Category Archives: Bandwidth

Twitch Stream about ZFS, zpool scrubbing, Hard drives, Data Centers, NVMe, Rack Servers…

Twitch stream on 2022-06-06 10:50 IST

In this very long session we went through actual errors in a ZFS pool, we check the Kernel, we remove and reinsert the drive, conduct zpool scrub… in the meantime I talked about Rack, Rack Servers, PSU, redundant components, ECC RAM…

News from the blog 2020-12-16

  • Happy Christmas!

This is the 3D tree that I bought, which is programmable in Python :)

  • If you’re into ZFS I recommend this video:

https://klarasystems.com/learning/webinars/best-practices-for-optimizing-zfs1/

Is a video from klarasystems about best practices for ZFS.

  • Amazing Apache Kafka resources can be found here:

https://developer.confluent.io/

https://developer.confluent.io/learn-kafka/

  • I decided to lower the price of my book to the minimum in LeanPub $5 USD while covid is going on in order to help people with their lives.
    https://leanpub.com/u/carlesmateo

I read with surprise that Comcast is capping the Internet use to 1.2TB per month, and that they will be charging excess.

So… if I contract a Backup with Carbonite or BackBlaze or DropBox or another company and I backup my 10TB files, Comcast will ruin me charging excesses…
Or if I work from home, or the family watches a lot of Netflix…
I can only thinK on their Cast Strategy of CastNumberOfClientsToBankrupcy.

A joke to indicate that I think they will loss clients.

Imagine yesterday I downloaded two images of Ubuntu, being 5 GB, installed Call of Duty in one computer 180 GB, installed few Xbox games 400 GB, listened to Spotify 10 Gb, watched youtube 3 GB, watched Netflix 4 GB, so 602 GB in one day.

Not counting the bandwidth WFH (Working from Home).

Not counting Windows Updates, TV updates, consoles updates, Android Updates, Ubuntu updates…

And this is done in the middle of the covid-19 pandemic, with so many people lock down at home, playing video games, watching movies, and requiring desperately distractions.

<irony>Well done Comcast!</irony>

The Ethernet standards group announces a new 800 GbE specification

Here is the link to the new: https://www.pcgamer.com/amp/the-ethernet-standards-group-developed-a-new-speed-so-fast-it-had-to-change-its-name/

This is a great new for scaling performance in the Data Centers. For routers, switches…

And this makes me think about all the Architects that are using Memcached and Redis in different Servers, in Networks of 1Gbps and makes me want to share with you what a nonsense, is often, that.

So the idea of having Memcache or Redis is just to cache the queries and unload the Database from those queries.

But 1Gbps is equivalent to 125MB (Megabytes) per second.

Local RAM Memory in Servers can perform at 24GB and more (24,000,000 Megabytes) per second, even more.

A PCIE NVMe drive at 3.5GB per second.

A local SSD drive without RAID 550 MB/s.

A SSD in the Cloud, varies a lot on the provider, number of drives, etc… but I’ve seen between 200 MB/s and 2.5GB/s aggregated in RAID.

In fact I have worked with Servers equipped with several IO Controllers, that were delivering 24GB/s of throughput writing or reading to HDD spinning drives.

If you’re in the Cloud. Instead of having 2 Load Balancers, 100 Front Web servers, with a cluster of 5 Redis with huge amount of RAM, and 1 MySQL Master and 1 Slave, all communicating at 1Gbps, probably you’ll get a better performance having the 2 LBs, and 11 Front Web with some more memory and having the Redis instance in the same machine and saving the money of that many small Front and from the 5 huge dedicated Redis.

The same applies if you’re using Docker or K8s.

Even if you just cache the queries to drive, speed will be better than sending everything through 1 Gbps.

This will matter for you if your site is really under heavy load. Most of the sites just query the MySQL Server using 1 Gbps lines, or 2 Gbps in bonding, and that’s enough.

Scaling phantomjs with PHP

One of my clients had a problem with a Phantomjs Software.

I was asked to help in their project, that was relying on one of its features.

Phantomjs is an interesting project, but unfortunately it has not had enough maintenance and a terrible lack of sufficient documentation. The last contributions to repo are from mid May, with small frequency. (Latest releases are from Feb 2015, see the Phantomjs releases on github)

The Software from my client ran well for certain requests, but not for others and after a random time, seconds, or minutes, it became irresponsible.

My client wanted to fix that or to use nodejs to scale their phantom code or in the worst case to rewrite the code in nodejs. And it was urgent, because they were losing a lot of money because of their programs malfunctioning.

I began to investigate. That’s the history of how I fixed…

Connections being irresponsible

My client was using the Phantomjs webserver.

The problem with Phantom’s webserver is that it has a hard limit of 10 concurrent connections. After that all the next http connections are queried until one becomes free.

So if you do a telnet to that port, the connection is accepted, but nothing happens. Even sending malformed GET requests.

My guess was that something in the process of parsing the requests was wrong, and then some of those 10 connections became frozen. I started to debug.

I implemented a timedout that will quit the worker after some time.

mTimerExit = setTimeout(forceExitByTimeout, DEFAULT_TIME_TO_EXIT);

Before exiting is important to clear the timers

clearTimeout(mTimerExit);

I also implemented a debug mode to see what was going on with a method consoleDebug that basically did console.log according to if a parameter debug was set to true.

My quickwin system was working, but many urls still were not being parsed by the phantomjs Engine.

Connecting with nodejs

My client had the bad experience of previous versions of Phantomjs crashing a lot.

So it has the idea of running nodejs as the main webserver, for scaling, and invoking Phantomjs from it.

I did several work in this line.

I tried to link with nodejs with products like:

1) https://github.com/alexscheelmeyer/node-phantom

Unfortunately those packets are no longer maintained, having seen the last update from 2013.

It doesn’t work. I found no documentation, and no traces on errors.

I also got errors like:

XMLHttpRequest cannot load http://localhost:8888/start Origin file:// is not allowed by Access-Control-Allow-Origin

And had to figure out what parameters to tune. I did by starting phantomjs with the param:

 --web-security=false 

In the js scene products and packages are changing very fast and sadly often breaking retrocompatibility.

So you better have a very well defined package.json that installs exactly the software version that you need, or soon, when you deploy to another server it will be a disaster.

2) https://www.npmjs.com/package/ghost-town

Ghost Town is a product that allows to run phantomjs from inside nodejs.

It is a company maintained product, by a contributor, Teddy.

He was very nice replying my questions, but it didn’t help.

The process was failing with no debug, no info.

The package really lacks documentation, and has only the same sample across all the web.

