Apache Spark Reviews

With spark SQL we've now the capabilities to analyse very large quantities of data located in S3 on Amazon at very low cost comparing other solution we checked. 

We also use our own Spark cluster to aggregate data on near real time and save the result on MySQL database. 

We've started new projects using the machine learning library ML.

Until Spark we didn't have the ability to analyse this quantity of data we're talking about two TB/hour. So we're now able to produce a lot of reports, and are also able to develop machine learning based analysis to optimize our business. 

We've central access to every piece of data in the company including finance, business, debug etc. and the ability to join all this data together.

Spark is actually very good for batch analysis much more good than Hadoop, it's much simple, much more quicker etc., but it actually lacks the ability to perform real-time querying like Vertica or Redshift.  

Also, it is more difficult for an end user to work with Spark than normal database. even comparing with analytic database like Vertica or Redshift.

We're now using Spark-Streaming and Spark-SQL for almost 2 years. 

We're working on AWS so we need to have a managed environment. We've choose to go with a solution based on Chef to deploy and configure the spark clusters. Tip : if you don't have any devops you can use the ec2 script (provided by spark distro) to deploy cluster on amazon. We've tested it and work perfectly.  

Spark Streaming is difficult to stabilize as you're always dependant to your stream flow. If you start to be late on the consumer you've a serious problem. We've encountered a lot of stability issue to configure it as expected

It's linked to stability in our case it's takes time to evaluate what is the correct size of the cluster you need. It's very important to always add to you jobs monitoring to be able to understand what's the problem. We use datadog as monitoring platform

Yes to make this job we've used a MySQL database. We switch because MySQL is not a scalable solution and we've reach it's limits.

Setup a spark cluster can be difficult. it's related to your clustering strategy. There is 4 solution at least. 

ec2 script : work only on Amazon AWS

Standalone : manually configuration (hard)

Yarn : to leverage your already existing Hadoop environment.

Mesos : to use with your other Mesos ready application

We use Databricks as online DB ad hoc query. It's work on AWS as managed service, it manage for you the cluster creation, configuration and monitoring.

Give a notebook oriented user interface to query any data source using Spark: DB, Parquet, CSV, Avro etc...

Yes we've started to evaluate analytics databases : vertica, exasol, and other for all the them the price was an issue regarding the quantity of data we want to manipulate.

We use this solution for it's anti-money laundering and direct marketing features within a banking environment.

This solution provides a clear and convenient syntax for our analytical tasks.

This solution currently cannot support or distribute neural network related models, or deep learning related algorithms. We would like this functionality to be developed.

There is also limited Python compatibility, which should be improved.

We have used this solution for around seven years, through several versions.

We have found this solution to be stable during our time using it.

This is a very scalable solution from our experience.

We implemented the solution using our in-house team, but the UI was developed using a third party contractor.

The deployment time of this solution is dependent on the requirements of an organization, and the compatibility of the systems they will be using alongside this solution. We would recommend that these are clearly defined when designing the product for the businesses needs.

I would rate this solution a nine out of ten.

In AI deployment, a key step is aggregating data from various sources, such as customer websites, debt records, and asset information. Apache Spark plays a vital role in this process, efficiently handling continuous streams of data. Its capability enables seamless gathering and feeding of diverse data into the system, facilitating effective processing and analysis for generating alerts and insights, particularly in scenarios like banking. 

The most crucial feature for us is the streaming capability. It serves as a fundamental aspect that allows us to exert control over our operations.

It requires overcoming a significant learning curve due to its robust and feature-rich nature.

We have been using it for two years now.

It provides excellent stability. We never faced any issues with it.

It ensures outstanding scalability capabilities.

Opting for Apache Spark, an open-source solution, provides a distinct advantage by offering control over the code. This means you can identify issues, make necessary fixes, and determine what aspects to accept as they are. In contrast, dealing with a vendor may limit control, requiring you to submit requests and advocate for changes based on your business volume with them. This dependency on volume can potentially compromise control. To safeguard both your customers and your business, the choice of an open-source solution like Apache Spark allows for more autonomy and control over the technology stack.

