Bitcoin Forum

Most Online Today: 1828.

The last time so many people where online as Today was on 2011-06-08 !!!

See: https://bitcointalk.org/index.php?action=stats

Is this thanks to Bitpay being at the CES?

Ohh, and BUY,BUY,BUY!!!

This was Top of Hacker News for a while earlier.
http://news.ycombinator.com/item?id=5043122 (http://news.ycombinator.com/item?id=5043122)


Would have driven a lot of traffic.

Explains why I can't load anything.

This development is certainly welcomed and I'm sure CES has something to do with it.  Hopefully the conspiracy theorists don't scare too many of them off with their wild rants :D

Why the hell is the forum unable to cope with the traffic given how much has been donated to upgrade the software?  Seriously, the software upgrade seems like vapor-ware at this point.

I've been struggling with loading pages too and had been wondering why.  I'm not meaning to detract from what repentance is saying but I guess an inconvenience with forum response times is a small price to pay for the additional interest and subsequent impact on the growth and spread of Bitcoin :)

That explains everything

Repentance, software update doesn't matter when a lot of ppl use a website, for so many ppl we would need better hardware, not software

the forum needs to be running on http://www.litespeedtech.com/ instead of apache

which i have offered up a license 100% free for use on the forum (i'd even pay the support/updates contract)  (for sale in this thread https://bitcointalk.org/index.php?topic=122271.0)




no. first step is replacing apache. then adding hardware if it is still a problem



put nicely apache is retarded in memory and cpu usage

OK, but are we looking at a huge amount for better hardware?  It seems crazy to be collecting thousands of dollars for better software if the hardware still can't handle more than 2,000 people online.

Whether it's hardware or software the forum needs, funds have been collected and at some point a decision to spend them on upgrading them needs to be made.

+1

Wow, this does bring back flashbacks to June ....

now I know why the site's running slow.

Won't get too excited until we beat the all time record...

CES baby!

http://www.theverge.com/2013/1/11/3864592/bitcoin-punks-go-corporate-at-ces

8-CPU Lease LSWS Enterprise
8-CPU Leased license of LiteSpeed Web Server Enterprise Edition
$92.00 USD Monthly
$999.00 USD Annually

Seeing as these forums are making more than this weekly, seems like this should be whats purchased.

I believe that the bottleneck is the database, not the web server.

no reason.

even if the server has 8 cores. the two core version i am offering FOR FREE (i own a ton of them for my past business that are sitting unused) would work perfect

a server with 8 cores would run the web server on two and php/mysql on the other 6. the litespeed core license is just for the webserver itself php/mysql and such will still use all cores


i ran 1.5gbit/sec of entirely php (no static) traffic through litespeed for years

well how to fix that?

if that's the case

is it mysql or something similar?

run it on a second server connected to the first over the local network of the datacenter


and use ssd drives on the database server

If it's just the database, you could always move to a MySQL cluster set up. Dedicated web server upfront (I like nginx and no license needed, still better than apache) and then the cluster sits behind and databases away.

How bottlenecked is the database? I only see it slow when slashdotting or something like this happens.

It may have something to do with the fact that Moot accepts Bitcoins for his 4chan Pass and says so under every post form. Thus millions of people see the word Bitcoin everytime they write a post on 4chan. CES is nothing compared to that.

More memory, faster CPU, faster disk. Maybe the mysql configuration could be improved, but I don't know much about that. mysqld seems to be CPU-bound.

The traffic will probably go back down soon. If not, I'll find a solution.


It's not much better (if any better) than the best open-source solutions. And I trust open source software more. If Apache does become the bottleneck at some point, I'd probably switch to nginx.

well how long has that been there

you're quote seems a bit messed up. someone elses text my name quoting it? whatever though

1M4yNbSCwSMFLF9BaLqzoo2to1WHtZrPke is the forums right? $74,789 USD

don't tell me it's acceptable to just hope that site traffic goes down so it is not slow anymore

what are we going to do when another wordpress type announcement happens? just not have a loading forum any longer?


a second server dedicated to mysql could be had for what? a couple hundred a month tops?


i get trusting open source software more but you can't claim it's not much better then open source in some cases.

It's been that way for a while now. Admin should be able to inform where traffic is coming from.


Wait for traffic to go down?
Stop being cheap, your new car can wait. The more users that are here, the more the word of bitcoin is spread. The more that sign up, the more the forums make in profit. The forums are making a killing in profit, pay to upgrade them.

Static stuff is loaded instantly, the site is db bound.

Is mysql query cache enabled? AFAIK it is disabled by default and can really improve DB performance.

You can check with a SQL query:
(0=disabled)

I doubt the webserver is the bottleneck. Most like the db backend is.

tbh: the forum is blazingly fast at the moment. I had experienced some minor delays loading threads (maybe 1-2 seconds) earlier today, but nothing serious.

How bad is/was it for others?

Either way, hardware, software, or editing the db code to speed it up, if he cant do it himself, admin should pay someone to fix the issue. Id be pretty upset if I was paying for advertisements, and the site is hardly running. New members that visit sites that are frozen just leave, it gives bad reputation.

Its been very slow for me most of the day, keep getting server not responding message.
 

For some time the forum felt like 1998 with my 56K modem.

Also super fast for me now. How can response times vary so much within the same day?
DDOS?

