Google Cloud Datalab Reviews

It's for our daily data processing, and there's a batch job that executes it. The process involves more than ten servers or systems. Some of them use a mobile network, some are ONTAP networks, and they're on some kind of system. Everything comes through the Microsoft service. It's not a tentative process, but when the accounting system requires it, a batch job processes the data and secures it in Datalab.

In GCP, there's an engine called Dataflow. You need to put some scripts in it. This process happens automatically every day around midnight. It gets executed, processes the data, and returns it to the database.

The dashboards are good. When you want to show data in a dashboard, like in pipelines, it can be done by the end-user as well. You can only put limited data (JSON) in Datalab. 

For example, if you want to show captions in a bar search, you need to provide the information to the data engine. You need to specify that it is a relation in GCP. It will be automatically displayed in the dashboard. That is the advantage of using GCP. 

I believe GCP is more costly than other cloud platforms because of this feature for end-users. The data visualization is more meaningful for the end-user compared to Azure Cloud. I have worked with Azure Cloud as well, and Google Cloud provides more meaningful information in its dashboards.

There are a lot of AI feature as well. One AI feature I found is auto-completion. GCP also uses AI for recording purposes and data management in Data Proc, maintaining logs. You just need to mention where to store the logs, either in the cloud or in a data directory. It's very easy.

However, there are limitations with GCP's AI. You need to configure it based on the limits of the nodes. If it goes beyond the limit, it should go to the next node, but that doesn't always happen properly. It could be due to the wrong configuration, but it's not immediately moved to a separate node. You need to restart the server and reconnect it to another node.

Access is always via URL, and unless your network is fast, it would be a little tough in India. In India, if we had a faster network, it would be easier. In a big data environment, like when forcing your database with over a billion records, it can be tough for the end-user to manage the data. You need to have a single entity system in each environment. It's not because of GCP, but it would be great to have options like MongoDB or other similar tools in GCP. Then, we wouldn't always need to connect to the cloud and execute SQL queries.

Even if your application is always connected to its database, the processing can be cumbersome. It shouldn't be so complicated. Once the data is collected, it should be easily sorted.

I have been using it since 2017. 

For me, it has been a stable product. We can manage it as long as it's there. Maybe some requests may not be provided in the database or reporting for Cloud, but overall, I can manage it.

Maybe if the data becomes larger, there could be changes or it might not be able to store it anymore. But as of now, there are no complaints, so I am happy.

Last time, I scaled up to 94 nodes. I'm not sure how much it can be scaled for the hardware. Normally, on my system, it's connected to Linux, and I can configure up to 77 nodes. It will work properly until then, but if you go beyond 99 nodes, you will need a separate server. You need to configure the next server and install the unit. There are some steps you need to follow. Then, you need to set up your microservices so they can be executed when you call one. That also needs to be configured. 

If your system is using the same microservices, you need to pass the same information you have. Generally, they can do cross-solution directly. Once they do the cross-origin request (CORS), they will connect. So, microservices and the database are separate, and they can connect to the database via microservices only. You need to pass it through cross-routing.

We make a call and send them the ticket number, and they use the address there. We get the response immediately. We just make and raise the ticket, then make a call, and they will prioritize it based on urgency, find it separately, and then we will get the response immediately.

Positive

The solutions are all the same, with the same features and purpose. We can achieve the same results using other tools like SQL Server or microservices. I am interested in using GCP because it's available everywhere, and I can access it wherever I go. I choose it for the performance demands of my work.

The initial setup is not a big thing. The documentation is there. I've watched videos and referred to the documentation. You can do it yourself. If you have any issues, you can optimize it in Google Cloud. It's a bit different from other platforms, but if you follow the documentation, you can do it. Everything is documented by Google Cloud, so you can set it up yourself.

You can do it yourself. 

Overall, I would rate it a nine out of ten. 

Google Cloud is very good. Once you go through the features of Google Cloud, it's a good idea to get a GCP certification so you have an idea of how it can be configured. Most things will be covered in the certification. Once you have around 80-84% knowledge of GCP, you can start using it.  

Other services, like S3 in AWS or Azure Cloud, provide the same. But I am confident in GCP. Azure Cloud is also good for me, but I haven't done any bucket tracking directly on it. Both Azure Cloud and GCP are good for me.

Our main use cases involve transferring workloads from AWS and Univision to Google Cloud Datalab. Before coming to the setting we utilised Google Datalab for looker and handling separated tables for research and development scenarios. Currently in association with AWS and Univision, we are focusing and migrating, defining data set tables and moving data to the Google hosted platform.