The system's smooth operation relies on deploying a comprehensive container with Kubernetes clusters, configured with essential toolsets. Instrumentation data from the backend is fed back to a central framework equipped with specific tools for driving various processes. In a case involving a customer with Red Hat and Postini clusters, the OpenShift Container Platform, comprising Kubernetes clusters, is used. The tools manage onboarding, infrastructure provisioning, certificate management, authorization control, etc. The deployment spans multiple independent data centers, like telecom circles in India, requiring unique approaches for various tasks, including disaster recovery planning and central alerting, facilitated through SaaS. The deployment process typically takes approximately forty to forty-five days for six thousand servers.

It provides a dual advantage by saving both time and money while enhancing performance, particularly by leveraging my skill sets. 

It is an open-source solution, it is free of charge.

I would give it a rating of seven out of ten, which, by my standards, is quite high.

I mainly use Spark to prepare data for processing because it has APIs for data evaluation. 

The most valuable feature is that Spark uses Scala, which has good data evaluation functions. Spark also supports good distribution on the clusters and provides optimization on the APIs.

Spark could be improved by adding support for other open-source storage layers than Delta Lake. The UI could also be enhanced to give more data on resource management.

I've been using Spark for six years.

Generally, Spark works correctly without any errors. It may give out some errors if your data changes, but in that case, it's a problem with the configuration, not with Spark.

The cloud version of Spark is very easy to scale.

The initial setup is not complex, but it depends on the product's component on the architecture. For example, if you use Hadoop, setup may not be easy. Deployment takes about a week, but the Spark cluster can be installed in the virtual architecture in a day.

Spark can handle small to huge data and is suitable for any size of company. I would rate Spark as eight out of ten. 

When we receive data from the messaging queue, we process everything using Apache Spark. Data Bricks does the processing and sends back everything the Apache file in the data lake. The machine learning program does some kind of analysis using the ML prediction algorithm.

The memory processing engine is the solution's most valuable aspect. It processes everything extremely fast, and it's in the cluster itself. It acts as a memory engine and is very effective in processing data correctly.

There are lots of items coming down the pipeline in the future. I don't know what features are missing. From my point of view, everything looks good.

The graphical user interface (UI) could be a bit more clear. It's very hard to figure out the execution logs and understand how long it takes to send everything. If an execution is lost, it's not so easy to understand why or where it went. I have to manually drill down on the data processes which takes a lot of time. Maybe there could be like a metrics monitor, or maybe the whole log analysis could be improved to make it easier to understand and navigate.

There should be more information shared to the user. The solution already has all the information tracked in the cluster. It just needs to be accessible or searchable.

I started using the solution about four years ago. However, it's been on and off since then. I would estimate in total I have about a year and a half of experience using the solution.

The stability of the solution is very, very good. It doesn't crash or have glitches. It's quite reliable for us.

The scalability of the solution is very good. If a company has to expand it, they can do so.

Right now, we have about six or seven users that are directly on the product. We're encouraging them to use more data. We do plan to increase usage in the future.

I'm a developer, so I don't interact directly with technical support. I can't speak to the quality of their service as I've never directly dealt with them.

We did previously use a lot of different mechanisms, however, we needed something that was good at processing data for analytical purposes, and this solution fit the bill. It's a very powerful tool. I haven't seen other tools that could do precisely what this one does.

The initial setup isn't too complex. It's quite straightforward.

We use CACD DevOps from deployment. We only use Spark for processing and for the Data Bricks cluster to spin off and do the job. It's continuously running int he background.

There isn't really any maintenance required per se. We just click the button and it comes up automatically, with the whole cluster and the Spark and everything ready to go.

I'm unsure as to how much the licensing is for the solution. It's not an aspect of the product I deal with directly.

We're customers and also partners with Apache.

While we are on version 2.6, we are considering upgrading to version 3.0.

I'd rate the solution nine out of ten. It works very well for us and suits our purposes almost perfectly.

We are working with a client that has a wide variety of data residing in other structured databases, as well. The idea is to make a database in Hadoop first, which we are in the process of building right now. One place for all kinds of data. Then we are going to use Spark.

I have worked with Hadoop a lot in my career and you need to do a lot of things to get it to Hello World. But in Spark it is easy. You could say it's an umbrella to do everything under the one shelf. It also has Spark Streaming. I feel the streaming is its best feature because I have extracted to enter data and analysis within Spark Stream.

