I have so many thoughts and ideas about where my passion will lead me next. I haven’t yet settled on any one thing for a new career, so I went back to the basics. Listen. Listen to my community. I envision my community as marginalized people in tech. So I have started a meetup group where we can get together and talk. Where we can listen to each other. Where we help each other. Join me and let’s go on this journey to our futures together.
A gathering place for people forging new paths after harassment at work.
This is a safe space – no hate speech, bullying, harassment, or discrimination is tolerated. We value input from a variety of identities and will center the views, needs, and decisions of those who are not cishet white men.
I’m a 50 year old white woman leaving the tech world. As I talk about the harassment, bullying, and discrimination I’ve faced over the years other women open up about their own experiences. So many of us have no place to talk to others with the same experiences. Let’s share our stories, our growth, our pain and joy. This is a place to talk about surviving and thriving, about careers, family, friends, life, work, play, and about disrupting the white patriarchy to nurture a new way of doing things.
#Words4Justice
Befriending Dragons – Life After Workplace Harassment
Bellevue, WA 3Members
A gathering place for people forging new paths after harassment at work.This is a safe space – no hate speech, bullying, harassment, or discrimination is tolerated. We value …
I no longer feel safe, comfortable, or valued working in tech. Going forward I’ll be working to actively disrupt tech culture and systems to reduce harassment and discrimination. Keep an eye on #Words4Justice. đ
Be kind. Be brave. Go beyond ally to accomplice to actively disrupt bullying and discrimination.
After all these years soaring through the data world, from SQL Server 1.11 all the way through todayâs modern Big Data technologies, I am making a flight adjustment. My next adventure will be in the land of the Windows Hypervisor: Hyper-V. Last week I started working with my new team and I am already learning to corral and wield a whole new world of acronyms, technologies, and scenarios. As a software engineer on the quality team Iâll help define and implement test scenarios that lead to better customer experiences across multiple products.
I won’t be leaving data behind! This new role has a lot of data aspects and of course the hypervisor underlies many of the world’s data systems! It’s been great working with the #SQLFamily over the years and I look forward to continuing to work with you all!
Weâre good people. As good people we donât want to think we do things that have negative consequences for others. But sometimes our subconscious can fool us. What we intend isnât always what happens. We think weâre making a totally rational decision based on our conscious values â but subtle, unconscious bias creeps in. Yes, even for good people. For 20+ years folks at Harvard have been using something called the Implicit Association Test (IAT) to help us identify our biases.
The human brain works with patterns and categories. Itâs how we make it through the day. We are bombarded with 100s of thousands of data points every day â we canât possibly think through each one every time. We unconsciously assign data points, including our perception of people, into buckets. Those buckets have values and characteristics assigned to them that may or may not reflect the individual person we put in that bucket.
This automatic assignment is called intuitive thinking or system 1 thinking. Itâs easy and takes little effort. It serves us well and lets us take on many tasks every day. However, it also sometimes leads us down the path of thinking weâve chosen the âbestâ person when weâre really hired someone who meets some set of assumptions.
Sometimes we use slow thinking, or system 2 thinking. Itâs rarely a conscious decision, something just makes us take some extra time and we usually donât even realize it. Thatâs when we stop to question what weâre doing â maybe we adjust which categories we put someone in or we adjust the category or the values and judgments associated with it. Weâre good people but system 2 thinking is tiring and we just canât do it all the time.
Diversity Matters
Why does diversity matter at work? Personally, when weâre on a diverse team we tend to have higher personal and job satisfaction. Diverse teams are interesting and we often learn more. People who donât feel like theyâre the âonly oneâ of something (gender, sexual orientation, race, introvert/extravert, etc.) relax, contribute more, and are more productive. And study after study shows that more diverse teams lead to better products and a better bottom line.
Companies with women on their boards have higher ROIs, more diverse companies tend to perform above average, and letâs face it â we donât have enough STEM graduates to fill needed jobs if we donât encourage a more diverse group of people to enter the field.
Mind Tricks
But weâre good people and we donât make these snap judgments. We are rational and we always know why we made a decision. Or do we?
Optical illusions fool us all the time. Even knowing those lines are all the same length, did you have to measure them just to be sure? The same thing happens in our interactions with people. Whatâs the first thing that comes to mind for single parent, introvert, doctor, CEO, or programmer? That first thing hints at your categories â the categories built up by a lifetime of media saturation filled with type-cast actors.
Back to the science of bias. Letâs think about resumes. In one study, resumes were handed out to academics who were asked to rate the job candidates for competency, hireability, mentoring, and suggested salary. Some resumes were for John and some for Jennifer. Professors of all genders rated Jennifer 25% less competent and less likely to be hired. They rated John worth about $4000 more. When asked why they gave ratings their justifications sounded rational butâŚ. 4 industry publications was awesome for John and 4 was just not enough for Jennifer. They are good people but they (we!) are at the mercy of their subconscious and years of societal conditioning.
Moving On
Weâre good people so what do we do?