Since we are working extensively with the solution in managing data sets and regional databases, we focus a lot on handling different table segmenting data based on ids and customising the architecture for unit projects. There was a challenge which included and rose from the rapid and regressive pace of development since we transitioned from AWS to GCP. Adapting to the logical data structure is a complex process because it is fast paced and dynamic due to the project nature. Google Cloud Datalab helps us to incorporate different components and utilise univision simultaneously. It is very customizable.

We have also encountered challenges during our transition period in terms of data control and segmentation. The management of each channel and data structure as it has its own unique characteristics requires very detailed and precise control. The allocation should be appropriate and the complexity increases due to the different time zones and geographic locations of our clients. The process usually involves migrating the existing database sets to gcp and ensure data integrity is maintained. This is the only challenge that we faced while navigating the integers of the solution and honestly, it was an interesting and unique experience. 

I have been using Google Cloud Datalab for quite some time now.

Regarding the technical support, I have worked with the various vendors that are Microsoft, AWS and Google while engaging in architecting projects. My approach always involved maintaining an open mind towards technology and considering various factors such as feasibility as specific use cases. In my opinion different vendors have specialised expertise and some collaborations can be challenging as they focus on specific areas without holistic understanding of the full cycle. This specialisation has led to issues even in Google Cloud data lab and their support. 

I believe that rating depends on the individual experience and this is a very subjective matter. Understanding factors like collaboration and approach, plays a very important role. In my opinion, Google might have missed some opportunities and is facing challenges compared to competitors like AWS and Azure. However, based on the open source and their association with Docker I will rate them around 6.5 to 7.

The solution is really useful. It’s an easy way to get information. I use it as a reference for analytics, sourcing information, and research.

The APIs are valuable. I also use the open-source things available in the solution.

The product must be made more user-friendly. Sometimes, we have to go a roundabout way and read a lot of instruction that isn't necessary. Generally, if people use the information, they have some knowledge about it. If they don't, there should be an introduction section specifically for it. The rest of it should skip all the instructions.

I have been using the solution since it was started.

The tool’s stability is pretty good. I rate the stability an eight out of ten.

I’m the only user. I use AI a lot. The tool’s scalability depends on who we have contracts with.

We can't talk to anybody. It's all AI-based. It would be good if the vendor kept up on their models. If the AI hasn’t been trained, it tends to go in circles. When we ask to speak to customer support, the AI goes round and round to keep us from doing it.

Now that I am a partner, I have a higher level of support. I can get a hold of administrators directly. However, in the beginning, the usability was bad. The AI sent us in circles when we asked for customer support because it didn't want us to talk to support.

Negative

The initial setup was straightforward. I wish the websites would ask us in the beginning whether or not we know how to use the tool. It could skip all the demos and introductions if we know it. The deployment time depends on how fast we enter the information. If we use templates, it is easy. Depending on our speed, we could set it up in minutes or hours.

I can do the deployment myself. We need to do regular upkeep. We can opt-in for automatic updates from Google. It depends on what cloud we use.

I use the free API versions. The product is cheap. I rate the pricing a two out of ten.

I use the products from Microsoft, Google, and IBM. I use many open-source tools, copy their APIs, and read their documents. Microsoft is easier for API and copying information because we can go over to the documents and click copy.

In Google, we must enter our payment methods even if we're not paying for a service and want to use the free service. It is annoying. If we don't use Google-specific things, it won’t allow us to connect to other cloud services without paying for the subscription. Microsoft is 99% free.

I use a lot of Atlassian and Jira products. I do a lot of code for a lot of bigger-name websites. I apply it through Jira, which is connected to my Google Cloud, which is also connected to my Azure. The tool’s usability is pretty good. Overall, I rate the product an eight out of ten.

Currently, we're currently provisioning some ML models into Vertex AI. The sales team will likely start using the other resources in a month or two. 

When deploying the model on-premises, with a global team, we faced time lag issues. On Google Cloud, there is a facility to select different locations, making the application available on servers almost anywhere. That's why we shifted to the cloud.

It's a data science platform, and we use various resources from Google Cloud, depending on project permissions.

In MLOps, when we are designing the data pipeline, the designing of the data pipeline is easy in Google Cloud.

There is room for improvement in the graphical user interface. So that the initial user would use it properly, that would be a good option.

We have been using this solution for a year now. We use the latest version. 

It is a stable solution. 

The scalability is average. We have not faced any issues with scalability. 

There are more than 500 end users using this solution in our company. It is an integral part of the daily operations. The usage pattern is not a one-time thing; employees regularly access and utilize the application.  