I provide this ghost-town code sample, in case it is useful for people looking for more:

var phantomClusterOptions = require("./phantomClusterOptions");
var town = require("ghost-town")(phantomClusterOptions);
var alerts = require("./qualitynodephantom"); // Do not ad .js
var PORT = 8080;

if (town.isMaster) {
    var express = require('express');
    var app = express();
    app.get('/', function(req, res) {
        // Every request comes here
        var data = {url:req.query.url,device:req.query.quality};

        town.queue(data, function(err,result) {
            res.set('Content-Type', 'text/plain');
            if (!err) {
                res.send(result);
            } else {
                res.send(err);
            }
        }, phantomClusterOptions.pageTries);

    });
    app.listen(PORT);
    console.log('App running');
} else {
    town.on("queue", function(page, data, callback) {
        town.phantom.set('onError', function(msg,trace){});
        // quality is the exported method, you pass the useful page object as parameter
        quality(page, data, function(str){
            callback(null, str);
        });
    });
    town.on("error", function(err) {console.log("error");});
}

And the file phantomClusterOptions has:

//Options here https://github.com/buzzvil/ghost-town
phantomClusterOptions = {
  //phantomBinary:'./phantomjs', //if you want to use a different phantomjs version
  //phantomBinary:'/usr/bin/phantomjs', 
  workerDeath: 3, //number of times that instance of phantom will be reused
  pageTries:5, //tries to the page before rejecting
  pageCount: 1, //number of pages analysed concurrently by the same phantom instance (1 is recommended)
  // This is for versions 1.9 and older of ghost-town
  //phantomFlags:['--load-images=no', '--local-to-remote-url-access=yes', '--ignore-ssl-errors=true', '--web-security=false', '--debug=true'] //flags (http://phantomjs.org/api/command-line.html)
// For v.2 and newer versions
  phantomFlags: {"load-images" : false, "local-to-remote-url-access" : true, "ignore-ssl-errors" : true, "web-security" : false, "debug" : true}
}
module.exports = phantomClusterOptions;

3) Other products

https://www.npmjs.com/package/node-phantom-simple

https://github.com/sgentle/phantomjs-node

I tried to debug with node debugger from command-line:

 

node debug myapp.js

blog-carlesmateo-com-node-debug-command-line

And with node-debug (very nice integration with Chrome):

node-debug myapp.js

blog-carlesmateo-com-debug-node-ghost-town

But I was unable to see what was failing. The nodejs App was up, and the ghost-town queue was increased, but apparently the worker processing the queue was not working or unable to execute phantomjs. But I saw no errors. When I switched the params for ghost-town to v.2, I got some exception, and it really looks like is unable to execute Phantom, or perhaps phantomjs could not exec the .js due to some dependencies problem.

(throw err and error spawn EACCES)

blog-carlesmateo-com-ghost-town-throw-err-error-spawn-eaccessAlso:

Error: /mypath/node_modules/ghost-town/node_modules/phantom/node_modules/dnode/node_modules/weak/build/Release/weakref.node: undefined symbol: node_module_register
    at Module.load (module.js:356:32)
    at Function.Module._load (module.js:312:12)
    at Module.require (module.js:364:17)
    at require (module.js:380:17)
    at bindings (/mypath/node_modules/ghost-town/node_modules/phantom/node_modules/dnode/node_modules/weak/node_modules/bindings/bindings.js:76:44)
    at Object.<anonymous> (/mypath/node_modules/ghost-town/node_modules/phantom/node_modules/dnode/node_modules/weak/lib/weak.js:7:35)
    at Module._compile (module.js:456:26)
    at Object.Module._extensions..js (module.js:474:10)
    at Module.load (module.js:356:32)
    at Function.Module._load (module.js:312:12)

/mypath/node_modules/ghost-town/node_modules/phantom/node_modules/dnode/node_modules/weak/node_modules/bindings/bindings.js:83
        throw e

But I was unable to find more info on the net, I tried to install additional modules and I even straced the processes but I didn’t find the origin of the problem.

I was using:

npm install browserify express ghost-town phantom socket.io URIjs
async dnode forever node-phantom request underscore.string waitfor

About CentOs and Ubuntu

Some SysAdmins love CentOs. I’m in love with Ubuntu.

Basically, is per the packages system. They are really well maintained.

Ubuntu has LTS Long Time Support versions, that last for 5 years.

And in the other hand, they release a new version every 6 months, and if you install a modern server, you have the latest stable packages of Software.

Working with Open Source, this is a really important point. As I have access to modern versions of PHP, Apache, Tomcat, etc…

To use phantomjs with CentOS you have to download the sources and compile it, it took like an hour in a Cloud commodity Virtual Server, and there were problems of dependencies. Also using a phantomjs compiled with a CentOS system didn’t worked with a Server with a different CentOS version. So it was a bit painful to distribute across heterogeneous machines.

With an Ubuntu 14.04 LTS, just:

sudo apt-get install phantomjs

did the trick installing phantomjs (1.9.0-1)

Scaling with PHP

So we had the decision to make between:

  • rewriting completely the application to nodejs, that certainly would take time
  • to invest more time trying to determine why workers freeze under phantomjs

Phantomjs is a headless WebKit scriptable so it was very convenient.

Nodejs is built on Chrome’s Javascript runtime, so it would do what we want to.

As we had a time-constraint and for my client was very important to have the system working asap.

So I decided to debug a bit more.

I found that url’s were being stop loading at the event page.onNavigationRequested

So I could keep all the url and after a timedout could force a page.open(url) inside the event if it stopped (timedout)

mPage.onNavigationRequested = function(url, type, willNavigate, main) {

That was working, finally, but was not my favourite solution. I wanted to understand why it was failing initially.

The lack of documentation was frustrating, but debugging the problematic urls, I found that they were doing several redirections, and after some I was getting SSL certificate error on one of the destination urls.

The thing had to be with chain certificates bad configured.

As nowadays there many cheap SSL certificates providers, based on chain certificates, and many sites are configuring them wrong, phantomjs was sensible to that and stopping following urls.

I already had the param:

--ignore-ssl-errors=true

But investigating I found a very interesting contribution on stackoverflow from user Micah:

http://stackoverflow.com/questions/12021578/phantomjs-failing-to-open-https-site

Note that as of 2014-10-16, PhantomJS defaults to using SSLv3 to open HTTPS connections. With the POODLE vulnerability recently announced, many servers are disabling SSLv3 support.

To get around that, you should be able to run PhantomJS with:

phantomjs --ssl-protocol=tlsv1

Hopefully, PhantomJS will be updated soon to make TLSv1 the default instead of SSLv3.

I decided to give a try to forcing the version of SSL to TLSV1:

--ssl-protocol=tlsv1

And it worked. It did the trick. All the urls were now being parsed right and following the redirects to the end (or to my timedout).

The problem and the solution has been there since 2015 October, and the default use of tlsv1 has not been implemented as default in Phantomjs. That lack of maintenance I found disappointing.

That is why, when recently a multinational interviewed me, and asked me about technologies like nodejs I told them that I’m conservative until it is clear that the version has been proved as stable. And I told that, in any case, a member of the company should me a core member of the contributors to the technology. They were surprised but they shouldn’t! they should have known what I told!. I explained them that if you use a new technology in production, at least you should have a member of your staff in the core of that product. So you pay a guy to build an Open Source technology, basically. This warranties you that if a heavy bug or security flaw appears, you’ll not be screwed until the release. You guy can fix it immediately and share the solution with the community.

Companies like google, Facebook or Amazon do that.

That conservativeness is what I drawn in an interview with Facebook Operations, where I was asked about an scenario where I would be requested by some Developers and DevOps to upgrade the Load Balancers Software. They were more for the action, and I told that LB are critical and I was replied that everything in FB was critical. I argued that if a chat component fails, only the chat fails, but if the Load Balancers fail, everything will fail as they are the entrance point. I had the confirmation that I was right when some months ago they had an outage for hours.

Sometimes you have to keep strong, defend your point, because you know you’re right. Even if you are in front of a person that doesn’t see the things like you and will take a decision that will let you out. Being honest is priceless.

Scaling Phantomjs with PHP

So cool, the system was working fine.

But there was something that could be improved.

As Phantomjs had the limit of 10 connections in their webserver, that was the maximum concurrent connections that it can serve at the same time, and so it was a bottleneck.