Some More Statistics:

Somehow the Bitcoin Google Stats for today somehow seem wired:

http://anonymouse.org/cgi-bin/anon-www.cgi/http://img593.imageshack.us/img593/9603/googletrend.png

http://www.google.com/trends/explore#q=bitcoin&date=today%203-m&cmpt=q


Strange day.

it wont show what the traffic actually as for at least 48 hours, but im guessing big.

Thanks, I'll try enabling that!

I agree more funds should be spent upgrading the server the forum runs on.

Keep wishing.... I'm pretty sure they've raised like 35,000 bitcoins ....

I hope they all buy a copy of Bitcoin Step by Step.  :-)

haha, what ARE the funds spent on?

Most of yesterday, and some this morning, I was getting that Google warning page that states the page took too long to load. I was blaming my computer. Now I have to apology to her.

Disable it.

I have no words to describe how dumb its to enable the query cache on such a website. Every time there is an update to a table the whole cache is invalidated (removed from the ram). Every time someone access the website the new cache is created, to be invalidate one second after.

It's counter productive. It only works with websites with no/infrequent upgrades.

If you are tiny on budget (why the hell are you using softlayer/rackspace?) and you want to improve the database performance, convert everything to InnoDB and set innodb_buffer_pool to your free ram. I suppose that the free ram should be bigger than the size of the database. If not, buy a new dedicated. Even an ovh.com/us should do, you can buy 2 monster servers for the price you pay now and setup a failover to withstand "poor" hardware iusses.

I could do that for you, I enjoy doing these things, but I'm sure you wouldn't trust me enought. So please study and fix this.

From dev.mysql.com: "When tables are modified, any relevant entries in the query cache are flushed." Not the whole cache ... but maybe I'm too dumb to understand your point of a complete cache invalidation when someone just accesses the forum.

However, I would suggest to: Enable it; Check and optimize the hit ratio; See how it performs; Disable it if it makes things worse

Cheers!

25% of queries are delivered from the query cache, which seems pretty good.

http://www.mysqlperformanceblog.com/2011/04/10/should-we-give-a-mysqlquery-cache-a-second-chance/

25% of queries delivered from cache doesn't necessarily means better performances.

I don't know about how optimized the indexes are for this decade-old board, but i'm sure that you would see a huge benefit converting to a 2013 board (xenforo is my bet, IPB of VB if you wish) and using InnoDB.

Seriously ?
Why not simply use NGINX with PHP-FPM instead (If not already using it) ?

NGINX is already used by many large sites and is blazing fast - I don't think there will be much debate over it.

PHP-FPM is slightly faster than SPAWN-FCGI or custom script dispatcher, but consumes much less memory (as all php processes share memory).

+1

https://i.imgur.com/k4s1z.png

Also, moving from MySQL to MariaDB could be helpful. "Normal" MySQL is not really developed anymore, as Oracle does not give a duck about it - they only want your money by selling you closed source extensions.

I have already moved my 2 servers with 50+ million views/month to MariaDB and I am serving ~35 million views with single quad-core server (but of course heavy caching is involved).

;)

I use XtraDB instead of the standard InnoDB. I serve 40 req/s med and 100 req/s top on a single server. It heavily depends on the patter of the visits.

So is there any word on where the hits came from?
Can admin let us know where the hits came from?

MariaDB contains XtraDB (it replaces InnoDB "drop in", plus it is binary compatibile, or at least MariaDB devs claim so)

The right approach to handling load spikes on a forum like this would be full-page caching in the web layer, using a caching reverse proxy in front of whatever webserver you use now. I do not know the guest:logged in user ratios here, but you could probably get away with caching only guest contents (probably less headaches that way).

Look at Varnish, which is written specifically for this job, or at Nginx which should also work. Here is an example and some numbers (http://fennb.com/microcaching-speed-your-app-up-250x-with-no-n) for you.

We need more servers!

reddit and Hacker News.

It is a good idea, however it will not be enough.
There should also PHP-side cache based on Memcached to offload MySQL/MariaDB servers.

I don't even use Varnish and still I can serve hundereds of requests per second on single server. You can get performance increase of thousand percent and more. Been there, done that.

Check this out:

This is a single Quad Core server, it contains MariaDB, NGINX, Apache and PHP - all on a single server. It serves total of ~1.400.000 views/day (when you count google/yahoo bots).
And i still think i can squeeze out much more speed using NoSQL.

Also, my MariaDB tables contain ~15GiB of data, with total of 65 million records.

He already stated he would eventually move to nginx if apache became the bottleneck.

Think about what servers you are on now.

https://www.ovh.com/us/dedicated-servers/

OVH allows huge bandwidth, you can get a full 100 Mbps line, intel e3 1245v2 processor with 2x SSD 120GB drives for under $100 a month. That would be a heck of a DB server. Mirror raid that for read performance and data security. Have that sit behind the main server as the DB server, and toss another one out in front with nginx as the main server. for $180 a month... you've got a solid set up. Each server has 32 GB of ram in it.

Their virtual rack price is a little bit expensive compared to most solutions (But their servers are much cheaper) Virtual rack you can communicate between the two servers on a local lan and not use the public bandwidth and shield the DB server a little bit. Local connection is at a gig. It appears even without the virtual rack though, that you can communicate with local IPs to other servers. Have not tested that and not as secure as a virtual rack.

If you're hosting with softlayer, you are probably already paying for a similar processor for about $200 a month, and tapped at 5 TB bandwidth with a fraction of the ram.

Late November 2012, I think? I doubt it's related. I remember the passes being on sale, with Bitcoin, right before the holidays.