Amazon EMR is a cloud-native big data platform for processing vast amounts of data quickly, at scale. Using open source tools such as Apache Spark, Apache Hive, Apache HBase, Apache Flink, Apache Hudi (Incubating), and Presto, coupled with the scalability of Amazon EC2 and scalable storage of Amazon S3, EMR gives analytical teams the engines and elasticity to run Petabyte-scale analysis.
N/A
Apache Spark
Score 8.9 out of 10
N/A
Apache Spark is a multi-language engine for executing data engineering, data science, and machine learning on single-node machines or clusters.
N/A
Everpure FlashBlade
Score 9.9 out of 10
N/A
Everpure (formerly Pure Storage) offers FlashBlade, a scale-out file and object storage – architected to consolidate complex data silos (like backup appliances and data lakes) while accelerating tomorrow's discoveries and insights.
Director of Customer Operations & Account Management
Chose Amazon EMR (Elastic MapReduce)
EMR is more suited for developers. Databricks feel more for data science-oriented with its notebooks and customs visualizations. With EMR you can more easily add additional capacity on-damnd on the instance. With others is a more cumbersome process. And then, you can also …
The alternatives to EMR are mainly hadoop distributions owned by the 3 companies above. I have not used the other distributions so it is difficult to comment, but the general tradeoff is, at the cost of a longer setup time and more infra management, you get more flexible …
1. Apache Spark is almost 100 % faster than Hadoop. 2. Apache Spark is more stable than Amazon EMR. 3. The end to end distributed machine library is more robust in Apache Spark.
Verified User
Engineer
Chose Apache Spark
Apache Spark has much more better performance and features if we compare with Hive or map/reduce kind of solutions. Spark has many other features for machine learning, streaming.
Even with Python, MapReduce is lengthy coding. Combination of Python with Apache Spark will not only shorten the code, but it will effectively increase the speed of algorithms. Occasionally, I use MapReduce, but Apache Spark will replace MapReduce very soon. It has many …
There are a few newer frameworks for general processing like Flink, Beam, frameworks for streaming like Samza and Storm, and traditional Map-Reduce. I think Spark is at a sweet spot where its clearly better than Map-Reduce for many workflows yet has gotten a good amount of …
We are running it to perform preparation which takes a few hours on EC2 to be running on a spark-based EMR cluster to total the preparation inside minutes rather than a few hours. Ease of utilization and capacity to select from either Hadoop or spark. Processing time diminishes from 5-8 hours to 25-30 minutes compared with the Ec2 occurrence and more in a few cases.
Well suited: To most of the local run of datasets and non-prod systems - scalability is not a problem at all. Including data from multiple types of data sources is an added advantage. MLlib is a decently nice built-in library that can be used for most of the ML tasks. Less appropriate: We had to work on a RecSys where the music dataset that we used was around 300+Gb in size. We faced memory-based issues. Few times we also got memory errors. Also the MLlib library does not have support for advanced analytics and deep-learning frameworks support. Understanding the internals of the working of Apache Spark for beginners is highly not possible.
If data storage, access, and security [are] of the highest priority to your business then Pure Storage FlashBlade is an excellent tool that must be considered. Analytics or sharing that requires the fastest speeds available will benefit from the NVMe solid-state drives they use which are far superior to spinning rust. It is less ideal for those who do not require such time-critical work.
EMR does well in managing the cost as it uses the task node cores to process the data and these instances are cheaper when the data is stored on s3. It is really cost efficient. No need to maintain any libraries to connect to AWS resources.
EMR is highly available, secure and easy to launch. No much hassle in launching the cluster (Simple and easy).
EMR manages the big data frameworks which the developer need not worry (no need to maintain the memory and framework settings) about the framework settings. It's all setup on launch time. The bootstrapping feature is great.
It would have been better if packages like HBase and Flume were available with Amazon EMR. This would make the product even more helpful in some cases.
Products like Cloudera provide the options to move the whole deployment into a dedicated server and use it at our discretion. This would have been a good option if available with EMR.
If EMR gave the option to be used with any choice of cloud provider, it would have helped instead of having to move the data from another cloud service to S3.
When reporting out a user has exceeded there quote, it only references the UID. It would certainly be nice it calls out the UID name that is clearly present in the Dashboard.
The ability to determine a snapshot total size would be helpful.
Proactive reachout to discuss new versions and assist in planning the upgrade would be a key win.
Documentation is quite good and the product is regularly updated, so new features regularly come out. The setup is straightforward enough, especially once you have already established the overall platform infrastructure and the aws-cli APIs are easy enough to use. It would be nice to have some out-of-the-box integrations for checking logs and the Spark UI, rather than relying on know-how and digging through multiple levels to find the informations
If the team looking to use Apache Spark is not used to debug and tweak settings for jobs to ensure maximum optimizations, it can be frustrating. However, the documentation and the support of the community on the internet can help resolve most issues. Moreover, it is highly configurable and it integrates with different tools (eg: it can be used by dbt core), which increase the scenarios where it can be used
I give the overall support for Amazon EMR this rating because while the support technicians are very knowledgeable and always able to help, it sometimes takes a very long time to get in contact with one of the support technicians. So overall the support is pretty good for Amazon EMR.
1. It integrates very well with scala or python. 2. It's very easy to understand SQL interoperability. 3. Apache is way faster than the other competitive technologies. 4. The support from the Apache community is very huge for Spark. 5. Execution times are faster as compared to others. 6. There are a large number of forums available for Apache Spark. 7. The code availability for Apache Spark is simpler and easy to gain access to. 8. Many organizations use Apache Spark, so many solutions are available for existing applications.
Without exception, the contacts with support have been quick and extremely knowledgeable. I do not fear getting an underqualified engineer to assess or work on my arrays. In addition to this support structure, the sales engineers are top notch as well.
Snowflake is a lot easier to get started with than the other options. Snowflake's data lake building capabilities are far more powerful. Although Amazon EMR isn't our first pick, we've had an excellent experience with EC2 and S3. Because of our current API interfaces, it made more sense for us to continue with Hadoop rather than explore other options.
Spark in comparison to similar technologies ends up being a one stop shop. You can achieve so much with this one framework instead of having to stitch and weave multiple technologies from the Hadoop stack, all while getting incredibility performance, minimal boilerplate, and getting the ability to write your application in the language of your choosing.
The NetApp a800 we tested was 14% faster than Pure FlashBlade with NFS workloads. However, NetApp lacked ease of administration and performing simple tasks such as creating multiple NFS volumes required scripting from the command line. Our flashblade contained 15 baldes and our NetApp was a clustered pair with each half containing 24 nvme devices.
It was obviously cheaper and convenient to use as most of our data processing and pipelines are on AWS. It was fast and readily available with a click and that saved a ton of time rather than having to figure out the down time of the cluster if its on premises.
It saved time on processing chunks of big data which had to be processed in short period with minimal costs. EMR solved this as the cluster setup time and processing was simple, easy, cheap and fast.
It had a negative impact as it was very difficult in submitting the test jobs as it lags a UI to submit spark code snippets.
We were able to consolidate 5 different storage platforms of lesser performance onto a single Flashblade and achieve much, much lower latency and higher throughput.
We've been able to reduce the amount of training and configuration required to just Pure Flashblade, instead of 5 different vendors and products.
In addition to our core use cases, Flashblade has capabilities that we are pursuing for some new projects, i.e. analytics data store and the object store features.
