Looking for some advice on how best to architect/design and build our pipeline.
After some initial testing, we're not getting the results that we were expecting. Maybe we're just doing something stupid, or our expectations are too high.
Our data/workflow:
Google DFP writes our adserver logs (CSV compressed) directly to GCS (hourly).
A day's worth of these logs has in the region of 30-70 million records, and about 1.5-2 billion for the month.
Perform transformation on 2 of the fields, and write the row to BigQuery.
The transformation involves performing 3 REGEX operations (due to increase to 50 operations) on 2 of the fields, which produces new fields/columns.
What we've got running so far:
Built a pipeline that reads the files from GCS for a day (31.3m), and uses a ParDo to perform the transformation (we thought we'd start with just a day, but our requirements are to process months & years too).
DoFn input is a String, and its output is a BigQuery TableRow.
The pipeline is executed in the cloud with instance type "n1-standard-1" (1vCPU), as we think 1 vCPU per worker is adequate given that the transformation is not overly complex, nor CPU intensive i.e. just a mapping of Strings to Strings.
We've run the job using a few different worker configurations to see how it performs:
5 workers (5 vCPUs) took ~17 mins
5 workers (10 vCPUs) took ~16 mins (in this run we bumped up the instance to "n1-standard-2" to get double the cores to see if it improved performance)
50 min and 100 max workers with autoscale set to "BASIC" (50-100 vCPUs) took ~13 mins
100 min and 150 max workers with autoscale set to "BASIC" (100-150 vCPUs) took ~14 mins
Would those times be in line with what you would expect for our use case and pipeline?
You can also write the output to files and then load it into BigQuery using command line/console. You'd probably save some dollars of instance's uptime. This is what I've been doing after running into issues with Dataflow/BigQuery interface. Also from my experience there is some overhead bringing instances up and tearing them down (could be 3-5 minutes). Do you include this time in your measurements as well?
BigQuery has a write limit of 100,000 rows per second per table OR 6M/per minute. At 31M rows of input that would take ~ 5 minutes of just flat out writes. When you add back the discrete processing time per element & then the synchronization time (read from GCS->dispatch->...) of the graph this looks about right.
We are working on a table sharding model so you can write across a set of tables and then use table wildcards within BigQuery to aggregate across the tables (common model for typical BigQuery streaming use case). I know the BigQuery folks are also looking at increased table streaming limits, but nothing official to share.
Net-net increasing instances is not going to get you much more throughput right now.
Another approach - in the mean time while we work on improving the BigQuery sync - would be to shard your reads using pattern matching via TextIO and then run X separate pipelines targeting X number of tables. Might be a fun experiment. :-)
Make sense?
Related
I am trying to optimize my Apache Beam pipeline on Google Cloud Platform Dataflow.
Background information: I am trying to read streaming data from PubSub Messages, and aggregate them based on 3 time windows: 1 min, 5 min and 60 min. Such aggregations consists of summing, averaging, finding the maximum or minimum, etc. For example, for all data collected from 1200 to 1201, I want to aggregate them and write the output into BigTable's 1-min column family. And for all data collected from 1200 to 1205, I want to similarly aggregate them and write the output into BigTable's 5-min column. Same goes for 60min.
The current approach I took is to have 3 separate dataflow jobs (i.e. 3 separate Beam Pipelines), each one having a different window duration (1min, 5min and 60min). See https://beam.apache.org/releases/javadoc/2.0.0/org/apache/beam/sdk/transforms/windowing/Window.html. And the outputs of all 3 dataflow jobs are written to the same BigTable, but on different column families. Other than that, the function and aggregations of the data are the same for the 3 jobs.
However, this seems to be very computationally inefficient, and cost inefficient, as the 3 jobs are essentially doing the same function, with the only exception being the window time duration and output column family.
Some challenges and limitations we faced was that from the Apache Beam documentation, it seems like we are unable to create multiple windows of different periods in a singular dataflow job. Also, when we write the final data into big table, we would have to define the table, column family, column, and rowkey. And unfortunately, the column family is a fixed property (i.e. it cannot be redefined or changed given the window period).
Hence, I am wondering if there is a way to only use 1 dataflow job (i.e. 1 Apache Beam pipeline) that fulfils the objective of this project? Which is to aggregate data on different window periods, and write them to different column families of the same BigTable.