// Sample code to create a webserver from PhantomJS
mWebserver = require('webserver');
mServer = mWebserver.create();
console.log("Server created");
//consoleDebug('Debug enabled');
mService = mServer.listen(8080,{'keepAlive': true}, function(request, response) {
    //consoleDebug('URL:' + request.url);
    s_params = request.url;
    doRender(s_params, function(res) {
        //consoleDebug('Response from URL:' + request.url + ' (processed)');
        writeStringResponse(response,res);
    });
    //consoleDebug('URL:' + request.url + ' ready for processing');
});

I decided to do propose to the company to use one of my tricks.

To launch phantomjs from PHP.

This is doing a wrapper to launch Phantomjs from commandline, and getting the response. I did the same in my CQLSÍ Cassandra wrapper around cqlsh before Cassandra drivers for PHP were available. I did also this to connect the payment gateway of a bank, written in C, with the Java libraries from Ticketing Solutions in 1999.

That way the server would be able to process as many concurrent Phantomjs instances as we want, as each one would be running in its own process.

I modified the js code to remove the webserver functionality and to get parameters from command line.

var system = require('system');
var args = system.args;
var b_debug_write = false;

if (args.length < 2) {
    console.log("Minim 2 parameters");
    console.log("call with: phantomjs program.js http://myurl.com quality");
    console.log("Parameter debug is optional");
    args.forEach(function(arg, i) {
            console.log(i + ': ' + arg);
        });
    // Exit with error level 1
    phantom.exit(1);
}

var s_url = args[1];
var s_quality = args[2];

if (args.length > 3) {
    // Enable debug
    b_debug_write = true;
}

consoleDebug("Starting with url:" + s_url + " and quality:" + s_quality);

Then the PHP code:

<?php
/**
 * Creator: Carles Mateo
 * Date: 2015-05-11 11:56
 */

// Report all PHP errors
error_reporting(E_ALL);

$b_debug = false;

if (!isset($_GET['url']) || !isset($_GET['quality'])) {
    echo 'Invalid parameters';
    exit();
}

if (isset($_GET['debug'])) {
    $b_debug = true;
}

$s_url = $_GET['url'];
$s_quality = $_GET['quality'];

// Just in case is not decoded by the PHP installed
$s_url = urldecode($s_url);
// reencode url
$s_url = urlencode($s_url);

$s_script = '/mypath/myapp_commandline.sh';

$s_script_with_params = $s_script.' '.$s_url.' '.$s_quality;

if ($b_debug == true) {
    $s_script_with_params .= ' debug';
    echo 'Executing '.$s_script_with_params."<br />\n";
}

//$message=shell_exec("/var/www/scripts/testscript 2>&1");
$s_message = shell_exec($s_script_with_params);

header("Content-Type: text/plain");
echo $s_message;

And finally the bash script myapp_comandline.sh:

#!/bin/bash

PATH_QUALITY=/mypath/
#tlsv1 is recommended to avoid problems with certificates
PARAMETERS="--local-to-remote-url-access=yes --ignore-ssl-errors=true --web-security=false --ssl-protocol=tlsv1"
cd $PATH_QUALITY

#echo "Debug param1=$1 param2=$2 param3=$3"

if [ -z "$3" ]
  then
    phantomjs $PARAMETERS quality.js $1 $2
  else
    echo "Launching phantomjs with debug. url=$1 quality=$2"
    phantomjs $PARAMETERS quality.js $1 $2 $3
fi

If you don’t need to load the images you can speed up the thing with parameter:

--load-images=false

So finally we were able to use only 285 MB of RAM to handle more than 20 concurrent phantomjs processes.

blog-carlesmateo-com-scaling-phantomjs-with-php-htop

Stopping definitively the massive Distributed DoS attack

I explain some final information and tricks and definitive guidance so you can totally stop the DDoS attacks affecting so many sites.

After effectively mitigating the previous DDoS Torrent attack to a point that was no longer harmful, even with thousands of requests, I had another surprise.

It was the morning when I had a brutal increase of the attacks and a savage spike of traffic and requests of different nature, so in addition to the BitTorent attack that never stopped but was harmful. I was on route to the work when the alarms raised into my smartphone. I stopped in a gas station, and started to fix it from the car, in the parking area.

Service was irresponsible, so first thing I did was to close the Firewall (from the Cloud provider panel) to HTTP and HTTPS to the Front Web Servers.

Immediately after that I was able to log in via SSH to the Servers. I tried to open HTTPS for our customers and I was surprised that most of the traffic was going through HTTPS. So I closed both, I called the CEO of the company to give status, and added a rule to the Firewall so the staff on the company would be able to work against the Backoffice Servers and browse the web while I was fixing all the mess. I also instructed my crew and gave instructions to support team to help the business users and to deal with customers issues while I was stopping the attack.

I saw a lot of new attacks in the access logs.

For example, requests like coming from Android SDK. Weird.

I blocked those by modifying the index.php (code in the previous article)

// Patch urgency Carles to stop an attack based on Torrent and exploiting sdk's
// http://blog.carlesmateo.com
if (isset($_GET['info_hash']) || (isset($_GET['format']) && isset($_GET['sdk'])) || (isset($_GET['format']) && isset($_GET['id']))) {

(Note: If your setup allows it and the HOST requested is another (attacks to ip), you can just check requestes HOST and block what’s different, or if not also implement an Apache rule to divert all the traffic to that route (like /announce for Bittorrent attack) to another file.)

Then, with the cron blocking the ip addresses all improved.

Problem was that I was receiving thousands of requests per second, and so, adding those thousands of Ip’s to IPTABLES was not efficient, in fact was so slow, that the process was taking more than an hour. The server continued overwhelmed with thousands of new ip’s per second while it was able to manage to block few per second (and with a performance adding ip’s degraded since the ip number 5,000). It was not the right strategy to fight back this attack.

Some examples of weird traffic over the http logs:

119.63.196.31 - - [01/Feb/2015:06:54:12 +0100] "GET /video/iiiOqybRvsM/images/av
atar/0097.jpg HTTP/1.1" 404 499 "-" "Baiduspider-image+(+http://www.baidu.com/se
arch/spider.htm)\\nReferer: http://image.baidu.com/i?ct=503316480&z=0&tn=baiduim
agedetail"
119.63.196.126 - - [01/Feb/2015:07:15:48 +0100] "GET /video/Ig3ebdqswQI/images/avatar/0114.jpg HTTP/1.1" 404 499 "-" "Baiduspider-image+(+http://www.baidu.com/search/spider.htm)\\nReferer: http://image.baidu.com/i?ct=503316480&z=0&tn=baiduimagedetail"
180.153.206.20 - - [30/Jan/2015:06:34:34 +0100] "GET /contents/all/tokusyu/skin_detail.html HTTP/1.1" 404 492 "-" "Mozilla/4.0"
180.153.206.20 - - [30/Jan/2015:06:34:35 +0100] "GET /contents/gift/ban_gift/skin_gift.js HTTP/1.1" 404 490 "http://top.dhc.co.jp/contents/gift/ban_gift/skin_gift.js" "Mozilla/4.0"
180.153.206.20 - - [30/Jan/2015:06:34:36 +0100] "GET /contents/all/content/skin.js HTTP/1.1" 404 483 "http://top.dhc.co.jp/contents/all/content/skin.js" "Mozilla/4.0
220.181.108.105 - - [30/Jan/2015:06:35:22 +0100] "GET /image/DbLiteGraphic/201405/thumb_14773360.jpg?1414567871 HTTP/1.1" 404 505 "http://image.baidu.com/i?ct=503316480&z=0&tn=baiduimagedetail" "Baiduspider-image+(+http://www.baidu.com/search/spider.htm)"
118.249.207.5 - - [30/Jan/2015:06:39:26 +0100] "GET /widgets.js HTTP/1.1" 404 0 "http://www.javlibrary.com/cn/vl_star.php?&mode=&s=azgcg&page=2" "Mozilla/5.0 (MSIE 9.0; qdesk 2.4.1266.203; Windows NT 6.1; WOW64; Trident/7.0; rv:11.0; QQBrowser/8.0.3197.400) like Gecko"
120.197.109.101 - - [30/Jan/2015:06:57:20 +0100] "GET /ads/2015/20minutes/publishing/stval/3001.html?pos=6 HTTP/1.1" 404 542 "-" "20minv3/6 CFNetwork/609.1.4 Darwin/13.0.0"
101.226.33.221 - - [30/Jan/2015:07:00:04 +0100] "GET /?LR_PUBLISHER_ID=29877&LR_SCHEMA=vast2-vpaid&LR_PARTNERS=758858&LR_CONTENT=1&LR_AUTOPLAY=1&LR_URL=http://play.retroogames.com/penguin-hockey-lite/?pc1oibq9f5&utm_source=4207933&entId=7a6c6f60cd27700031f2802842eb2457b46e15c0 HTTP/1.1" 200 11783 "http://play.retroogames.com/penguin-hockey-lite/?pc1oibq9f5&utm_source=4207933&entId=7a6c6f60cd27700031f2802842eb2457b46e15c0" "Mozilla/4.0"
180.153.206.20 - - [30/Jan/2015:07:03:10 +0100] "GET /?metric=csync&p=3030&s=6109924323866574861 HTTP/1.1" 200 11783 "http://v.youku.com/v_show/id_XODYzNzA0NTk2.html" "Mozilla/4.0"
180.153.114.199 - - [30/Jan/2015:07:05:18 +0100] "GET /?LR_PUBLISHER_ID=78817&LR_SCHEMA=vast2-vpaid&LR_PARTNERS=763110&LR_CONTENT=1&LR_AUTOPLAY=1&LR_URL=http://www.dnd.com.pk/free-download-malala-malala-yousafzai HTTP/1.1" 200 11783 "http://www.dnd.com.pk/free-download-malala-malala-yousafzai" "Mozilla/4.0"
60.185.249.230 - - [30/Jan/2015:07:08:20 +0100] "GET /scrape.php?info_hash=8%e3Q%9a%9c%85%bc%e3%1d%14%213wNi3%28%e9%f0. HTTP/1.1" 404 459 "-" "Transmission/2.77"
180.153.236.129 - - [30/Jan/2015:07:11:10 +0100] "GET /c/lf-centennial-services-hong-kong-limited/hk041664/ HTTP/1.1" 404 545 "http://hk.kompass.com/c/lf-centennial-services-hong-kong-limited/hk041664/" "Mozilla/5.0 (compatible; MSIE 9.0; Windows NT 6.1; Trident/5.0); 360Spider(compatible; HaosouSpider; http://www.haosou.com/help/help_3_2.html)"
36.106.38.13 - - [30/Jan/2015:07:18:28 +0100] "GET /op/icon?id=com.wsw.en.gm.ArrowDefense HTTP/1.1" 404 0 "-" "Apache-HttpClient/UNAVAILABLE (java 1.4)"
180.153.114.199 - - [30/Jan/2015:07:19:12 +0100] "GET /?metric=csync&p=3030&s=6109985239380459533 HTTP/1.1" 200 11783 "http://vox-static.liverail.com/swf/v4/admanager.swf" "Mozilla/4.0"
180.153.206.20 - - [30/Jan/2015:07:39:56 +0100] "GET /?metric=csync HTTP/1.1" 20
0 11783 "http://3294027.fls.doubleclick.net/activityi;src=3294027;type=krde;cat=
krde_060;ord=94170148950.07014?" "Mozilla/4.0"
220.181.108.139 - - [30/Jan/2015:07:47:08 +0100] "GET /asset/assetId/5994450/size/large/ts/1408882797/type/library/client/WD-KJLAK/5994450_large.jpg?token=4e703a62ec08791e2b91ec1731be0d13&category=pres&action=thumb HTTP/1.1" 404 550 "http://www.passion-nyc.com/" "Baiduspider-image+(+http://www.baidu.com/search/spider.htm)"

And over HTTPS:

120.197.26.43 - - [29/Jan/2015:00:04:25 +0100] "GET /c/5356/cc.js?ns=_cc5356 HTTP/1.1" 404 11676 "http://m.accuweather.com/zh/cn/guangzhou/102255/current-weather/102255?p=huawei2" "HUAWEI Y325-T00_TD/V1 Linux/3.4.5 Android/2.3.6 Release/03.26.2013 Browser/AppleWebKit533.1 Mobile Safari/533.1;"
120.197.26.43 - - [29/Jan/2015:00:05:32 +0100] "GET /c/5356/cc.js?ns=_cc5356 HTTP/1.1" 404 11674 "http://m.accuweather.com/zh/cn/guangzhou/102255/weather-forecast/102255" "HUAWEI Y325-T00_TD/V1 Linux/3.4.5 Android/2.3.6 Release/03.26.2013 Browser/AppleWebKit533.1 Mobile Safari/533.1;"
117.136.1.106 - - [29/Jan/2015:06:35:05 +0100] "GET /v2.2/237613769760602?format=json&sdk=android&fields=supports_attribution%2Csupports_implicit_sdk_logging%2Cgdpv4_nux_content%2Cgdpv4_nux_enabled%2Candroid_dialog_configs HTTP/1.1" 404 6304 "-" "FBAndroidSDK.3.20.0"
117.136.1.106 - - [29/Jan/2015:06:35:11 +0100] "GET /v2.2/237613769760602?format=json&sdk=android&fields=supports_attribution%2Csupports_implicit_sdk_logging%2Cgdpv4_nux_content%2Cgdpv4_nux_enabled%2Candroid_dialog_configs HTTP/1.1" 404 6308 "-" "FBAndroidSDK.3.20.0"
117.136.1.107 - - [29/Jan/2015:06:35:36 +0100] "GET /v2.2/1493038024241481?fields=name,supports_implicit_sdk_logging,gdpv4_nux_enabled,gdpv4_nux_content,ios_dialog_configs,app_events_feature_bitmask&format=json&sdk=ios HTTP/1.1" 404 6731 "-" "FBiOSSDK.3.22.0"
180.175.55.177 - public [29/Jan/2015:07:40:23 +0100] "GET /public/checksum?a=8&t=p&v=2.5.0&c=d62c8d04f74057d51872d5dc5cdab098 HTTP/1.0" 404 39668 "https://rc-regkeytool.appspot.com/public/checksum?a=8&t=p&v=2.5.0&c=d62c8d04f74057d51872d5dc5cdab098" "Mozilla/5.0 (iPhone; U; CPU like Mac OS X; en) AppleWebKit/420+ (KHTML, like Gecko) Version/3.0 Mobile/1C28 Safari/419.3"

So, I was seeing requests that typically do applications like Facebook, smartphones with Android going to google store, and a lot of other applications and sites traffics, that was going diverted to my Ip.

So I saw that this attack was an attack associated to the ip address. I could change the ip in an extreme case and gain some hours to react.

