Online Learning Algorithms

Xin Hunt describes the benefits of online learning algorithms:

A few examples of classical online learning algorithms include recursive least squares, stochastic gradient descent and multi-armed bandit algorithms like Thompson sampling. Many online algorithms (including recursive least squares and stochastic gradient descent) have offline versions. These online algorithms are usually developed after the offline version, and are designed for better scaling with large datasets and streaming data. Algorithms like Thompson sampling on the other hand, do not have offline counterparts, because the problems they solve are inherently online.

Let’s look at interactive ad recommendation systems as an example. You’ll find ads powered by these systems when you browse popular publications, weather sites and social media networks. These recommendation systems build customer preference models by tracking your shopping and browsing activities (ad clicking, wish list updates and purchases, for example). Due to the transient nature of shopping behaviors, new recommendations must reflect the most recent activities. This makes online learning a natural choice for these systems.

My favorite online learning algorithm at the moment is Online Passive-Aggressive Algorithms.  Not just because that name describes my Twitter feed.

Optimizing Apache Flink

Ivan Mushketyk has a few tips for speeding up programs using Apache Flink:

One more way to optimize your Flink application is to provide some information about what your user-defined functions are doing with input data. Since Flink can’t parse and understand code, you can provide crucial information that will help to build a more efficient execution plan. There are three annotations that we can use:

  1. @ForwardedFields: Specifies what fields in an input value were left unchanged and are used in an output value.

  2. @NotForwardedFields: Specifies fields that were not preserved in the same positions in the output.

  3. @ReadFields: Specifies what fields were used to compute a result value. You should only specify fields that were used in computations and not merely copied to the output.

Click through for his four tips.

Installing The Azure ML Workbench

Leila Etaati walks us through setting up the Azure ML workbench:

In Microsoft ignite 2017, Azure ML team announce new on-premises tools for doing machine learning. this tools much more comprehensive as it provides

1- a workspace helps data wrangling

2- Data Visualization

3-Easy to deploy

4-Support Python codes

in this post and next posts, I will share my experiment with working this tools.

Click through for the step-by-step installation guide.

When The Maximum Workspace Memory Isn’t The Internal Pool Maximum

Lonny Niederstadt answers the call from someone who needs the combination of Perfmon and DMV data:

When is a maximum not really the maximum?
When it’s a maximum for an explicitly or implicitly modified default.
Whether “the definitive documentation” says so or not.

Yesterday on Twitter #sqlhelp this question came up.



Aha! I thought to myself.  For this I am purposed! To show how Perfmon and DMV data tie out!

Read on for the simple form of the answer, followed by the complication which makes life interesting.

Checking If UseLargeFRS Is Enabled

Cody Konior has a quick post showing how to see if you have the UseLargeFRS switch turned on for your SQL Server disks:

PureStorage has a pretty cool post that mentions the importance of formatting SQL Server disks with a 64KB clusters and the /L flag (also known as the UseLargeFRS switch on PowerShell’s Format-Volume cmdlet).

Why UseLargeFRS? It’s to help avoid DBCC CHECKDB failures on large/busy databases. But how do you work out whether it’s enabled or not? PowerShell to the rescue!

You can’t work this out remotely but you can do it locally.

Read on for the script.

Tracking Replication Throughput

Phil Grayson has a couple of replication-related queries:

It’s a right laugh when replication goes wrong, if you haven’t been there and use replication heavily, you’re in for the night / early morning of your life. Like any issue, the cause of replication problems could be one of many, but checking what commands are being passed through the distributor is always worthwhile, as an inefficient query could be updating the same data over and over, putting unnecessary strain on the distributor, infrastructure or the subscribers.

Read on for those queries.

XESmartTarget: Event Handling With SQL Server

Gianluca Sartori introduces a new tool to help with event handling in SQL Server:

I am proud to introduce XESmartTarget: the easiest way to interact with extended events sessions without writing a single line of code.

XESmartTarget uses the Extended Events streaming API to connect to a session and execute some actions in response to each event. The actions you perform can be configured in a simple .json file, which controls the Response types and all their properties.

For instance, if you have a session to capture deadlocks, you may want to be notified whenever such an event is captured. All you need to do is configure XESmartTarget to send an email using the EmailResponse class. You can do that by creating a simple .json file with all the parameters needed

This looks quite interesting.

Replacing DAX PathContains With OR

Chris Koester shows the performance benefits of replacing the PathContains function in DAX with a simple OR operator:

This post shows how you can generate optimized multi-value DAX parameters in SSRS and achieve greater performance compared to the DAX PathContains function. This will be a short post that provides the SSRS expression to convert multiple SSRS parameters into a double-pipe delimited string for use in a DAX query. In other words, the goal is to use the DAX OR operator (||) instead of the PathContains function. I’m assuming the reader has experience with SSRS, so not all steps will be shown.

Read on for the example, which ended up being a 16X performance improvement.

Detecting RAID Failures

Randolph West has an interesting after-the-fact test to determine whether data on a RAID array which experienced major failure is recoverable:

We took a look through some random files on the disk. I was looking for evidence that the VMDK file that had been copied was taken from a RAID array in a corrupt state, namely that one of the disks had failed, and that the second disk had failed during the rebuild of the array.

The easiest way to see this is to look for a JPEG or PNG file over 256 KB in size. Most RAID block sizes are multiples of 64 KB, usually 128 KB or 256 KB. Each block is split over the individual physical disks, with a parity bit, so for a particular block of data, if the RAID array has failed, you will see a corrupt image, or the image won’t be viewable at all.

Randolph presents an interesting smoke test here.  Read the whole thing.

Installing dbatools

Chrissy LeMaire has a good walkthrough on how to install a module from the Powershell gallery, using dbatools as an example:

It’s easy for PowerShell toolmakers to forget that new users may have questions about ExecutionPolicy and Repositories. If dbatools is the first module you’ve ever installed and used, this guide is intended to help you with all of your install/setup questions and concerns.

In order to emulate what your experience may be like, I spun up a fresh Windows 10 instance in Azure and went through all of the required steps which include addressing:

  • Setting the Execution Policy

  • Explicitly trusting Microsoft’s PowerShell Gallery repository

  • Installing dbatools

  • Explicitly trusting dbatools as a Publisher, before first use

If you’ve never installed dbatools before, Chrissy shows just how easy it can be.


October 2017
« Sep Nov »