I was considering using Split stream: first window by 1-min, then split into 3 streams (1 write to bigtable for 1-min interval, another for 5-min aggregation, and another for 60-min aggregation). However, the problem is that we are working with streaming data and not batch data.
Thank you
I have a Spark script that reads data from amazon S3 and then writes in another bucket usion parquet format.
This is what the code looks like:
File = "LocationInFirstBucket.csv.gz"
df_ods = spark.read.csv(File, header=True, sep=";")
df_ods.repartition(25).write.format("parquet").mode("OverWrite").save("AnotherLocationInS3")
My question is: how does the repartition argument (here 25) affects the execution time? Should I increase it so the script runs faster?
Second question: Would it be better if I cache my df before the last line?
Thank you
In typical setups neither repartition nor cache will help you in this specific case. Since you read data from non-splittable format:
File = "LocationInFirstBucket.csv.gz"
df_ods = spark.read.csv(File, header=True, sep=";")
df_ods will have only one partition.
In such case repartitioning would make sense, if you performed any actual processing on this data.
However if you just write to distributed file system repartitioning will simply double the cost - you have to send data to other nodes first (that involves serialization, deserialization, network transfer, write to disk) and then still write to distributed file system.
There are of course edge cases when this makes sense. If network connecting your cluster is much faster than network connection your cluster to S3 nodes, effective latency might be a bit lower.
As of caching ‒ there is no value in caching here at all. Caching Dataset is expensive, and makes sense only if persisted data is reused.
Answer 1 :- Repartition of 25 or more or less it depends on how much data you have and no. of executors you provided. If your Spark code run in the cluster with more than one executor and it is not repartitioned then repartitioning will speedy to writing parallel your data.
Answer 2 :- There is no need to cache df before the last line because you are using only single action in your code. If you will perform multiple actions on your DF and don't want it will recalculate as the number of actions then you will Cache it.
The thing here is that Spark can parallelize writing to a certain point since one file can't be written by multiple executors at the same time.
Repartition helps you in this parallelization because it will write 25 different files (one for each partition). If you increase the number of partitions you will increase the number of written files hence speeding up the execution. This comes with a price because of the reading time will increase with the number of files to be read.
The limit is the number of executors you are running your job with, e.g. if you are running with 25 executors then setting repartition to 26 will not help you because to write the 26th partition one of the previous 25 would have to be finished.
For the other question, I don't think .cache() will help you because Spark is lazy, maybe this article can help you further.
I noticed that running a SELECT count(*) FROM myTable on my larger BQ tables yields long running times, upwards of 30/40 seconds despite the validator claiming the query processes 0 bytes. This doesn't seem quite right when 500 GB queries run faster. Additionally, total row counts are listed under details -> Table Info. Am I doing something wrong? Is there a way to get total row counts instantly?
When you run a count BigQuery still needs to allocate resources (such as: slot units, shards etc). You might be reaching some limits which cause a delay. For example, the slots default per project is 2,000 units.
BigQuery execution plan provides very detail information about the process which can help you better understand the source of the delay.
One way to overcome this is to use an approximate method described in this link
This Slide by Google might also help you
For more details see this video about how to understand the execution plan
In Google Dataflow, i have a job that basically looks like this:
Dataset: 100 rows, 1 column.
Recipe: 0 steps
Output: New Table.
But it takes between 6-8 minutes to run. What could be the issue?
Usually times are in minutes, not in seconds for Dataprep/dataflow setup.
These solutions are for large data sets and the duration stays constant even if you have 10 times the size.
DataPrep creates for you a DataFlow workflow, and provisions a few VMs for you, that takes time, usually that phase could be in the minute mark. And only a bit later is scaling that up to 50 or 1000 boxes.
I just wrote my first hadoop job. It processes many files and generates multipleoutput files for each input file. I am running it on a two node cluster and it takes about 10 minutes for my largest input set. Looking at the counters below, what are the optimizations I can do to make it run faster? Are there any specific indicators which one should look for in these counters-
Version: 2.0.0-mr1-cdh4.1.2
Map task Capacity:20
Reduce task Capacity:20
Avg task per node:20
We can see here that most of data reduction happens in the map phase (number of map output bytes is much less then HDFS read bytes, The same about map input records - it is much lower then map input record). We also see that a lot of CPU time spent. We also see low number of shuffling bytes
So this job is:
a) A lot of data reduction is done on Map phase.
b) The job is CPU bound.
So I think code of mapper and reducer should be optimized. I/O probably is not important for this job.