So if those requests were coming from legitimate applications I saw two possibilities:

1. A zombie network controlled by pirates through malware/virus, that changes the /etc/hosts so those devices go to my ip address. Unlikely scenario, as there was a lot of mobile traffic and is not the easiest target of those attacks

2. A Dns Spoofing attack targeting our ip’s

That is an attack that cheats to Dns servers to make them believe that a name resolves to a different ip. (that’s why ssl certificates are so important, and the warnings that the browser raises if the server doesn’t send the right certificate, it allows you to detect that you’re connecting to a fake/evil server and avoid logging etc… if you have been directed a bad ip by a dns poisoned)

I added some traces to get more information, basically I dump the $_SERVER to a file:

$s_server_dump = var_export($_SERVER, true);

file_put_contents($s_debug_log_file, $s_server_dump.”\n”, FILE_APPEND | LOCK_EX);

array (
  'REDIRECT_SCRIPT_URL' => '/announce',
  'REDIRECT_SCRIPT_URI' => 'http://jonnyfavorite3.appspot.com/announce',
  'REDIRECT_STATUS' => '200',
  'SCRIPT_URL' => '/announce',
  'SCRIPT_URI' => 'http://jonnyfavorite3.appspot.com/announce',
  'HTTP_HOST' => 'jonnyfavorite3.appspot.com',
  'HTTP_USER_AGENT' => 'Bittorrent',
  'HTTP_ACCEPT' => '*/*',
  'HTTP_CONNECTION' => 'closed',
  'PATH' => '/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin',
  'SERVER_SIGNATURE' => '<address>Apache/2.4.7 (Ubuntu) Server at jonnyfavorite3.appspot.com Port 80</address>',
  'SERVER_SOFTWARE' => 'Apache/2.4.7 (Ubuntu)',
  'SERVER_NAME' => 'jonnyfavorite3.appspot.com',
  'SERVER_ADDR' => 'deleted',
  'SERVER_PORT' => '80',
  'REMOTE_ADDR' => '60.219.115.77',
  'DOCUMENT_ROOT' => 'deleted',
  'REQUEST_SCHEME' => 'http',
  'CONTEXT_PREFIX' => '',
  'CONTEXT_DOCUMENT_ROOT' => 'deleted',
  'SERVER_ADMIN' => 'deleted',
  'SCRIPT_FILENAME' => 'deleted',
  'REMOTE_PORT' => '27361',
  'REDIRECT_QUERY_STRING' => 'info_hash=%26%27%0F%C2%EB%03J%E2%1F%B0%28%2B%29d%7C%8C%FE%C8l%E9&peer_id=%2DSD0100%2Dj%7E%C3%1C%14%FFsj%DA9%0B%27&ip=10.155.31.47&port=9978&uploaded=2244978664&downloaded=2244978664&left=2570039513&numwant=200&key=511&compact=1',
  'REDIRECT_URL' => '/announce',
  'GATEWAY_INTERFACE' => 'CGI/1.1',
  'SERVER_PROTOCOL' => 'HTTP/1.0',
  'REQUEST_METHOD' => 'GET',
  'QUERY_STRING' => 'info_hash=%26%27%0F%C2%EB%03J%E2%1F%B0%28%2B%29d%7C%8C%FE%C8l%E9&peer_id=%2DSD0100%2Dj%7E%C3%1C%14%FFsj%DA9%0B%27&ip=10.155.31.47&port=9978&uploaded=2244978664&downloaded=2244978664&left=2570039513&numwant=200&key=511&compact=1',
  'REQUEST_URI' => '/announce?info_hash=%26%27%0F%C2%EB%03J%E2%1F%B0%28%2B%29d%7C%8C%FE%C8l%E9&peer_id=%2DSD0100%2Dj%7E%C3%1C%14%FFsj%DA9%0B%27&ip=10.155.31.47&port=9978&uploaded=2244978664&downloaded=2244978664&left=2570039513&numwant=200&key=511&compact=1',
  'SCRIPT_NAME' => '/index.php',
  'PHP_SELF' => '/index.php',
  'REQUEST_TIME_FLOAT' => 1422528394.036,
  'REQUEST_TIME' => 1422528394,
)

Some fields were replaced with ‘deleted’ for security.

Ok, so the client was sending a HOST header

jonnyfavorite3.appspot.com

This address resolves as:

ping jonnyfavorite3.appspot.com
PING appspot.l.google.com (74.125.24.141) 56(84) bytes of data.
64 bytes from de-in-f141.1e100.net (74.125.24.141): icmp_seq=1 ttl=52 time=1.53 ms

So to google appspot.

Next one:

array (
  'REDIRECT_SCRIPT_URL' => '/v2.2/1493038024241481',
  'REDIRECT_SCRIPT_URI' => 'https://graph.facebook.com/v2.2/1493038024241481',
  'REDIRECT_HTTPS' => 'on',
  'REDIRECT_SSL_TLS_SNI' => 'graph.facebook.com',
  'REDIRECT_STATUS' => '200',
  'SCRIPT_URL' => '/v2.2/1493038024241481',
  'SCRIPT_URI' => 'https://graph.facebook.com/v2.2/1493038024241481',
  'HTTPS' => 'on',
  'SSL_TLS_SNI' => 'graph.facebook.com',
  'HTTP_HOST' => 'graph.facebook.com',
  'HTTP_ACCEPT_ENCODING' => 'gzip, deflate',
  'CONTENT_TYPE' => 'multipart/form-data; boundary=3i2ndDfv2rTHiSisAbouNdArYfORhtTPEefj3q2f',
  'HTTP_ACCEPT_LANGUAGE' => 'zh-cn',
  'HTTP_CONNECTION' => 'keep-alive',
  'HTTP_ACCEPT' => '*/*',
  'HTTP_USER_AGENT' => 'FBiOSSDK.3.22.0',
  'PATH' => '/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin',
  'SERVER_SIGNATURE' => '<address>Apache/2.4.7 (Ubuntu) Server at graph.facebook.com Port 443</address>',
  'SERVER_SOFTWARE' => 'Apache/2.4.7 (Ubuntu)',
  'SERVER_NAME' => 'graph.facebook.com',
  'SERVER_ADDR' => 'deleted',
  'SERVER_PORT' => '443',
  'REMOTE_ADDR' => '223.100.159.96',
  'DOCUMENT_ROOT' => 'deletd',
  'REQUEST_SCHEME' => 'https',
  'CONTEXT_PREFIX' => '',
  'CONTEXT_DOCUMENT_ROOT' => 'deleted',
  'SERVER_ADMIN' => 'deleted',
  'SCRIPT_FILENAME' => 'deleted',
  'REMOTE_PORT' => '24797',
  'REDIRECT_QUERY_STRING' => 'fields=name,supports_implicit_sdk_logging,gdpv4_nux_enabled,gdpv4_nux_content,ios_dialog_configs,app_events_feature_bitmask&format=json&sdk=ios',
  'REDIRECT_URL' => '/v2.2/1493038024241481',
  'GATEWAY_INTERFACE' => 'CGI/1.1',
  'SERVER_PROTOCOL' => 'HTTP/1.1',
  'REQUEST_METHOD' => 'GET',
  'QUERY_STRING' => 'fields=name,supports_implicit_sdk_logging,gdpv4_nux_enabled,gdpv4_nux_content,ios_dialog_configs,app_events_feature_bitmask&format=json&sdk=ios',
  'REQUEST_URI' => '/v2.2/1493038024241481?fields=name,supports_implicit_sdk_logging,gdpv4_nux_enabled,gdpv4_nux_content,ios_dialog_configs,app_events_feature_bitmask&format=json&sdk=ios',
  'SCRIPT_NAME' => '/index.php',
  'PHP_SELF' => '/index.php',
  'REQUEST_TIME_FLOAT' => 1422528399.4159999,
  'REQUEST_TIME' => 1422528399,
)

So that was it, connections were arriving to the Server/Load Balancer requesting addresses from Google, Facebook, thepiratebay main tracker, etc… tens of thousands of users simultaneously.