Take the IATs â there are many, take at least a couple and understand your unexpected biases. Talk about this with others so we all become comfortable talking about our subtle biases. Work to consciously update your mental categories â seek out images and reminders of people who are different and successful. Now that you know your own categories a bit better, be more mindful about switching to system 2 thinking. Reach out to one person and mentor them. Spend time with someone who makes you uncomfortable. Pay attention to the âfirstsâ (the first autistic character on Sesame Street, the first black President, the first whatever) and see if that helps you update your mental categories.
Increase the pipeline. Participate in groups that help kids learn to code. Recruit beyond your normal network, post jobs on diversity sites, and consider non-traditional backgrounds. Join diverse groups that donât match your own diversity.
Be careful with words. Is someone bossy or exhibiting leadership? Is someone aggressive or a go-getter? Are they emotional or passionate. You may be surprised how you assign different words for the same behavior in unexpected ways.
When you post a job, only list something as ârequiredâ if it truly is. Women for example tend to only apply if they meet almost all the requirements, men tend to apply if they meet a few. Do you really require Java experience or do you need a good coder who is willing to learn new things? Donât ask for a specific type of leader, look for someone who can lead in any of many productive ways. Explicitly state that you value a diverse team. And beware of subtle stereotypes â words like best, rock star, action-oriented define a particular picture but may not represent what youâre really looking for.
When reviewing resumes, have HR take off names, cities, and years. Before you pick up a resume decide on your priorities â does experience or willingness to learn matter more for example? Look for people who fill gaps rather than trying to replicate people you already have. And remember, system 2 thinking is tiring so do this when youâre alert and can take the time to think about what youâre doing.
For the interviews, have a diverse group participate. Simply looking at picture of or talking about diverse people before starting interviews increases the chance you hire with diversity in mind. Donât confuse either confidence or âgeek credâ with competence. Keep an open mind about different ways of approaching problems â itâs the result that matters.
Many flowers make a beautiful bouquet – @IsisAnchalee
SQL Server is a great technology â Iâve been using it since 1993 when the user interface consisted of a query window with the options to save and execute and not much else. With every release thereâs something new and exciting and thereâs always something to learn about even the most familiar of features. However, not everyone uses SQL Server for every storage and compute opportunity â sad but true.
So what is a SQL geek to do in the face of all the new options out there â many under the umbrella of Big Data (distributed processing)? Why just jump right on in and learn it! No one can know all the pieces because itâs a big, fluid, messy collection of âthingsâ. But donât worry about that, start with one thing and build from there. Even if you never plan to implement a production Big Data system you need to learn about it â because if you donât have some hands-on experience with it then someone who does have that experience will be influencing the decision makers without you. For a SQL Pro I suggest Hive as that easy entry point. At some point maybe Spark SQL will jump into that gap, but for now Hive is the easiest entry point for most SQL pros.
For more, I refer you to the talk I gave at the Pacific Northwest SQL Server User Group meeting on October 14, 2015. Excerpts are below, the file is attached.
Look, itâs SQL!
SELECT score, fun
FROM toDo
WHERE type = ‘they pay me for this?’;
Hereâs how that code looks from Visual Studio along with the links to how you find the output and logs:
And yet itâs more!
CREATE EXTERNAL TABLE IF NOT EXISTS toDo
(fun STRING,
rank INT COMMENT ‘rank the greatness’,
type STRING)
COMMENT ‘two tables walk into a bar….’
ROW FORMAT DELIMITED FIELDS TERMINATED BY ‘,’
STORED AS TEXTFILE
LOCATION ‘/data/demo/’;
A mix of old and new
— read some data
SELECT ‘you cannot make me ‘, score, fun, type
FROM toDo
WHERE score <= 0
ORDER BY score;
SELECT ‘when can we ‘, score, fun, type
FROM toDo
WHERE score > 0
DISTRIBUTE BY score SORT BY score;
Thatâs Hive folks!
Hive
on Hadoop
on HDInsight
on AzureBig Data in the cloud!
Data exploration, analytics and reporting, new data-driven actionable insights
Rapid iterating
Unknown unknowns
Flexible scaling
Data driven actions for early competitive advantage or first to market
Low number of direct, concurrent users
Low cost data archival
Hadoop Anti-PatternsâŚ.
Replace system whose pain points donât align with Hadoopâs strengths
OLTP needs adequately met by an existing system
Known data with a static schema
Many end users
Interactive response time requirements (becoming less true)
Your first Hadoop project + mission critical system
Azure has so much more
Go straight to the business code
Scale storage and compute separately
Open Source
Linux
Managed and unmanaged services
Hybrid
On-demand and 24×7 options
SQL Server
Itâs a Polyglot
Stream your data into a lake
Pick the best compute for each task
One of the frequently overlooked yet essential best practices for Hadoop is to prefer fewer, bigger files over more, smaller files. How small is too small and how many is too many? How do you stitch together all those small Internet of Things files into files “big enough” for Hadoop to process efficiently?
