Databricks ne vous convainc pas tout à fait ?
Comparer avec une alternative populaire
Alternatives à Databricks
Avis sur Databricks
Powerful tool for dev ops of machine learning models
Commentaires : Overall, my experience with Databricks has been very positive. It is a powerful tool to enable data scientists without a lot of data engineering skills. However, you need to be a data scientist or machine learning engineer to be able to take advantage of its power for machine learning.
I love how easy it is to deploy auto-scaling machine learning models. After a machine learning model is trained, you can just click a button to deploy the model, I believe in a container, and have it auto scale as needed. You can also specify the minimum and maximum size of the deployment to reduce costs but to keep up with the workload as necessary. It is also built around Spark, so tasks involving "big data" aren't an issue.
Some of the cons are that the primary language is Java/Scala, whereas many data scientists are using python or R, which run slower on Databricks than Java and Scala. Also, the main interface via coding, which can limit a lot of citizen data scientists.
An unified platform to develop high quality analysis
Commentaires : Databricks is allowing data analysis that other systems could not perform at the same performance because it is a platform that integrates huge amounts of cloud data with Scala, Python, SQL or R notebooks in a user-friendly interface. Due to the features of Databricks, daily work seems more efficient and less bureaucratic.
What I like most about Databricks is the amount of integrations the platform provides to the user. With Databricks, you can create datasets, develop machine learning models, and analyze performance automatically by setting up a job periodically. Whether the user is an engineer, data scientist, or business analyst, Databricks can streamline everyone's work.
What I least like about Databricks is the instability that usually occurs when there are too many users trying to run their notebooks on the same cluster at the same time.
Very good to handle very big data
Commentaires : While it supports python, when I need to use it, I ultimately prefer to sample or aggregate and export data to work in another environment. For this end, it works very well.
- Enables simultaneous collaborative work with colleagues - Easy to mix spark queries and python for extra analyses and plots - Handful visualization modes for query results (tables and plots with aggregations)
- Hard to manage notebook workspace - Sometimes it gets really slow to run queries - AFAIK, there aren't visualization options for datasets (without running queries)
Powerful tool for data analysis
Commentaires : Overall databricks is very good, and if optimized correctly, can let you work with big scale datasets.
I love the z indexing, which allows for really fast querying of data. Optimized by spark it is great.
The data visualization are subpar. I wish there were better libraries to integrate and visual the data.
Alternatives envisagées :
Modern Analytics with High Flexibility
Commentaires : Positive - after implementing at our company, significant data automation has reduce the amount of time it takes to get in the proper format in the correct people's hands. No more late information that has negative consequences,
Databricks was able to pull data from our core and create specialized dashboarding / reporting that automated a host of manual process that took hours per week. It is now totally hands off and management get review the data in just a few clicks.
It can be extremely confusing given the sheer breadth of tools available. The initial setup and connections certainly require an experienced professional, but once up and running, less-technical users can utilize.