Clearly there was a kind of DNS spoofing attack. And looking at the ACCEPT_LANGUAGE I saw that clients were using Chineses language (zh-cn).

Changing the ip address will take some time, as some dns have 1 hour or several hours of TTL to improve page speed of customers, and both ip’s have to coexist for a while until all the client’s dns have refreshed, browsers have refreshed their cache of dns names, and dns of partners consuming our APIs have refreshed.

You can have a virtual host just blank, but for some projects your virtual hosts needs to listen for all the request, for example, imagine that you serve contents like barcelona.yourdomain.com and for newyork.yourdomain.com and whatever-changing.yourdomain.com like carlesmateo.yourdomaincom, anotheruser.yourdomain.com, anyuserwilhaveownurl.yourdomain.com, that your code process to get the information and render the pages, so it is not always possible to have a default one.

First thing I had to do was, as the attack came from the ip, was to do not allow any host requests that did not belong to the hosts served. That way my framework will not process any request that was not for the exact domains that I expect, and I will save the 404 process time.

Also, if the servers return a white page with few bytes (header), this will harm less the bandwidth if under a DDoS. The bandwidth available is one of the weak points when suffering a DDoS.

To block hosts that are not expected there are several things that can be done, for example:

  • Nginx can easily block those requests not matching the hosts.
  • Modify my script to look at
    HTTP_HOST

    So:
    if (isset($_SERVER['HTTP_HOST']) && ($_SERVER['HTTP_HOST'] != 'www.mydomain.com' && $_SERVER['HTTP_HOST'] != 'mydomain.com') {
    exit();
    }

    That way any request that is not for the exact domain will be halted quickly.
    You can also check instead strpost($_SERVER['HTTP_HOST'], ‘mydomain.com’) for variable HTTP_HOST

  • Create an Apache default virtual host and default ssl. With a white page the load was so small that even tens of thousands ip’s cannot affect enough the server. (Max load was around 2.5%)

Optimization: If you want to save IOPS and network bandwidth to the storage you can create a ramdisk, and have the index.html of the default virtual host in there. This will increase server speed and reduce overhead.

If you’re in a hurry to stop the attack you can also:

  • Change the Ip, get a new ip address, and change the dns to the new ip, and hope this one is not attacked
  • An emergency solution, drastic but effective, would be to block entire China’s ip ranges.

Some of the webs I helped reported having attacks from outside China, but nothing related, the gross attacks, 99%, come from China.

Sadly Amazon’s Cloud Firewall does not allow to deny certain ip addresses, only allow services through a default deny policy.

I reopened the firewall in 80 and 443, so traffic to everyone, checked that the Service was Ok and continued my way to the office.

When later I spoke with some friends, one of them told me about the Chinese government firewall causing this sort of problems worldwide!. Look at this interesting article.

http://blog.sucuri.net/2015/01/ddos-from-china-facebook-wordpress-and-twitter-users-receiving-sucuri-error-pages.html

It is not clear if those DDoS attacks are caused by the Chinese Firewall, by a bug, or by pirates exploiting it to attack sites for money. But is clear that all those attacks and traffic comes from China ip addresses. Obviously this DNS issues are causing the guys browsing on China to not being able of browsing Facebook, WordPres, twitter, etc…

Here you can see live the origin of the world DDoS attacks:

http://www.digitalattackmap.com/#anim=1&color=0&country=ALL&list=0&time=16465&view=map

And China being the source of most of the DDoS attacks.

blog-carlesmateo-com-2015-02-01-15-23-33-china-source-of-attacks

The Cloud is for Scaling

Last Update: 2022-08-03 I added the Enterprises, as more and more services have been provided by Amazon and the other major Cloud Providers and they offer a suitable solution for Enterprises requiring high availability in services managed by Amazon, without the need to configure and maintain their own setups manually.

dell-blades-m4110The Cloud is for Startups, and for Scaling. Nothing more.

In the future will be used by phone operators, to re-dimension their infrastructure and bandwidth in real time according to demand, but nowadays the Cloud is for Startups.

Examine the prices in my post in cmips, take a look, examine the performance also of the different CPU. You see that according to CMIPS v.1.03 a Desktop Processor Intel i7-4770S, worth USD $300, performs better than an Amazon M2 High Memory Quadruple Extra Large and than a Rackspace First gen. 30 GB RAM 8 Cores?.

Today the public cost of an Amazon M2 High Memory Quadruple Extra Large running for a month is USD $1,180.80 so USD $1.64 per hour and the Rackspace First Generation 30 GB RAM 8 Cores 1200 GB of disk costs is USD $1,425.60 so USD $1.98 per hour running.

And that’s the key, the cost per hour.

Because the greatness, the majesty of the Cloud is that you pay per hour, you pay as you need, or as you go. No attaching contracts. All on demand.

I had my company at a time where the hosting companies and the Data Centers were forcing customers to sign yearly contracts. What if a company only needs to host their Servers for three months? What if they have to close?. No options. You take it or you leave it.

Even renting a dedicated hosting was for at least a month or more, and what if the latency was not good? What if the bandwidth of the provider was not enough?.

Amazon irrupted in the market with strength. I really like that company because they grew the best eCommerce company for buying books, they did a system that really worked, and was able to recommend very useful computer books, and the delivery, logistics was so good, also post-sales service. They simply started to rent the same infrastructure they were using to attend their millions of customers and was a total success.

And for a while few people knew about Amazon deep technologies and functionalities, but later became a fashion.

Now people is using Amazon or whatever provider/Service that contains the word “Cloud” because the Cloud is in the mouth of everyone. Magazines and newspapers speak about the Cloud, so many many companies use it simply because everyone is talking about the Cloud. And those ISP that didn’t had a Cloud have invested heavily to create a Cloud, just because they didn’t want to be the ones without a Cloud, since everyone was asking for it and all the ISP companies were offering their “Clouds”.

Every company claims to have “Cloud” where the only many of them have is Vmware servers, Xen servers, Open Stack… running the tenants or instances of the customers always on the same host servers. No real Cloud, professional Cloud, abstract layered in a Professional way like Amazon, only the traditional “shared hosting” with another name, sharing CPU and RAM and Disk storage using virtual machines called instances.

So, Cloud fashion has become a confusing craziness where no one knows why they are in the Cloud but they believe they have to be in.

But do companies need the Cloud?. Cloud instances?

It depends. The best would be to ask that companies Why you choose the Cloud?.

If you compare the cost of having an instance in the Cloud, is much much more expensive than having a dedicated server. And for that high cost you don’t get more performance.

Virtualization is always slower and disk speed is always an issue in Cloud providers, where all the data travels via network from the disk cabins NAS to the Host servers running the guest instances. Data cannot be at local disks, since every time you start an instance, the resources like CPU and RAM are provisioned, and your instance run in totally different hardware. Only your data remain in the NAS (Network Attached Storage).

