Press "Enter" to skip to content

Day: November 23, 2018

Using Kafka To Drive ML Predictions

Published 2018-11-23 by Kevin Feasel

Kai Waehner shows us a model architecture for using Apache Kafka to generate predictions from trained models:

Kafka applications are event based, and leverage stream processing to continuously process input data. If you’re using Kafka, then you can embed an analytic model natively in a Kafka Streams or KSQLapplication. There are various examples of Kafka Streams microservices embedding models built with TensorFlow, H2O or Deeplearning4j natively.

It is not always possible or feasible to embed analytic models directly due to architectural, security or organizational reasons. You can also choose to use RPC to perform model inference from your Kafka application (bearing in mind the the pros and cons discussed above). You can visit my project for an example of gRPC integration between a Kafka Streams microservice and locally hosted TensorFlow Serving container for making predictions with a hosted TensorFlow model.

There are a couple separate and interesting patterns here.

Comments closed

Strategies For Dealing With Failed Projects

Published 2018-11-23 by Kevin Feasel

Edwin Thoen gives us a few tips for dealing with failing data science projects:

At the beginning of a project the levels enthusiasm and optimism are always at its peak. Especially in data science projects. Isn’t data the new oil? This is the time we are finally going to dig into that well and leverage our data in unprecedented ways! No setbacks are experienced yet. There is only one road ahead and it will lead us to success. Probably at this stage you, the data scientist, are already well aware of a number of project risks. You might want to keep these concerns to yourself, as you don’t want to come across as negative, or worse, someone who is not up to the job ahead. Please don’t, if you foresee possible problems at this stage and you don’t speak out, they can come back as a boomerang when the problems actually occur. Rather, invite all stakeholders to perform a risk analysis together.

This is good advice and applies outside of data science projects as well.  H/T R-bloggers

Comments closed

Kafka Analytics Patterns In HDP 3.1

Published 2018-11-23 by Kevin Feasel

George Vetticaden walks us through what’s coming with Apache Kafka in Hortonworks Data Platform 3.1:

A summary of these three new access patterns:

  • Stream Processing: Kafka Streams Support – With existing support for Spark Streaming, SAM/Storm, Kafka Streams addition provides developers with more options for their stream processing and microservice needs.

  • SQL Analytics: New Hive Kafka Storage Handler – View Kafka topics as tables and execute SQL via Hive with full SQL Support for joins, windowing, aggregations, etc.

  • OLAP Analytics: New Druid Kafka Indexing Service – View Kafka topics as cubes and perform OLAP style analytics on streaming events in Kafka using Druid.

Click through for high-level explanations of each.  George promises more detailed explanations as well.

Comments closed

DATEDIFF

Published 2018-11-23 by Kevin Feasel

Randolph West continues a series on covering dates and times, looking at DATEDIFF and DATEDIFF_BIG:

The only functional difference between them is that the DATEDIFF_BIG() returns values as a BIGINT, for results that exceed the boundary of an INT. Keep this in mind when deciding which one to use. For example, the maximum number of seconds an INT can hold is 68 years, while a BIGINT can comfortably store the number of seconds in 10,000 years. This becomes especially important when dealing with microseconds and nanoseconds.

The rest of the post will use DATEDIFF() to refer to both functions.

I think this might be the first time I’d read about DATEDIFF_BIG()and I’m not aware of ever having used it.  But hey, it could make sense if you need to track more than 2 billion microseconds.

Comments closed

Optimizing SQL Server Workloads On VMware

Published 2018-11-23 by Kevin Feasel

Jeff Mlakar shares a few tips on hosting SQL Server via VMware virtual machine:

Why is Over-Allocating a VM Bad?

Why is is so important to allocate just the right amount of resources for your guest VM running SQL Server? We all know the problem with an undersized system; however, what about an oversized system? What problem could that be? Here are a few common issues to consider:

  • Over-allocating CPU causes poor resource utilization across all the guest VMs

  • Over-allocating memory unnecessarily increases memory contention and overhead on other guest VMs

  • Having more vCPUs assigned to the VM can have an impact on licensing.

There are some good points in the post, so check it out.

Comments closed

The Risk Of Shadow IT In The Cloud

Published 2018-11-23 by Kevin Feasel

Kenneth Fisher walks us through the risk of increased Shadow IT with migration to the cloud:

Shadow IT has been, well, maybe not the bane of the IT department, but certainly a pain in the neck. On the off chance you’ve never heard of shadow IT do any of these sound familiar?

  • A user asks you to restore a corrupt database on a SQL Server you’ve never heard of and isn’t in your inventory. (And 50/50 odds there’s never been a backup taken.)

  • You do a licensing true-up and dozens of new SQL Servers suddenly show up.

  • You hear from a user: “We have this mission critical Access database that suddenly isn’t working. I know you don’t support access but you’re the database person so we need you to fix it.”

It’s an interesting short essay and worth thinking about if you’re in the cloud or moving that way.

Comments closed

Casting Constants And POWER()

Published 2018-11-23 by Kevin Feasel

Steve Jones walks us through a case of an unexpected error:

I ran into an interesting problem while working with the POWER() function. I was trying to do some binary conversions and had a statement like this to process powers of 2.

SELECT POWER(2, n)

This was designed to take a value and return a power of 2. I then used a different value to determine if this was added to my conversion factor or not. In trying to work with some larger numbers, I ran into this error:

Msg 232, Level 16, State 3, Line 3
Arithmetic overflow error for type int, value = 2147483648.000000.

Click through for a description of the problem as well as Steve’s solution.

Comments closed

Extended Events In Azure Data Studio

Published 2018-11-23 by Kevin Feasel

Jason Brimhall walks us through installation of the SQL Server Profiler on Azure Data Studio:

Azure Data Studio (ADS) is getting all sorts of love and attention these days. So much so that they have finally gotten around to adding Extended Events (XE) to the tool – sort of. Now we have the power to run traces on SQL Server via ADS.

The presence of XE in ADS comes via an extension and comes with a few other caveats. I will explore the extension for XE available in ADS in this article and discuss some of the caveats. As you read the article, it might be helpful to go ahead and download ADS if you do not already have it.

Jason points out the name of SQL Server Profiler and I’d like to add my own bit of irritation here.  “Don’t use Profiler, except the one good Profiler but not the Profiler you think you’re using unless you know not to use Profiler and use Profiler instead.”  Yeah, that’s pretty clear.

Comments closed

dbatools: On The Way To 1.0

Published 2018-11-23 by Kevin Feasel

Chrissy LeMaire points out a bunch of changes to dbatools:

Aliases have been added for the changes, so these are not breaking changes:

  • Mismatched Copy commands have been renamed to match their corresponding Get command names (ie. Copy-DbaCentralManagementServer is now Copy-DbaCmsRegServer).

  • Most parameters named Password have been changed to SecurePassword. They’ve always been a SecureStringdata type but this makes that clear.

  • The parameters ExcludeAllSystemDb and ExcludeAllUserDb have been changed to ExcludeSystem and ExcludeUser, respectively.

These are some of the non-breaking changes, but this latest release has several breaking changes too.  Chrissy is promising no more breaking changes for a little while, so it’s probably a good time to upgrade and check those scripts to see what you need to change.

Comments closed