Alternatives to KSQL

The algorithms and data infrastructure at Stitch Fix is housed in #AWS. Data acquisition is split between events flowing through Kafka, and periodic snapshots of PostgreSQL DBs. We store data in an Amazon S3 based data warehouse. Apache Spark on Yarn is our tool of choice for data movement and #ETL. Because our storage layer (s3) is decoupled from our processing layer, we are able to scale our compute environment very elastically. We have several semi-permanent, autoscaling Yarn clusters running to serve our data processing needs. While the bulk of our compute infrastructure is dedicated to algorithmic processing, we also implemented Presto for adhoc queries and dashboards.

Beyond data movement and ETL, most #ML centric jobs (e.g. model training and execution) run in a similarly elastic environment as containers running Python and R code on Amazon EC2 Container Service clusters. The execution of batch jobs on top of ECS is managed by Flotilla, a service we built in house and open sourced (see https://github.com/stitchfix/flotilla-os).

At Stitch Fix, algorithmic integrations are pervasive across the business. We have dozens of data products actively integrated systems. That requires serving layer that is robust, agile, flexible, and allows for self-service. Models produced on Flotilla are packaged for deployment in production using Khan, another framework we've developed internally. Khan provides our data scientists the ability to quickly productionize those models they've developed with open source frameworks in Python 3 (e.g. PyTorch, sklearn), by automatically packaging them as Docker containers and deploying to Amazon ECS. This provides our data scientist a one-click method of getting from their algorithms to production. We then integrate those deployments into a service mesh, which allows us to A/B test various implementations in our product.

For more info:

• Our Algorithms Tour: https://algorithms-tour.stitchfix.com/
• Our blog: https://multithreaded.stitchfix.com/blog/
• Careers: https://multithreaded.stitchfix.com/careers/

#DataScience #DataStack #Data

As a frontend engineer on the Algorithms & Analytics team at Stitch Fix, I work with data scientists to develop applications and visualizations to help our internal business partners make data-driven decisions. I envisioned a platform that would assist data scientists in the data exploration process, allowing them to visually explore and rapidly iterate through their assumptions, then share their insights with others. This would align with our team's philosophy of having engineers "deploy platforms, services, abstractions, and frameworks that allow the data scientists to conceive of, develop, and deploy their ideas with autonomy", and solve the pain of data exploration.

The final product, code-named Dora, is built with React, Redux.js and Victory, backed by Elasticsearch to enable fast and iterative data exploration, and uses Apache Spark to move data from our Amazon S3 data warehouse into the Elasticsearch cluster.

I have recently started using Confluent/Kafka cloud. We want to do some stream processing. As I was going through Kafka I came across Kafka Streams and KSQL. Both seem to be A good fit for stream processing. But I could not understand which one should be used and one has any advantage over another. We will be using Confluent/Kafka Managed Cloud Instance. In near future, our Producers and Consumers are running on premise and we will be interacting with Confluent Cloud.

Also, Confluent Cloud Kafka has a primitive interface; is there any better UI interface to manage Kafka Cloud Cluster?

We currently have 2 Kafka Streams topics that have records coming in continuously. We're looking into joining the 2 streams based on a key with a window of 5 minutes based on their timestamp.

Should I consider kStream - kStream join or Apache Flink window joins? Or is there any other better way to achieve this?

Lumosity is home to the world's largest cognitive training database, a responsibility we take seriously. For most of the company's history, our analysis of user behavior and training data has been powered by an event stream--first a simple Node.js pub/sub app, then a heavyweight Ruby app with stronger durability. Both supported decent throughput and latency, but they lacked some major features supported by existing open-source alternatives: replaying existing messages (also lacking in most message queue-based solutions), scaling out many different readers for the same stream, the ability to leverage existing solutions for reading and writing, and possibly most importantly: the ability to hire someone externally who already had expertise.

