Fast again
January 30, 2010
For the past month, I’ve been working on making the BookMooch web site fast again. I think I’ve finally licked it.
I keep a daily chart of how many seconds it takes to load up the home page. Here is what the chart looks like over a year and a half:
You can see that the speeds were quite good from October 2008 until August 2009 when they started to get worse (1 and 2 seconds being typical) but still pretty good. Then, something happened around December 2009, when things got really bad, taking typically more than 5 seconds to load the BookMooch home page.
Initially, I thought it was simply that we needed a new server. So, I upgraded the server, and that helped a little bit, but not nearly enough.
Then I thought that this was caused by the search feature not being fast enough to keep up. So, I rewrote parts of it and removed a few extraneous features that took a lot of time, That speed a search for “Bertrand Russell” from 2.6 seconds to only 0.6 seconds. Much faster!
Unfortunately, that didn’t help either: the web site was still slow.
Then I started looking at the usage logs to see what was going on. I had 16 CPUs on the server, and when I shut BookMooch down, 14 of them were doing full text searches. Hmm.. And… 10 of those searches came from one search engine named cuil.com. Apparently, they’re a brand new search engine, and are trying to distinguish themselves by deeply indexing the web. Their indexer was hitting my full text searching feature pretty hard.
So, I banned cuil.com completely from my site. I also banned Google and Yahoo and Baidu from using the full text feature to index BookMooch. Those sites can (and do) still index BookMooch, but they stay away from pages that are full text searches (indexing book details pages, mostly). A normal page on BookMooch takes about 2/100th of a second of the server’s time, whereas a complicated full text search can take seconds.
Bang! That was it!
The load on the server plummeted, which you can see on this chart:
There are two days of pre-ban usage on the left, where the load is very high, around 17, and then after the ban on the right side, where the server load drops to about 1/4 the previous level, to about 4. The server was a lot less stressed out.
Going back to my chart of how many seconds it takes to load the BookMooch home page, we have good news:
You can see that it used to take more than 5 seconds to load the home page, but for two days now it’s been less than 1/5th of a second to do so.
Thanks, everyone, for bearing with me as I figured this all out.
Two changes have occurred to the search results, as part of making searching work faster:
a) the search results page no longer displays 3 “related searches” for you automatically. If you want to see related searches, there’s a “RELATED SEARCHES” button toward the top right of the page.
b) the bottom of the search results page used to show you the other languages that the books from your search were in. This is no longer shown. It was noisy and not very useful and turned out to take a lot of time to calculate. I think most people use “advanced search” if they want to narrow their search down to books of a particular language.
Now that I’ve fixed the speed problems, I’m back on BM, working on new features.
One last piece of good news: now that BM is fast again, usage of the web site has climbed 45% in the past two days.
-john
Well done!
John
Thank you so much for your hard work. I’ve always appreciated you but since the sudden and totally unexpected death “Frugal” you reach much, much higher in my appreciation for you!
A Thomas
As always, thank you for all your wonderful work to keep this site humming along. I am always impressed with your technical abilities. Have a great weekend.
Thanks John I love fast. This may explain why PBS is so slow right now.
Bravo, John!
Thank you!
I’ve been in a similar boat with the sites on work on, both cuil and msn are typically, in my experience, “bad” netizens. Often hitting the sites with far too many requests and completely ignoring robots.txt. Really happy, though, that you found the problem and things are running super fast now!
Thanks for this John! We really appreciate your efforts in keeping Bookmooch alive and well!
Thank you so much for all your hard work! Wow..the improvement is wonderful! I also love how we can keep up via Twitter..I have gotten quite a few books that I would never have known existed!! Thank you again…
Hi John,
Thank you very much. Your efforts are appreciated.
Hurray! Thanks
What a treat to have someone care so much and work so hard to improve things – you are appreciated.
Wow! What a saga! Thank you for persisting and getting Bookmooch fast again.
YOU ARE AMAZING. MANY MANY THANKS!
Again Thank you for all you do!! Love this site!! We appreciate you more than you will ever realize!
Thank you so much for your efforts. Much appreciated.
Great work!
BookMooch is worth waiting for even if it itakes longer. There’s nothing like it.
Ditto X however many compliments above mine!
Thanks for all your hard work – it’s a lot quicker now!
What a pleasure to be able to know what is going on from such a caring person!!!
Thank God! Bookmooch is one of my favorite sites and it was a little frustrating to have it run so slow. But now its back to normal if not blazing speeds!
I have a couple of friends working for cuil and google. I should tell them about this interesting episode.
Thanks, thought it was my computer
Thanks, John! Makes me a very happy camper.
Thank you so much for everything you do to keep this site up and running!
It’s sooo fast! I love it. This is the way a website should be. I know there was trouble with the website before because sometimes it would have trouble loading. One time recently I couldn’t mark a book received and I could barely get into Book Mooch that day. Here’s hoping it stays fast for a long, long time.
Awesome! I’ve been kinda staying away from BM the past few months, due to the slowness. Time to get back to searching other peoples’ inventories for books
Just joined a few weeks ago, love the site, but thought it was super slow–it’s working wonderfully now!
good to know that. thanks John!
John,
You are the best. Your dedication to improving the site, which is such a boon to readers like myself, is not only admirable but inspiring. Thanks!!!!!
Thank you! I would love to have a fast BM Message Board also. It’s almost impossible slow at the moment and emails is not an option because they are to many.
But I would love to read and learn from other members more often via a MB.
Nice work!!
I am so happy to see the pages loading quickly again. I almost left. So happy you are dedicated and fixed the problem. I have been referring people on freecycle to list books here.
Oh, thank heaven… I thought it was my fault for visiting the Web site too much.
Thanks for all your hard work, John!
Thanks for your dedication to provide this superior service. You are doing an awesome job!
Great work John! I thought it was my computer and was seriously trying to decide if it was time for ME to update.
I agree with all the other thankful Moochers. I hope you’ve got your Chief Detective Inspector’s badge for solving this!
Thank you, I had no idea what was going on – Seriously appreciate your efforts.
Thanks John, you are so dedicated, does it feel like work or is it purely pleasure?!!! Either way much appreciated by the masses.
WOW! Talk about the proverbial needle in the haystack! Thanks so much for your time. This is truly an awesome website and it is great to have it speedy again.
yey john!!!!!
it all sounds very complicated but the speed is good.Always a pleasure , always a happy moocher -except with the won’t send to my cuntry stuff
oops …i meant country….
I work for a company with a very highly trafficked web site, and I run a few small web sites. My company site’s RSS feeds were hit hard by a service called Collecta that more than doubled our already high traffic load until we banned them. The big and small ones have also been hit aggressively by Cuil. The Bing/MSN/Live spiders also submit lots of requests using fake referrers.