So unless you run your in-the-Cloud instance in a special provider that offers local disks, like DigitalOcean that offers SSD but monthly paying, (and so you pay the price by losing the hardware abstraction capability because you’re attached to the CPU that has the disk connected, and also you loss the flexibility of paying per hour of use, as you go), then you’ll face a bottleneck that is the hard disk performance (that for real takes all the data from NAS, where is stored, through the local network).

So what are the motivations to use the Cloud?. I try to put some examples, out of these it has no much sense, I think. You can send me your happy-in-Cloud scenarios if you found other good uses.

Example A) Saving initial costs, avoid contract attachment and grow easily own-made

Imagine a Developer that start its own project. May be it works, may be not, but instead of having a monthly contract for a dedicated server, he starts with an Amazon Free Tier (better not, use Small instance at least) and runs a web. If it does not work, simply stop the instance and pay no more. If the project works and has more and more users he can re-dimension the server with a click. Just stop the instance, change the type of instance, start it again with more RAM and more CPU power. Fast.

Hiring a dedicated server implies at least monthly contracts, average of USD $100 per month, and is not easy to move to a bigger server, not fast and is expensive as it requires the ISP tech guys to move the data, to migrate from a Server to another.

Also the available bandwidth is to be taken in consideration. Bandwidth is expensive and Amazon can offer 150 Mbit to smaller machines. Not all the Internet Service Providers can offer that bandwidth even with most advanced packets.

If the project still grows, with a click, in seconds, 20 instances with a lot of bandwidth can be deployed and serving traffic to your customers very quick.

You save the init costs of buying Servers, and the time to deal with hardware, bandwidth limitations and avoid contracts, but you pay an hourly rate a lot more expensive. So in the long run is much much expensive using Amazon and less powerful than having dedicated servers. That happened to Zynga, that was paying $63M annually to Amazon and decided to step back from Amazon to their own Data Centers again. (another fortune tech link)

The limited CPU power was also a deal breaker for many companies that needed really powerful CPU and gigs of RAM for their Database Servers. Now this situation is much better with the introduction of the new Servers.

This developer can benefit from doing bacups with a click, cloning, starting instances from an image, having more static ip’s with a click, deploying built-in (from the Cloud provider) load balancers, using monitoring services like CloudWatch, creating Volumes and attaching to the servers for additional space…

Example B) An Startup with fluctuating number of users and hopes of growing

Imagine an Startup with a wonderful Facebook Application.

During 80% of the day has few visits, may be only need 3 Servers, but during 20% of the hours of the day from 10:00 to 15:00 users connect like hell, so they need 20 servers to attend this traffic and workload, and may be tomorrow needs 30 servers.

With the Cloud they pay for 3 servers 24 hours per day and for the other 17 servers only pay the hours they are on, that’s 5 hours per day. Doing that they save money and they have an unlimited * amount of power. (* There are limits for real, you have to specially request authorisation to run more than default max. servers for the zone, that is normally 20 instances for Amazon. Also it can happen theoretically that when you request new instances the Zone has no instances available).

So well, for an Startup growing, avoiding hiring 20 dedicated servers and instead running into the Cloud as many as they need, for just the time they need, Auto-Scaling up and down, and can use the servers NOW and pay the next month with Visa card, all of that can make a difference for a growing Startup.

If the servers chosen are not powerful enough that is solved with a click, changing instance type. So fast. A minute.

It’s only a matter of money.

Example C) e-Learning companies and online universities

e-Learning platforms also get benefits from the Auto-Scaling for the full occupation hours.

The built-in functionalities of the Cloud to clone instances is very useful to deploy new web servers, or new environments for students doing practices, in the case of teaching Information Technology subjects, where the users need to practice against a real server (Linux or windows).

Those servers can be created and destroyed, cloned from the main -ready to go- template. And also servers can be scheduled to stop at a certain hour and to start also, so saving the money from the hours not needed.

Example D) Digital agencies, sports and other events

When there is an Special event, like motorcycle running, when a Football Team scores, when there is an spot in tv announcing a product…

At those moments the traffic to the site can multiply, so more servers and more bandwidth have to be deployed instantly. That cannot be done with physical servers, hardware, but is very easy to provision instances from the Cloud.

Mass mailing email campaigns can also benefit from creating new Servers when needed.

Example E) Proximity and SEO

Cloud providers have Data Centres everywhere. If you want to have servers in Asia, or static content to be deployed faster, or in South-America, or in Europe… the Cloud providers have plenty of Data Centers all over the world.

Example F) Game aficionado and friends sharing contents

People that loves cooperative games can find the needed hungry bandwidth and at a moderate price. If they run their private server few hours, at night, from 22:00 to 01:00 as example, they will benefit from a great bandwidth from the big Cloud provider and pay only 3 hours per day (the exceed of traffic uses to be paid in most providers, but price of additional GB uses to be really really competitive).

Friends sharing contents in an Ftp also, can benefit from this Cloud servers, but probably they will find more easy to use services like Dropbox.

Example G) Startup serving contents

An Startup serving videos, images, or books, can benefit not only from the great bandwidth of big Cloud providers (this has been covered before), but for a very cheap price for exceeding Gigabyte transferred.

Local ISP can’t offer 150 Mbit for an instance of USD $20 and USD $0.12 per additional GB transferred.

Many Cloud providers also allow unlimited incoming traffic from the Internet, and from Server to Server through private ip’s.

Example H) Enterprises that want to outsource the solutions and having better availability

Since I wrote this article in 2013, many CSP have been incorporation more and more powerful services to their catalalog.

For instance Amazon has service like DynamoDB, RDS, Load Balancers able to balance servers in different availability zones, with Auto-Scaling, SQS for Message Queues, ElastiCache, Lambda, S3…

Many Enterprise find convenient to have all their services offloaded to Amazon, Google Cloud, Microsoft Azure… so they require less Operations and System Administration Engineers.

Other cases

For other cases Dedicated Servers are much more Powerful, faster and cheaper, at the price of being “static” in the sense of attached, not layer abstracted, but all the aspects of your Project have to be taken in count before deciding stepping into or out of the Cloud.

In general terms I would say that the Cloud is for Scaling, but in 2022 it’s also to outsource Operation Services, to fully managed services by the CSPs.

NAS and Gigabit

Note 2019-05-28: This article was written in 2013. It is still valid, but since then 10Gbps have dropped in price a lot. In my DRAID Solution I’ve qualified 10GbE based in copper, RJ45, and for fiber: 10, 25, 40 and 100 Gbps. Mellanox switches and NICs are a reference. 10GbE based in copper are cheap and easy to deploy, as you can reuse existing infrastructure and grow your segments. https://en.wikipedia.org/wiki/10_Gigabit_Ethernet

I’ve found this problem in several companies, and I’ve had to show their error and convince experienced SysAdmins, CTOs and CEO about the erroneous approach. Many of them made heavy investments in NAS, that they are really wasting, and offering very poor performance.

Normally the rack servers have their local disks, but for professional solutions, like virtual machines, blade servers, and hundreds of servers the local disk are not used.

NAS – Network Attached Storage- Servers are used instead.

This NAS Servers, when are powerful (and expensive) offer very interesting features like hot backups, hot backups that do not slow the system (the most advanced), hot disk replacement, hot increase of total available space, the Enterprise solutions can replicate and copy data from different NAS in different countries, etc…

Smaller NAS are also used in configurations like Webservers’ Webfarms, were all the nodes has to have the same information replicated, and when a used uploads a new profile image, has to be available to all the webservers for example.

