Generating big data sets for search engines

NOTE: This is the English version. You will find the French version further down in this article.

When proposing our expertise search, we are often asked to do performance evaluations on large datasets, for instance in Proof of Concepts. For a recent customer request, in order to gain time and to not use sensitive customer data, we have used log-synth, a random data generator developed by Ted Dunning. We are describing here how to use log-synth in order to generate a 100.000 lines data set.

The first step, which we don’t document here, is about downloading log-synth, unzipping it and building it with maven.

Continue reading

Enterprise Search Europe in London – Open source focus

NOTE: this post has a French version at the bottom of this page.

Enterprise Search Europe is the largest european event dedicated to Enterprise Search. Looking at this year’s agenda, I have the feeling a particular highlight will be given to open source. As in the recent years, several case studies are dedicated to open source, but in addition, the keynote will be focused on it. Charlie Hull, CEO and cofounder of Flax, expert in open source enterprise search, will be sharing his thoughts on the future of search and the link betweeb search and big data. Other open source tracks include a migration from Exalead to Apache Solr (the talk will be given by France Labs, yeeepieeeee), and a round table on open source implementation. You can find more details on the ESEU 2015  programme page.

Continue reading

Tutorial – Deploying Solrcloud 5 on Amazon EC2

NOTE: There is a French version to this tutorial, which you’ll find on the second half of this blog entry.

In this tutorial, we’ll be setting up a Solrcloud cluster on Amazon EC2.
We’ll be using Solr 5.1, the embedded Jetty, Zookeeper 3.4.6 on Debian 7 instances.
This tutorial explains step by step how to reach this objective.

We’ll be installing a set of 3 machines, with 3 shares and 2 replicas per shard, which gives us a total of 9 shards.
We’ll also be installing a Zookeeper ensemble of 3 machines.

This architecture will be flexible enough to allow for a fail-over of one or two machines, depending on whether we’re at the indexing phase or at the querying phase:

  • Indexing: a machine can fail without impacting the cluster (the zookeeper ensemble of 3 machines allows for one machine down). The updates are successfully broadcasted to the machines still running.
  • Querying: two machines can fail without impacting the cluster. Since each machine hosts 3 shards, a search query can be processed without problems, the only constraints being a slower response time due to the higher load on the remaining machine.

Continue reading

Datafari on Docker

NOTE: This is the English version. For the French version, please scroll down.

UPDATE 08/08/16 : update of the post for Datafari v3

UPDATE 01/04/16 : beware that there is a bug with Docker toolbox 1.9.1 for the use of Cassandra (which is a component of Datafari). Update your Docker to 1.10+
https://github.com/docker/docker/issues/18180

This time, we’ll talk about the release of Datafari on Docker.

If you don’t know it yet, Docker is an emulation mechanism that works at a low level of the Linux kernel, hence making it faster than widespread technologies of virtualisation such as VMWare. As its name suggests, you can “dock” applications in an isolated manner, and it will work as a standalone system on your OS.

Although we recommend installing Datafari alone on systems when used in a productive environment, using Datafari on Docker allows you to quickly install Datafari without impacting the configuration and packages in place in your system. Just download the docker image, and the remainder is being taken care of by Docker.

Continue reading

Schemaless Solr

NOTE: French version at the bottom of this page.

We can often see on the web that Elasticsearch is really cool because it is schemaless, and Solr is not. Although Elasticsearch is cool for many reasons, we want to remind you that Solr is also schemaless since July 2013 (Solr 4.4).

To remind you what schemaless means: Without manually editing the Solr schema, it can recognize some data types  automatically when receiving data to be indexed. Those types are: Boolean, Integer, Long, Float, Double, and Date

That’s pretty convenient for quick prototyping. Still, as for Elasticsearch, Continue reading

Tutorial for setting up SolrCloud on Amazon EC2

UPDATE: This tutorial is based on Solr 4. If you want to use Solr 5, we strongly recommend to use our recent blog entry to set up Solrcloud 5 on Amazon EC2

NOTE: There is French version to this tutorial, which you’ll find on the second half of this blog entry.

In this tutorial, we’ll be installing a SolrCloud cluster on Amazon EC2.
We’ll be using Solr 4.9, Tomcat 7 and Zookeeper 3.4.6 on Debian 7 instances.
This tutorial will explain how to achieve this result.
We’ll be installing a set of 3 machines with 3 shards and 2 replicas per shard, thus creating a set of 9 shards.
We’ll also be installing a Zookeeper ensemble of 3 machines.

Continue reading

Tutorial for combining ManifoldCF and Elasticsearch for files search

With the arrival of Manifold CF 1.0 (now already in v1.6.1), the open source community is looking for tutorials to combine it with Elasticsearch. That’s the intent of this tutorial, which will drive you through the different steps required to make it work.

First, we’ll recap the installation process of Manifold CF (we’ll call it MCF later on). Second, we will install ElasticSearch with the attachment plugin so that it handles rich document indexing. Third, we’ll configure MCF so that it crawls a windows file share and indexes documents in ElasticSearch. In this tutorial, when I specify installation directory such as apache-manifoldcf-1.6.1, you have to complete with the absolute path of the installation directory.
Continue reading

Binary version of SolrMeter for Solr 4

NOTE: English version on top, French version below.

We have noticed that in Solr 4, there is problem with the UI related to cache hit ratio evaluation of SolrMeter. Digging a bit, the problem is due to a type change between Solr 3 and Solr 4. SolrMeter expects a string, whereas Solr4 sends back a float. More precisely, Solr 4 does that within its request handler mbean, in the cache sub category.

We’re now using a patch available for this bug, created by Javier Mendez, see his contribution on this google group.

Still, there is no binary version of SolrMeter, hence this blog. Continue reading