We ultimately migrated to Kafka in early- to mid-2016, citing both industry trends in companies we'd talked to with similar durability and throughput needs, the extremely strong documentation and community. We pored over Kyle Kingsbury's Jepsen post (https://aphyr.com/posts/293-jepsen-Kafka), as well as Jay Kreps' follow-up (http://blog.empathybox.com/post/62279088548/a-few-notes-on-kafka-and-jepsen), talked at length with Confluent folks and community members, and still wound up running parallel systems for quite a long time, but ultimately, we've been very, very happy. Understanding the internals and proper levers takes some commitment, but it's taken very little maintenance once configured. Since then, the Confluent Platform community has grown and grown; we've gone from doing most development using custom Scala consumers and producers to being 60/40 Kafka Streams/Connects.

We originally looked into Storm / Heron , and we'd moved on from Redis pub/sub. Heron looks great, but we already had a programming model across services that was more akin to consuming a message consumers than required a topology of bolts, etc. Heron also had just come out while we were starting to migrate things, and the community momentum and direction of Kafka felt more substantial than the older Storm. If we were to start the process over again today, we might check out Pulsar , although the ecosystem is much younger.

To find out more, read our 2017 engineering blog post about the migration!

I need to build the Alert & Notification framework with the use of a scheduled program. We will analyze the events from the database table and filter events that are falling under a day timespan and send these event messages over email. Currently, we are using Kafka Pub/Sub for messaging. The customer wants us to move on Apache Flink, I am trying to understand how Apache Flink could be fit better for us.

I have to build a data processing application with an Apache Beam stack and Apache Flink runner on an Amazon EMR cluster. I saw some instability with the process and EMR clusters that keep going down. Here, the Apache Beam application gets inputs from Kafka and sends the accumulative data streams to another Kafka topic. Any advice on how to make the process more stable?

My background is in Data analytics in the telecom domain. Have to build the database for analyzing large volumes of CDR data so far the data are maintained in a file server and the application queries data from the files. It's consuming a lot of resources queries are taking time so now I am asked to come up with the approach. I planned to rewrite the app, so which database needs to be used. I am confused between MongoDB and Druid.

So please do advise me on picking from these two and why?

My process is like this: I would get data once a month, either from Google BigQuery or as parquet files from Azure Blob Storage. I have a script that does some cleaning and then stores the result as partitioned parquet files because the following process cannot handle loading all data to memory.

The next process is making a heavy computation in a parallel fashion (per partition), and storing 3 intermediate versions as parquet files: two used for statistics, and the third will be filtered and create the final files.

I make a report based on the two files in Jupyter notebook and convert it to HTML.

• Everything is done with vanilla python and Pandas.
• sometimes I may get a different format of data
• cloud service is Microsoft Azure.

What I'm considering is the following:

Get the data with Kafka or with native python, do the first processing, and store data in Druid, the second processing will be done with Apache Spark getting data from apache druid.

the intermediate states can be stored in druid too. and visualization would be with apache superset.

To provide employees with the critical need of interactive querying, we’ve worked with Presto, an open-source distributed SQL query engine, over the years. Operating Presto at Pinterest’s scale has involved resolving quite a few challenges like, supporting deeply nested and huge thrift schemas, slow/ bad worker detection and remediation, auto-scaling cluster, graceful cluster shutdown and impersonation support for ldap authenticator.

Our infrastructure is built on top of Amazon EC2 and we leverage Amazon S3 for storing our data. This separates compute and storage layers, and allows multiple compute clusters to share the S3 data.

We have hundreds of petabytes of data and tens of thousands of Apache Hive tables. Our Presto clusters are comprised of a fleet of 450 r4.8xl EC2 instances. Presto clusters together have over 100 TBs of memory and 14K vcpu cores. Within Pinterest, we have close to more than 1,000 monthly active users (out of total 1,600+ Pinterest employees) using Presto, who run about 400K queries on these clusters per month.

Each query submitted to Presto cluster is logged to a Kafka topic via Singer. Singer is a logging agent built at Pinterest and we talked about it in a previous post. Each query is logged when it is submitted and when it finishes. When a Presto cluster crashes, we will have query submitted events without corresponding query finished events. These events enable us to capture the effect of cluster crashes over time.

Each Presto cluster at Pinterest has workers on a mix of dedicated AWS EC2 instances and Kubernetes pods. Kubernetes platform provides us with the capability to add and remove workers from a Presto cluster very quickly. The best-case latency on bringing up a new worker on Kubernetes is less than a minute. However, when the Kubernetes cluster itself is out of resources and needs to scale up, it can take up to ten minutes. Some other advantages of deploying on Kubernetes platform is that our Presto deployment becomes agnostic of cloud vendor, instance types, OS, etc.