The Problem
One performance best practice for Hadoop is to have fewer large files as opposed to large numbers of small files. A related best practice is to not partition âtoo muchâ. Part of the reason for not over-partitioning is that it generally leads to larger numbers of smaller files.
Too small is smaller than HDFS block size (chunk size), or realistically small is something less than several times larger than chunk size. A very, very rough rule of thumb is files should be at least 1GB each and no more than maybe around 10,000-ish files per table. These numbers, especially the maximum total number of files per table, vary depending on many factors. However, it gives you a reference point. The 1GB is based on multiples of the chunk size while the 2nd is honestly a bit of a guess based on a typical small cluster.
Why Is It Important?
One reason for this recommendation is that Hadoopâs name node service keep track of all the files and where the internal chunks of the individual files are. The more files it has to track the more memory it needs on the head node and the longer it takes to build a job execution plan. The number and size of files also affects how memory is used on each node.
Letâs say your chunk size is 256MB. Thatâs the maximum size of each piece of the file that Hadoop will store per node. So if you have 10 nodes and a single 1GB file it would be split into 4 chunks of 256MB each and stored on 4 of those nodes (Iâm ignoring the replication factor for this discussion). If you have 1000 files that are 1MB each (still a total data size of ~1GB) then every one of those files is a separate chunk and 1000 chunks are spread across those 10 nodes. NOTE: In Azure and WASB this happens somewhat differently behind the scenes â the data isnât physically chunked up when initially stored but rather chunked up at the time a job runs.
With the single 1GB file the name node has 5 things to keep track of â the logical file plus the 4 physical chunks and their associated physical locations. With 1000 smaller files the name node has to track the logical file plus 1000 physical chunks and their physical locations. That uses more memory and results in more work when the head node service uses the file location information to build out the plan for how it will split out any Hadoop job into tasks across the many nodes. When weâre talking about systems that often have TBs or PBs of data the difference between small and large files can add up quickly.
The other problem comes at the time that the data is read by a Hadoop job. When the job runs on each node it loads the files the task tracker identified for it to work with into memory on that local node (in WASB the chunking is done at this point). When there are more files to be read for the same amount of data it results in more work and slower execution time for each task within each job. Sometimes you will see hard errors when operating system limits are hit related to the number of open files. There is also more internal work involved in reading the larger number of files and combining the data.
Stitching
There are several options for stitching files together.
Combine the files as they land using the code that moves the files. This is the most performant and efficient method in most cases.
INSERT into new Hive tables (directories) which creates larger files under the covers. The output file size can be controlled with settings like hive.merge.smallfiles.avgsize and hive.merge.size.per.task.
Use a combiner in Pig to load the many small files into bigger splits.
Dealing with Hadoop’s small files problem http://snowplowanalytics.com/blog/2013/05/30/dealing-with-hadoops-small-files-problem/Â âaggregating with the small files first reduced total processing time from 2 hours 57 minutes to just 9 minutes – of which 3 minutes was the aggregation, and 4 minutes was running our actual Enrichment process. Thatâs a speedup of 1,867%.â
The Small Files Problem http://blog.cloudera.com/blog/2009/02/the-small-files-problem/ âReading through files in a HAR is no more efficient than reading through files in HDFS, and in fact may be slower since each HAR file access requires two index file reads as well as the data file read (see diagram). And although HAR files can be used as input to MapReduce, there is no special magic that allows maps to operate over all the files in the HAR co-resident on a HDFS block.â
The key here is to work with fewer, larger files as much as possible in Hadoop. The exact steps to get there will vary depending on your specific scenario.
You can use the new Azure portal to create or edit Azure Data Factory components. Once you are done you may automate the process of creating future Data Factory components from PowerShell. In that case you can use the JSON files you edited in the portal GUI as configuration files for the PowerShell cmdlets. For example, you may try to create a new linked service using settings from C:CoolerHDInsight.JSON as specified below:
In that case you may see something like this error:
New-AzureDataFactoryLinkedService : Hub: {SomeName_hub} not found.
CategoryInfo               : CloseError: (:) [New-AzureDataFactoryLinkedService]. Provisioning FailedException
FullyQualifiedErrorIDÂ Â : Microsoft.Azure.Commands.DataFactories.NewAzureDataFactoryLinkedServiceCommand
If you check the JSON file that you exported from the portal and referenced in the PowerShell script, you will see it ends with something like this:
       “isPaused”: false,
“hubName”: “SomeName_hub”
}
}
The hubName is currently automatically generated based on the name of the Data Factory and should not be present in the JSON files used by PowerShell. Remove the comma on the line above the hubName and the entire line starting with hubName.
      ,
“hubName”: “SomeName_hub”
That will leave the end of the file looking something like this:
       “isPaused”: false
}
}
Check out all your other JSON files you are using for Data Factory components and do the same editing for any that have a hubName.
NOTE: This applies to Azure Data Factory as of April 2015. At some point the hubName should become a viable parameter usable by PowerShell.
a whole new world of acronyms, technologies, and scenarios. As a software engineer on the quality team Iâll help define and implement test scenarios that lead to better customer experiences across multiple products.
