<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>So The Site With the LAMP had three outages in three days. Did I say three in three? But my <a href="site-offline-twice-in-two-days.html">previous post</a> mentioned only two? Sorry about that. There was another one&nbsp;immediately&nbsp;after that was publish. &nbsp;In all three&nbsp;occasions&nbsp;downtime was caused by a surge in demand for Amazon EC2 spot instances.</p>
            <p>&nbsp;</p>
            <p></p>
            <figure style="text-align:center"><a href="http://images.raditha.com/blog/images/spot-pricing.png"><img alt="Amazon EC2 spot pricing" height="183" src="http://images.raditha.com/blog/images/spot-pricing-t.png" title="Amazon EC2 spot pricing" width="400"></a>
                <figcaption>EC2 Spot pricing, with several large spikes</figcaption>
            </figure>
            <p></p>
            <p>&nbsp;</p>
            <p>The chart shows several surges since the 27th of September. On the 29th, the price of an M1.large instance (the kind of virtual machine used to power this site) shot upto USD 15.00 per hour. Usually these are priced at USD 0.12 per hour, so it represents a 125 fold increase!</p>
            <p>&nbsp;</p>
            <p>I had set the upper limit for my instance at USD 2.00 per our thinking that a safe limit. And it really ought to be safe because the 'on demand' pricing is only 0.34 per hour. How in his right mind would pay slightly more for the some service when the same service is available by the same company for a much lower price? I am pretty sure a few sysadmins and network managers would have lost their jobs. Some people would have burnt up a months hosting budget in just in just five hours!</p>
            <p>&nbsp;</p>
            <p>While you might think that it's small businesses, free lancers or bloggers who rely on amazon spot instance, the truth is very different, even foursquare and many other large websites and web apps are running on Amazon EC2 spot instances. It looks like there might have been a DDoS attack on such a company. Whoever the real target was, I am pretty sure that they managed to keep their service online at great cost (or we might have heard about it). But looks like the attackers left behind a hell of a lot of collateral damage.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Who is the victim?

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>After running smoothly for more than an year, the Site with the LAMP was offline for around twelve hours yesterday, it was caused the by EC2 instance on which the site is hosted going offline &nbsp;for some unknown reason. It had restarted soon after but all the data in the instance storage had gone missing. &nbsp;Of course Amazon strongly advices against relying on instance storage, you are supposed to use EBS volumes but lazybones (or cheapskates) like me don't do that all the time.</p>
            <p>&nbsp;</p>
            <p>But why was the site offline for so long? Well it took me a few hours to notice the issue and a few hours more before I had a chance to attend to it. Priorities have changed, I know now that a site that gets a few hundred thousand page views a month being offline isn't going to be the end of the world. There was another reason that the restoration took a little bit longer than ideal, the last back up I had was a little bit outdated. I also took the opportunity to switch the OS. Previously the server was running Centos 5.4 , now it's running Amazon Linux. I wonder if I will ever get used to always having to <em>sudo</em></p>
            <p>&nbsp;</p>
            <p>Even then Google cache came to the rescus so I only lost a single blog post &nbsp;+ another one that had been in the draft stage. That's something that I can live with.</p>
            <p>&nbsp;</p>
            <p>But hang on a second, twice in two days you said, what happened next? I haven't the foggiest, the site went off line again the today and I had to repeat the whole process all over again. This time the outage was a 'mere' five hours</p>
            <p></p>

        </div>
    </div>


</body>

</html>

The october curse comes early?

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The Financial Times is a publication that appears to be staffed almost exclusively by PR men moonlighting for large companies and stock broker firms. At least 90% of space is set aside for press releases. That’s a lousy statistic but I am adhering to the same standard as the FT.</p>
            <p>Todays’ edition carries an article titled '<a href="http://www.ft.lk/2011/09/01/super-september/#comment-10201">Super </a>September' It argues that September is a great month for the Colombo Stock Exchange and says:&nbsp;<em>‘September has produced an average positive movement of near0% throughout 2002 and 2010 with last year producing the best return of near 24%’ </em>Mark Twain did speak of “Lies, Damn Lies, and Statistics” but the above statement is not statistics it’s just ignorance (or damn lies, take your pick)</p>
            <p>The FT doesn’t say what exactly they mean when the use the term average, but if you add up the reported percentage gains for September in each year and divide by 8, you can arrive at a figure of 9.76, which is the Arithmatic Mean for those figures. The trouble is you cannot use arithmatic means with percentages.</p>
            <p>Consider this. Suppose the price of a security rises 10% in 2008, followed by another0% in 2009 and then declines 10% in 2010 so going by the FT logic, you would say the share has gained an average of (10+10-10)/3 or 6.67% but in actual fact the price of the security at the end of 3 years will be 107.9! hardly a 20% gain is it? A 10% gain in one month followed by a 9% drop get’s you back to square one.</p>
            <p>There are plenty of studies showing that trading stragies based on <em>the hope </em>that&nbsp;prices will rise in a particular month (eg January Effect) or fall in a particular month (October curse) is not an effective trading strategy. Some authors just say don’t trade on hope.</p>
            <p>Many famous and gut wrenching market crashes have happened during the month of October, starting with the grand mother of them all the ‘Great Crash’ of 1929. Then there was the Black Monday of 1987. Old hat you say? Well there was the ‘Mini Crash’ of October, 1997 and very recently the &nbsp;Nikkei plunged 9.4% on October7, 2008</p>
            <p>Want something closer to home? The <strong>CSE</strong> all share plummeted 5% in a single day on October3, 2010. &nbsp;Only the automatic trading halt that’s in place prevented a bigger sell off. Immidiately afterwards the government intervened to stabilize the market. So at the end of the month the All Share Index had lost 4.5% . But look at the figure for 2008; &nbsp;there you can see a &nbsp;15% drop</p>
            <p>It will take a brave man to trade on this newspaper story, on the other hand fools rush in where angels fear to tread and the story could well become a self&nbsp;fulfilling&nbsp;prophesy. If that happens surely the&nbsp;October&nbsp;curse will be&nbsp;fulfilled&nbsp;as well?</p>
            <p>“Wake me up when September Ends” (Green Day (the Band))</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Monthly Archives for 2011 09

Who is the victim?

Site Offline Twice in Two Days

The october curse comes early?

The October Curse