I think for IT people it is good. The whole idea is that Spark works pretty easily, but a lot of people, including me, struggle to set things up properly. I like contributions and if you want to connect Spark with Hadoop its not a big thing, but other things, such as if you want to use Sqoop with Spark, you need to do the configuration by hand. I wish there would be a solution that does all these configurations like in Windows where you have the whole solution and it does the back-end. So I think that kind of solution would help. But still, it can do everything for a data scientist.

Spark's main objective is to manipulate and calculate. It is playing with the data. So it has to keep doing what it does best and let the visualization tool do what it does best.

Overall, it offers everything that I can imagine right now. 

I have been using Apache Spark for a couple of months.

In terms of stability, I have not seen any bugs, glitches or crashes. Even if there is, that's fine, because I would probably take care of it and then I'd have progressed further in the process.

I have not tested the scalability yet.

In my company, there are two or three people that are using it for different products. But right now, the client I'm engaged with doesn't know anything about Spark or Hadoop. They are a typical financial company so they do what they do, and they ask us to do everything. They have pretty much outsourced their whole big data initiative to us.

I have used MapReduce from Hadoop previously. Otherwise, I haven't used any other big data infrastructure.

In my work previously, not in this company, I was working with some big data, but I was extracting using a single-core off my PC. I realized over time that my system had eight cores. So instead, I used all of those cores for multi-core programming. Then I realized that Hadoop and Spark do the same thing but with different PC's. That was then I used multi-core programming and that's the point - Spark needs to go and search Hadoop and other things.

The initial setup to get it to Hello World is pretty easy, you just have to install it. But when you want to extract data from your HDFS and other sources then it is kind of tricky because you have to connect with those sources. But you can get a lot of help from different sources on the internet. So it's great. A lot of people are doing it.

I work with a startup company. You know that in startups you do not have the luxury of different people doing different things, you have to do everything on your own, and it's an opportunity to learn everything. In a typical corporate or big organization you only have restricted SOPs, you have to work within the boundaries. In my organization, I have to set up all the things, configure it, and work on it myself.

I would suggest not to try to do everything at once. Identify the area where you want to solve the problem, start small and expand it incrementally, slowly expand your vision. For example, if I have a problem where I need to do streaming, just focus on the streaming and not on the machine learning that Spark offers. It offers a lot of things but you need to focus on one thing so that you can learn. That is what I have learned from the little experience I have with Spark. You need to focus on your objective and let the tools help you rather than the tools drive the work. That is my advice.

On a scale of 1 to 10, I'd put it at an eight.

To make it a perfect 10 I'd like to see an improved configuration bot. Sometimes it is a nightmare on Linux trying to figure out what happened on the configuration and back-end. So I think installation and configuration with some other tools. We are technical people, we could figure it out, but if aspects like that were improved then other people who are less technical would use it and it would be more adaptable to the end-user.

Organisations can now harness richer data sets and benefit from use cases, which add value to their business functions.

Distributed in memory processing. Some of the algorithms are resource heavy and executing this requires a lot of RAM and CPU. With Hadoop-related technologies, we can distribute the workload with multiple commodity hardware.

Include more machine learning algorithms and the ability to handle streaming of data versus micro batch processing.

At times when users do not know how to use Spark and request a lot of resources, then the underlying JVMs can crash, which is a big sense of worry. 

No issues.

I use Spark to run automation processes driven by data.

Apache Spark helped us with horizontal scalability and cost optimizations.

The most valuable feature is the grid computing.

An area for improvement is that when we start the solution and declare the maximum number of nodes, the process is shared, which is a problem in some cases. It would be useful to be able to change this parameter in real-time rather than having to stop the solution and restart with a higher number of nodes.

I've been using Spark for around four years.

The initial setup was not easy, but we created a means of asking the user about their needs, making the setup much easier. We can now deploy the platform in thirty minutes using the public cloud or Kubernetes space.

Spark helps us reduce startup time for our customers and gives a very high ROI in the medium term.

Spark is an open-source solution, so there are no licensing costs. 

I would rate Apache Spark eight out of ten.