In this configurations servers save and retrieve the needed data from the NAS Servers, through LAN (Local Area Network).

The main error I have seen is that no one ever considers the pipe where all the data is travelling, so most configurations are simply Gigabit, and so are bottleneck.

dell-blade-servers-enclosureImagine a Dell blade server, like this in the image on the left.

This enclosure hosts 16 servers, hot plugable, with up to two CPU’s each blade, we also call those blade servers “pizza” (like we call before to rack servers).

A common use is to use those servers to have Vmware, OpenStack, Xen or other virtualization software, so the servers run instances of customers. In this scenario the virtual disks (the hard disk of the virtual machines) are stored in the NAS Server.

So if a customer shutdown his virtual server, and start it later, the physical server where its virtual machine is running will be another, but the data (the disk of the virtual server) is stored in the NAS and all the data is saved and retrieved from the NAS.

The enclosure is connected to the NAS through a Gigabit connection, as 10 Gigabit connections are still too expensive and not yet supported in many servers.

Once we have explained that, imagine, those 16 servers, each with 4 or 5 virtual machines, accessing to their disks through a Gigabit connection.

If only one of these 80 virtual machines is accessing to disk, the will be no problem, but if more than one is accessing the Gigabit connection, that’s a maximum of 125 MB (Megabytes) per second, will be shared among all the virtual machines.

So imagine, 70 virtual machines are accessing NAS to serve web pages, with not much traffic, OK, but the other 10 virtual machines are doing heavy data transmission: for example one is serving data through FTP server, the other is broadcasting video, the other is copying heavy log files, and so… Imagine that scenario.

The 125 MB per second is divided between the 80 servers, so those 10 servers using extensively the disk will monopolize the bandwidth, but even those 10 servers will have around 12,5 MB each, that is 100 Mbit each and is very slow.

Imagine one of the virtual machines broadcast video. To broadcast video, first it has to get it from the NAS (the chunks of data), so this node serving video will be able to serve different videos to few customers, as the network will not provide more than 12,5 MB under the circumstances provided.

This is a simplified scenario, as many other things has to be taken in count, like the SATA, SCSI and SAS disks do not provide sustained speeds, speed depends on locating the info, fragmentation, etc… also has to be considered that NAS use protocol iSCSI, a sort of SCSI commands sent through the Ethernet. And Tcp/Ip uses verifications in their protocol, and protocol headers. That is also an overhead. I’ve considered only traffic in one direction, so the servers downloading from the NAS, as assuming Gigabit full duplex, so Gigabit for sending and Gigabit for receiving.

So instead of 125 MB per second we have available around 100 MB per second with a Gigabit or even less.

Also the virtualization servers try to handle a bit better the disk access, by keeping a cache in memory, and not writing immediately to disk.

So you can’t do dd tests in virtual machines like you would do in any Linux with local disks, and if you do go for big files, like 10 GB with random data (not just 0, they have optimizations for that).

Let’s recalculate it now:

70 virtual machines using as low as 0.10 MB/second each, that’s 7 MB/second. That’s really optimist as most webservers running PHP read many big files for attending a simple request and webservers server a lot of big images.

10 virtual machines using extensively the NAS, so sharing 100 MB – 7 MB = 93 MB. That is 9.3 MB each.

So under these circumstances for a virtual machine trying to read from disk a file of 1 GB (1000 MB), this operation will take 107 seconds, so 1:47 minutes.

So with this considerations in mind, you can imagine that the performance of the virtual machines under those configurations are leaved to the luck. The luck that nobody else of the other guests in the servers are abusing the disk I/O.

I’ve explained you in a theoretical plan. Sadly reality is worst. A lot worst. Those 70 web virtual machines with webservers will be so slow that they will leave your company very disappointed, and the other 10 will not even be happier.

One of the principal problems of Amazon EC2 has been always disk performance. Few months ago they released IOPS, high performance disks, that are more expensive, but faster.

It has to be recognized that in Amazon they are always improving.

They have also connection between your servers at 10 Gbit/second.

Returning to the Blades and NAS, an easy improvement is to aggregate two Gigabits, so creating a connection of 2 Gbit. This helps a bit. Is not the solution, but helps.

Probably different physical servers with few virtual machines and a dedicated 1 Gbit connection (or 2 Gbit by 1+1 aggregated if possible) to the NAS, and using local disks as much as possible would be much better (harder to maintain at big scale, but much much better performance).

But if you provide infrastructure as a Service (IaS) go with 2 x 10 Gbit Fibre aggregated, so 20 Gigabit, or better aggregate 2 x 20 Gbit Fibre. It’s expensive, but crucial.

Now compare the 9.3 MB per second, or even the 125 MB theoretical of Gigabit of the average real sequential read of 50 MB/second that a SATA disk can offer when connected on local, or nearly the double for modern SAS 15.000 rpm disks… (writing is always slower)

… and the 550 MB/s for reading and 550 MB/s for writing that some SSD disks offer when connected locally. (I own two OSZ SSD disks that performs 550 MB).

I’ve seen also better configuration for local disks, like a good disk controller with Raid 5 and disks SSD. With my dd tests I got more than 900 MB per second for writing!.

So if you are going to spend 30.000 € in your NAS with SATA disks (really bad solution as SATA is domestic technology not aimed to work 24×7 and not even fast) or SAS disks, and 30.000 € more in your blade servers, think very well what you need and what configuration you will use. Contact experts, but real experts, not supposedly real experts.

Otherwise you’ll waste your money and your customers will have very very poor performance on these times where applications on the Internet demand more and more performance.

Why Cloud is not optional

data-center-companyIf you’re a Developer or an Entrepreneur to avoid Cloud is not an option.

It is a must to use the Cloud.

Why?.

Because if your project is a success, you’ll need to scale very fast from a single server to many, just to attend the increasing number of users.

And if you have a lot of users, you’ll need a lot of bandwidth.

Even if you have a single server, but want to serve video, will need a lot of bandwidth to serve data fast.

Here is where Cloud is not an option. It’s a must.

If you use a big server with a lot of RAM or CPUs, or several servers, Amazon EC2 is very expensive.

But to start with the needed power, and to be able to grow really fast, and to pay as you go there is no other option.

The smaller instance from Amazon, is able to serve 150 Mbit per second.

If you need to serve video, where would you be able to deliver it at 150 Mbit/sec rates at 17 € / 14.67 £ / US $23 per month?.

Nowhere else.

It’s not a matter of something fashion, there’s not alternative.

With the privative price of dedicated bandwidth in the data centres, no one else can offer something similar even for then times this price.

So if you need to be able to pass from a server to 20 or 100 with a click and within a minute, and you need to deliver contents very fast there is no other option.

With Amazon Cloud since you pay per hour, you can create 20 instances to face a rise of visitors, due to a campaign or because you have top visitors window of time, you can create instances as you need and destroy them when you don’t need them and pay only for the hours used.

So when your Start up is growing and low on money/resources, you save the costs of buying several physical server ($2,000 each), the time of installing, of replacing if a motherboard or disk crashes, and simply creating new instances in the Amazon Cloud as you need, and paying only for the time you use them.

So you can save the costs and grow as you need.

There are other benefits like you can use Amazon data centres in all-over the world, where the infrastructure is closer to your customers (reducing the latency and increasing the speed of servicing pages), the CDN service, load balancers…

The cost of the transferred Gigabyte is another reason.

One month I transferred 287 GB and paid only $50.

An small ISP can’t beat this nor even compete with this price and speed.