We use it at a global level with a scattered user base. This means that users don't all use the application at the same time. So, around 300 out of 500 employees use the solution, and this usage is spread out throughout the day.

We're using Vertex AI for MLOps in Google Cloud.

The initial setup is easy. I would rate my experience with the initial setup a nine out of ten because we'll have to depend on the DBS team because we are not the full expert in the cloud. 

So maybe there is a dependency on the cloud practitioner to provision all these facilities.

Previously, we used an on-premises server, but we faced some issues deploying our models. Our organization suggested moving to the cloud.

It's an enterprise edition of the cloud that has been provided to us.  

Deployment is an easy task on the cloud because the deployment link is readily available on Google Cloud. So, we only needed to embed that deployment link on different platforms, like for websites or anything else on the mobile application. It's easy.

For deployment, in the case of MLOps, I don't require any help because it's in the process. So, from model building to deployment, I can do it on my own.

For the maintenance part, we'll only look out at how the model is performing. So that is the only part of maintenance. Another case is when whenever there will be more users come to the platform, then we have the unicorn in place so that they will take care of that scaling part.

It is affordable for us because we have a limited number of users. 

Overall, I would rate the solution an eight out of ten. I would definitely recommend using the solution. 

We use it to perform Exploratory Data Analysis on our dataset, which focuses on machine learning. Through Datalab, we conduct an in-depth examination of the data, identifying key features for our machine-learning models. Additionally, we leverage it to enhance data quality by addressing missing values, applying preprocessing techniques, and ensuring that the data aligns with the specific requirements of our customers. This involves employing statistical methods to analyze data patterns and make informed decisions about data processing.

The standout feature is its robust infrastructure and environment. The infrastructure is highly reliable and efficient, contributing to a positive experience. Setting up the environment, including libraries for development in languages like Python, is seamless and enhances the overall usability. It allows users to concentrate on deriving value from data without the distraction of dealing with time-consuming tasks related to infrastructure and environment configuration.

Connectivity challenges for end-users, particularly when loading data, environments, and libraries, need to be addressed for an enhanced user experience. The short session timeout is disruptive, especially across different time zones, and improving session management would contribute to a smoother development process. Enhancements in the integration with BigQuery are essential, focusing on addressing performance and usability issues, especially with large datasets. On a broader scale, enriching the UI settings within the Google Cloud Console for Datalab is crucial. This includes implementing features for unified settings in environments and facilitating easier navigation and management for end-users.

I have been working with it for approximately three years.

I would rate its stability capabilities eight out of ten. The emphasis is on simplicity, eliminating the need for extensive knowledge of technologies like DevOps or intricate infrastructure setups.

Its interface is familiar to a broad user base, especially those involved in data-related fields worldwide. Its ease of use is attributed to its foundation in open-source technologies, making it a tool that resonates well with diverse users in the data community. Our client base is extensive. I would rate it nine out of ten.

The technical support is commendable, as they primarily focus on addressing technical issues related to infrastructure and the environment. I would rate it nine out of ten.

Positive

The initial setup can be somewhat inconvenient. Users might find it less user-friendly, particularly when encountering technical issues. For instance, when data errors occur, there's a need to restart the system, which involves either reinstalling certain components or reconfiguring the environment. This process can be time-consuming, requiring users to invest additional time in restarting and reconfiguring the environment.

The deployment time is impressively quick, typically taking ten to fifteen minutes. The process involves accessing the Google Cloud Console, navigating to the appropriate product section, and creating an instance. During instance creation, selecting the right specifications, such as CPU, RAM, and the required libraries is crucial. This can be done through manual configuration or opting for a predefined environment.

The pricing is quite reasonable, and I would give it a rating of four out of ten. It is all-inclusive, there are no additional payments required.

Overall, I would rate it eight out of ten.

We are using this solution to help manage personnel and to see if everyone is in the right place.

The biggest difference between this solution and competing products is the call-up feature, which is free to use.

All of the features of this product are quite good.

The interface should be more user-friendly.

The security should be easier to set up.

TensorBoard is available but it is hard to use.

I have been using Google Cloud Datalab for three months.

Google Cloud Datalab is quite stable. It is being used on a weekly basis.

We have fewer than five people working on this product right now because it is not in production.

We have not been in contact with technical support.

Prior to this solution, we used Google AutoML and we also tried BigQuery ML.

 I have also been using H2O.ai.

It was easy to set up.

We implemented this solution with our own in-house team.

I find that Google Cloud Datalab is okay for now. I would recommend that people use the Google Platform and especially machine learning.

I would rate this solution an eight out of ten.