Essentially the game is a database. Our characters are constantly requesting information from the server for various things which include our inventory when we look at it and our achievements for times we look at it or other things that require achievements.

By reducing how much information that queries go through it reduces the load on the server which in effect improves performance.

Will this be a fix-all for the performance? No. We probably will not notice it. But with multiple improvements across different areas, some of the issues we experience due to the server load will improve.

Beyond that, I would not expect more from Zenimax. They stated their answer in the Q&A and doubt they will find a need to expand on it.

Amottica wrote: »

Essentially the game is a database. Our characters are constantly requesting information from the server for various things which include our inventory when we look at it and our achievements for times we look at it or other things that require achievements.

By reducing how much information that queries go through it reduces the load on the server which in effect improves performance.

That's not how databases work, and that's not how achievements are implemented. First, they are loaded when you login and updated with events, not every time you open the window. Second, it is infinitesimal compared to all the other data that moves around (hello guild sales!)

On the database side, there are cache layers, multiple indexes, backend optimizers, different kinds of optimized views... there are so many facets to a good database design that simply deleting a few records is like throwing a pebble into the ocean. And any database like this will be monotonically increasing, so this is, at best, a pebble thrown into an ocean that's rising (hi, melting ice caps!)

BlossomDead wrote: »

It is a matter of querying the databases per character rather than per account which is more consuming in terms of resources. Nothing more complicated than that.

It eases bit the load on the backend and probably will speed up char load. Multiply that with millions of accounts and it might be a significant gain.

This is absolutely untrue, unless you are just terribly bad at software design. Have you considered that those same (thousands, not millions) of accounts doing hundreds of combat actions resulting in both server checks, processing and logged (!!!!!!) PER SECOND. How often do you think a player gets an achievement? I would it's no more than 2-3, probably less, per player per day. It's straight up lies.

Kaspy wrote: »

I would take a guess that this is pretty much because they are moving account and character data to some kinds of 'cold storage' options, which means much lower storage but higher I/O (read/write) costs. So reducing the need for I/O's means overal improvement of 'performance' costwise. From the point of view of individual players this probably won't yield any measurable benefits performancewise.

Cold storage is the exact opposite. It means that data is being semi-archived when a user is AFC (away from account) for a certain amount of time. When the user logs in again, that data is moved from cold storage back onto "hot" storage so it can be accessed more quickly (and expensively).

The performance increase comes simply from having less data and smaller/cleaner indexing on the hot storage instances, which means the can be more consolidated, both physically and logically.

remosito wrote: »

nightstrike wrote: »

spartaxoxo wrote: »

First. This is not a thread to discuss AWA. Please do not discuss it at all here. There are multiple threads that already exist for that. This is also not a thread to discuss whether or not these changes are worth it. That all belongs in the existing AWA threads.

This is a question mostly for the dev team, though anyone is obviously free to discuss perfomance alone in this thread. What exactly are the performance issues this change intends to solve? What will this allow you to do in the future? Does this have to do with console support? What do you mean it allows the game to be performant in the future as the most important aspect?

I guarantee the performance difference is negligible. It's just not how databases work.

it actually can be how a database works.

read access is sped up considerably if data is kept in memory. the bigger the database the less of it fits in memory (for same server/amountofram)..

same for indexes and memory..

if we just look at achieves db. in my case reduction of data is 90%. 10 uptodate u33 chars..

10chars->1account..

It would only do that if they loaded every character's data into memory when you log in every time with any character, which is crazy. Even if you needed the data from other characters, they could cache it client-side and read it from there as that data is contextually static.

I guess the things that I don't really understand from the perspective on getting performance gain, are things like...

"Number of achievements" were mentioned, and adding more, so from a lines in the database existence they are all still there. Character and date would be more of an attribute, no? So how would this even help with the size as the achievement still needs to exist to be able to get on the account? Every character and the UI and everything all still HAVE the line there.

Same thing with logging it or having it pop up. The game must have to check if the achievement is blank or not when the action is done in game... right? Anything that still IS an achievement will need that checkpoint, which means it will be read the same amount of times as now.

Plus, the data still needs to be logged somewhere if the Zone Guide is supposed to be the solution for tracking progress, so... where is that data living?

Seems to me like maybe it would have been better to completely remove certain types of achievements to address the issue. For example, one of the #1 achievements mentioned as "grindy" are the trophy ones. Remove those and give anyone who got them some crowns or something. Achievements that acknowledge stepping into a DLC. I like the "kill x number of x" in dungeon achievements, but maybe let's get rid of all of those, if it is an issue with having to watch for each kill. If we're deleting data, anyway...

Completely serious, btw. Wouldn't deleting some achievements altogether, and phasing those type of achievements out forevermore, be much more helpful in terms of increasing/maintaining performance and allowing more to be added?

Saieden wrote: »

remosito wrote: »

nightstrike wrote: »

spartaxoxo wrote: »

First. This is not a thread to discuss AWA. Please do not discuss it at all here. There are multiple threads that already exist for that. This is also not a thread to discuss whether or not these changes are worth it. That all belongs in the existing AWA threads.

