Update (5/30/13):
The voting was really, really tight, but Hortonworks came out on top. Voting was weighted, so each first place vote counted for 3 points, each second place for 2 and third place votes, obviously, received 1 point.
Here are the results:
1. Hortonworks: 763 points
2. Pivotal: 708 points
3. Cloudera 616 points
Digging into the results, it really looks like this was a battle between Pivotal and Hortonworks for the most part (since Cloudera received the bulk of the second place votes). Pivotal did well enough to get added to the Big Data 42 list (well, now 43). However, Cloudera, despite losing here, is still doing well on the Big Data 42 voting, so I’m not going to bump them from the list.
Over the past week or so, I’ve found myself revisiting the new Flickr and trying to figure out just why it feels like the product release falls short. A lot has changed, including what Flickr, at its core, is. A few particular pieces stand out, however, that should be discussed. Not so much as a critique of Flickr, but rather as lessons we can all benefit from as we think about our own products. In no particular order:
While I don’t particularly love the new profile design (the list of photos lacks hierarchy, focus, and context), its biggest problem is a technical one: speed. It doesn’t matter if people enjoy the new layout if the photos don’t load smoothly and quickly. It appears the new design required more photos per page than before in order to feel rich and full. But that meant implementing some lazy-loading that really gets crushed as the user scrolls. Tons of blank spaces and many seconds before the photos pop in.
We’re all responsible for the user experience (not just designers). Does the page respond quickly? What happens when it doesn’t? At Etsy, we have a special admin toolbar that measures page load times and lets us know when a page is above the acceptable load time (and we’re constantly working to shave milliseconds off load times). The most beautiful design in the world doesn’t matter if the person exits/bounces before it shows up.
via venturebeat.com
Plus I'm sure George Oates is cringing at the busy new design
“With the military and startups, there is a bit of romanticism, but once you are in, you clearly see the gravity of your decision and how it gets tested every day,”
via venturebeat.com
The Steve Jobs emails that show how to win a hard-nosed negotiation
via qz.com
Stats
- 500 million page views a day
- 15B+ page views month
- ~20 engineers
- Peak rate of ~40k requests per second
- 1+ TB/day into Hadoop cluster
- Many TB/day into MySQL/HBase/Redis/Memcache
- Growing at 30% a month
- ~1000 hardware nodes in production
- Billions of page visits per month per engineer
- Posts are about 50GB a day. Follower list updates are about 2.7TB a day.
- Dashboard runs at a million writes a second, 50K reads a second, and it is growing.
Software
- OS X for development, Linux (CentOS, Scientific) in production
- Apache
- PHP, Scala, Ruby
- Redis, HBase, MySQL
- Varnish, HA-Proxy, nginx,
- Memcache, Gearman, Kafka, Kestrel, Finagle
- Thrift, HTTP
- Func - a secure, scriptable remote control framework and API
- Git, Capistrano, Puppet, Jenkins
Hardware
- 500 web servers
- 200 database servers (many of these are part of a spare pool we pulled from for failures)
- 47 pools
- 30 shards
- 30 memcache servers
- 22 redis servers
- 15 varnish servers
- 25 haproxy nodes
- 8 nginx
- 14 job queue servers (kestrel + gearman)
![]()
This week, memcached, a piece of software that prevents much of the Internet from melting down, turns 10 years old. Despite its age, memcached is still the go-to solution for many programmers and sysadmins managing heavy workloads. Without memcached, Ars Technica would likely be unable to serve this article to you at all.
Brad Fitzpatrick wrote memcached for LiveJournal way back in 2003 (check out the initial CVS commit here). While waiting for new hardware to help save the site from being overloaded, Fitzpatrick realized that he had plenty of unused RAM spread across LiveJournal's existing servers. He wrote memcached to take advantage of this spare memory and lighten the load on the site.
memcached is a distributed in-memory key-value store that uses a very simple protocol for storing and retrieving arbitrary data from memory instead of from a filesystem. To store a value, a program connects to the memcached server on the default port of 11211 and issues a series of basic commands. (Note: a binary protocol is also supported.)
via arstechnica.com
![]()
PALO ALTO, Calif., May 21, 2013 — At a live event today, VMware (NYSE: VMW) CEO Pat Gelsinger unveiled VMware vCloud® Hybrid Service™, an Infrastructure as a Service (IaaS) cloud operated by VMware and built on the trusted foundation of VMware vSphere®, giving customers a common platform to seamlessly extend their data center to the cloud.
“VMware’s mission is to radically simplify IT and help customers transform their IT operations,” said Pat Gelsinger, CEO, VMware. “Today, with the introduction of the VMware vCloud® Hybrid Service™, we take a big step forward by coupling all the value of VMware virtualization and software-defined data center technologies with the speed and simplicity of a public cloud service that our customers desire.”
vCloud Hybrid Service will seamlessly extend VMware software used by hundreds of thousands of customers into the public cloud. This means customers will be able to extend the same skills, tools, networking and security models across both on-premise and off-premise environments.
via www.vmware.com
All of these questions are variations on "Why did Yahoo! spend one-third of their cash on hand to buy a company that by all accounts is about to run out of money?" Read this post, and hopefully these questions will not need to be asked again!
Join the @CloudFoundry team. Now is the time! http://cloudfoundry.com/jobs
The best Paas cloud is one that you can quickly and easily contribute to. Just ask @andypiper about his post: http://blog.cloudfoundry.com/2013/05/16/want-to-contribute-to-cloud-foundry-come-on-in/
QOTD: "If there's something you can do about it, then why be upset? If there's nothing you can do about it, then why be upset?"
![]()
Last Friday Facebook blocked Path’s “Find Friends” feature over brewing spam complaints. Shortly thereafter Dave Morin, Path’s CEO, proudly stated that “Path does not spam users,” but the tactics he is defending today are the very same practices that he himself cracked down on as “spam” when he was running the Facebook Platform. I watched the crack-down from the front-row at iLike, an early Facebook platform partner.
Compare Morin’s description of Path’s “feature” to the Facebook policy that was put in place while Morin was the head of developer relations for Facebook Platform. The official Facebook policy was that apps were forbidden from doing exactly what Morin now calls “not spam”:
Specifically, an app with a pre-selected (opt-out) checkbox sending messages to your entire address book if you simply pressed “OK” was what Facebook, under Morin’s oversight, considered to be a punishable violation of the terms of use. Let me say that again: the exact form of invitation interface Path uses was specifically called out and forbidden by Facebook under Morin oversight, in order to preserve the sanctity and user-experience of the Facebook platform ecosystem.





Recent Comments