#BigData #AWS #DataScience #DataEngineering

The algorithms and data infrastructure at Stitch Fix is housed in #AWS. Data acquisition is split between events flowing through Kafka, and periodic snapshots of PostgreSQL DBs. We store data in an Amazon S3 based data warehouse. Apache Spark on Yarn is our tool of choice for data movement and #ETL. Because our storage layer (s3) is decoupled from our processing layer, we are able to scale our compute environment very elastically. We have several semi-permanent, autoscaling Yarn clusters running to serve our data processing needs. While the bulk of our compute infrastructure is dedicated to algorithmic processing, we also implemented Presto for adhoc queries and dashboards.

Beyond data movement and ETL, most #ML centric jobs (e.g. model training and execution) run in a similarly elastic environment as containers running Python and R code on Amazon EC2 Container Service clusters. The execution of batch jobs on top of ECS is managed by Flotilla, a service we built in house and open sourced (see https://github.com/stitchfix/flotilla-os).

At Stitch Fix, algorithmic integrations are pervasive across the business. We have dozens of data products actively integrated systems. That requires serving layer that is robust, agile, flexible, and allows for self-service. Models produced on Flotilla are packaged for deployment in production using Khan, another framework we've developed internally. Khan provides our data scientists the ability to quickly productionize those models they've developed with open source frameworks in Python 3 (e.g. PyTorch, sklearn), by automatically packaging them as Docker containers and deploying to Amazon ECS. This provides our data scientist a one-click method of getting from their algorithms to production. We then integrate those deployments into a service mesh, which allows us to A/B test various implementations in our product.

For more info:

• Our Algorithms Tour: https://algorithms-tour.stitchfix.com/
• Our blog: https://multithreaded.stitchfix.com/blog/
• Careers: https://multithreaded.stitchfix.com/careers/

#DataScience #DataStack #Data

When I joined NYT there was already broad dissatisfaction with the LAMP (Linux Apache HTTP Server MySQL PHP) Stack and the front end framework, in particular. So, I wasn't passing judgment on it. I mean, LAMP's fine, you can do good work in LAMP. It's a little dated at this point, but it's not ... I didn't want to rip it out for its own sake, but everyone else was like, "We don't like this, it's really inflexible." And I remember from being outside the company when that was called MIT FIVE when it had launched. And been observing it from the outside, and I was like, you guys took so long to do that and you did it so carefully, and yet you're not happy with your decisions. Why is that? That was more the impetus. If we're going to do this again, how are we going to do it in a way that we're gonna get a better result?

So we're moving quickly away from LAMP, I would say. So, right now, the new front end is React based and using Apollo. And we've been in a long, protracted, gradual rollout of the core experiences.

React is now talking to GraphQL as a primary API. There's a Node.js back end, to the front end, which is mainly for server-side rendering, as well.

Behind there, the main repository for the GraphQL server is a big table repository, that we call Bodega because it's a convenience store. And that reads off of a Kafka pipeline.

We've been using PostgreSQL since the very early days of Zulip, but we actually didn't use it from the beginning. Zulip started out as a MySQL project back in 2012, because we'd heard it was a good choice for a startup with a wide community. However, we found that even though we were using the Django ORM for most of our database access, we spent a lot of time fighting with MySQL. Issues ranged from bad collation defaults, to bad query plans which required a lot of manual query tweaks.

We ended up getting so frustrated that we tried out PostgresQL, and the results were fantastic. We didn't have to do any real customization (just some tuning settings for how big a server we had), and all of our most important queries were faster out of the box. As a result, we were able to delete a bunch of custom queries escaping the ORM that we'd written to make the MySQL query planner happy (because postgres just did the right thing automatically).

And then after that, we've just gotten a ton of value out of postgres. We use its excellent built-in full-text search, which has helped us avoid needing to bring in a tool like Elasticsearch, and we've really enjoyed features like its partial indexes, which saved us a lot of work adding unnecessary extra tables to get good performance for things like our "unread messages" and "starred messages" indexes.

I can't recommend it highly enough.