This is a question mostly for the dev team, though anyone is obviously free to discuss perfomance alone in this thread. What exactly are the performance issues this change intends to solve? What will this allow you to do in the future? Does this have to do with console support? What do you mean it allows the game to be performant in the future as the most important aspect?

I guarantee the performance difference is negligible. It's just not how databases work.

it actually can be how a database works.

read access is sped up considerably if data is kept in memory. the bigger the database the less of it fits in memory (for same server/amountofram)..

same for indexes and memory..

if we just look at achieves db. in my case reduction of data is 90%. 10 uptodate u33 chars..

10chars->1account..

It would only do that if they loaded every character's data into memory when you log in every time with any character, which is crazy.

that wont help with indexes. you really want those in memory. a db 5x in size will have much bigger indexes..

as for client side... you dont really want client side caching for gating stuff.. mmos use server authoritative for a reason....

client side data just really aint safe from shenangians...

ajkb78 wrote: »

It shouldn't make any significant difference except for a small amount of storage (say 15 million accounts, average 5 toons, O(1000) achievements so around 75GB even if they really inefficiently use 1 byte per achievement). As far as data throughput goes it's the same. There's only ever 1 toon loaded in per account at a time so it makes no difference whether you're updating 1 account-wide achievement data blob or 1 character-wide achievement data blob. If there's a performance aspect to this, it probably means some aspect of data storage is being done fundamentally wrong. It would be better to separate the 2 issues, roll data architecture fixes into the general "year of fundamental fixes" plan and hold off on AWA until a future update later in the year when a better approach can have been defined, preferably with more consultation about what the player base actually want from AWA (I guarantee its not loading a new toon to find all your maps marked as complete with no way of telling what delves etc. that specific toon needs to complete just as one example).

achievements have timestamps... much more than 1 byte. plus counters obviously are more than one byte too. as are multipart progress ones. at least the way they currently encode them.. from memory when I did some custom adaptatios to kyomas global achievements.. plus charname resp. charid, achievenr., accountid depending on how they built their db tables..

I don't know anything about all this database design and caching stuff, but I ran into this in the feedback thread and thought it was probably relevant.

From Mr. Firor's post Update-26 launch update:

ZOS_MattFiror wrote: »

Now that the dust is settling from the launch of Update 26, I'd like to take a moment give everyone an update on what's been happening in the last week. Obviously it was one of the more bumpy releases we've had, and it's important that you all know the issues we found, what we did to fix them, and in general how things are going.

[snip]

The second problem we encountered was much more serious. Update 26 includes many new achievements, especially with a new Chapter and the new Antiquities system. This then combined with an issue where character records saved far more often than intended. These two factors together meant that save times grew over time, backing up all character loads and saves, which led to infinite load screens, timeouts while zoning from one place to another, etc. It took us a while to find the root cause and fix this one, but by Wednesday afternoon (after another maintenance that morning), it was fixed.

Now, at first, I assumed that the performance improvements sited as part of Update 33 were part of the PvP rewrite of foundational server code, but what if they aren't? If achievements still have the power to cause stability problems, and if a rewrite could have solved these problems by adding caching or whatever, then it makes sense to assume that they aren't rewriting achievements code, but rather just trying to reduce strain on a system that has already proven to have problems with achievements before.

I probably wouldn't have tried to sell that as "performance", personally. Sounds more like "stability" to me. Kind of like "the rewrite still hasn't happened, and we need to cut something in the meantime if we don't want our servers to melt when the next chapter is released."

wenchmore420b14_ESO wrote: »

There were some responses in one of the other threads from people who actually work in Programing and Data Coding in RL who broke down the technical aspects of what ZoS was claiming on performance increase , but going back to find and quote them, I found they seem to have been removed.

I invite anyone who does coding and programing as a living to post your opinions about this. Thank You!
Just my 2 drakes.

Wouldn't that just invite them to be removed again?

The truth is, the only people who really have all the information are ZOS, and they say "performance". Whether that's true or not, I'm sure the technical details would bore 99.99% of players, and just annoy the other 0.01% who think they could do it better.

Without some substantial improvement to the player experience, nobody cares.

They could just as easily do the Star Trek script thing and <insert tech speak here>, and it wouldn't be any less satisfying to hear.

Zephiran23 wrote: »

I'm confused how this actually helps performance. As an example:

Character A has completed 90% of Daedra Slayer, characters B-D have 2-3% each so there's only a small, <1% to go. Player creates a new character to play through Deadlands. Lots of deadra in Deadlands, even more if you can't ride away from them or ignore them with Shadow Rider. In order to complete the total number of kills/account for the achievement, the server now needs to keep a running total of kills on A-D as well as the new character E.

The server also needs to keep checking for other Slayer achievements as well, for all the Waking Flame acolytes etc and whether those count for daily or weekly Endeavors - which would seem to be a separate check added relatively recently.

I doubt it will help performance, but just to try and make this clearer:
Today, if you look at your achievements you'll see the numbers you said. After the update you'll only see 93% completed. The game won't be storing those hits separately.

SirBedevere wrote: »

They could just as easily do the Star Trek script thing and <insert tech speak here>, and it wouldn't be any less satisfying to hear.

The technical term is actually "technobabble"

I'll go hide in a corner now :P :P