour grapes however even at the time I noticed it seemed to play heavily throughout the conference call on my language skills – which are fine by the way.  Something just seemed “off” about the whole experience.

I had thought that was the end of it, and ended up in employment at another gaming company.  Now this is where it gets weird, directly after the media storm, escapist article CIG did contact my references – even though I didn’t get a job after the interview process.  This was several months later so have no idea why they asked for them without asking me first.

Now not sure if that relates to Derek’s understanding of HR at CIG.  For the references I assumed they had simply made a mistake, but the conference call itself seemed very odd.

[Author]

Posts