I am pretty new to spark and would like some advise on how to approach the following problem.
I have candle data (high, low, open, close) for ever minute of a trading day spread across a year. This represents about 360,000 data points.
What I want to do is run some simulations across that data (and possibly every data point) and what I would like is for a given data point, get the previous (or next) x data points and then run some code across that to give a result.
Ideally, this would be in a map style function but you cannot do a nested operation in Spark. The only way that I can think about doing it is to create a DataSet of the Candle as a key and have the related data un-normalised or partitioning it on every key - either way seems inefficient.
Ideally I am looking for something that does (Candle, List) -> Double or something similar.
I am sure there is a better approach.
I am using Spark 2.1.0 and using Yarn as the scheduling engine.
I've done a fair bit of time series processing in Spark, and have spent some time thinking about exactly the same problem.
Unfortunately, in my opinion, there is no nice way to process all of the data, in the way you want, without structuring it as you suggested. I think we just have to accept that this kinda thing is an expensive operation, whether we are using Spark, pandas or Postgres.
You may hide the code complexity by using Spark SQL window functions (look at rangeBetween / RANGE BETWEEN), but the essence of what you are doing cannot be escaped.
Protip: map the data to features->label once and write it to disk to make dev/testing faster!
Related
I have a kind of requirement but not able to figure out how can I solve it. I have datasets in below format
id, atime, grade
123, time1, A
241, time2, B
123, time3, C
or if I put in list format:
[[123,time1,A],[124,timeb,C],[123,timec,C],[143,timed,D],[423,timee,P].......]
Now my use-case is to perform comparison, aggregation and queries over multiple row like
time difference between last 2 rows where id=123
time difference between last 2 rows where id=123&GradeA
Time difference between first, 3rd, 5th and latest one
all data (or last 10 records for particular id) should be easily accessible.
Also need to further do compute. What format should I chose for dataset
and what database/tools should I use?
I don't Relational Database is useful here. I am not able to solve it with Solr/Elastic if you have any ideas, please give a brief.Or any other tool Spark, hadoop, cassandra any heads?
I am trying out things but any help is appreciated.
Choosing the right technology is highly dependent on things related to your SLA. things like how much can your query have latency? what are your query types? is your data categorized as big data or not? Is data updateable? Do we expect late events? Do we need historical data in the future or we can use techniques like rollup? and things like that. To clarify my answer, probably by using window functions you can solve your problems. For example, you can store your data on any of the tools you mentioned and by using the Presto SQL engine you can query and get your desired result. But not all of them are optimal. Furthermore, usually, these kinds of problems can not be solved with a single tool. A set of tools can cover all requirements.
tl;dr. In the below text we don't find a solution. It introduces a way to think about data modeling and choosing tools.
Let me take try to model the problem to choose a single tool. I assume your data is not updatable, you need a low latency response time, we don't expect any late event and we face a large volume data stream that must be saved as raw data.
Based on the first and second requirements, it's crucial to have random access (it seems you wanna query on a particular ID), so solutions like parquet or ORC files are not a good choice.
Based on the last requirement, data must be partitioned based on the ID. Both the first and second requirements and the last requirement, count on ID as an identifier part and it seems there is nothing like join and global ordering based on other fields like time. So we can choose ID as the partitioner (physical or logical) and atime as the cluster part; For each ID, events are ordered based on the time.
The third requirement is a bit vague. You wanna result on all data? or for each ID?
For computing the first three conditions, we need a tool that supports window functions.
Based on the mentioned notes, it seems we should choose a tool that has good support for random access queries. Tools like Cassandra, Postgres, Druid, MongoDB, and ElasticSearch are things that currently I can remember them. Let's check them:
Cassandra: It's great on response time on random access queries, can handle a huge amount of data easily, and does not have a single point of failure. But sadly it does not support window functions. Also, you should carefully design your data model and it seems it's not a good tool that we can choose (because of future need for raw data). We can bypass some of these limitations by using Spark alongside Cassandra, but for now, we prefer to avoid adding a new tool to our stack.
Postgres: It's great on random access queries and indexed columns. It supports window functions. We can shard data (horizontal partitioning) across multiple servers (and by choosing ID as the shard key, we can have data locality on computations). But there is a problem: ID is not unique; so we can not choose ID as the primary key and we face some problems with random access (We can choose the ID and atime columns (as a timestamp column) as a compound primary key, but it does not save us).
Druid: It's a great OLAP tool. Based on the storing manner (segment files) that Druid follows, by choosing the right data model, you can have analytic queries on a huge volume of data in sub-seconds. It does not support window functions, but with rollup and some other functions (like EARLIEST), we can answer our questions. But by using rollup, we lose raw data and we need them.
MongoDB: It supports random access queries and sharding. Also, we can have some type of window function on its computing framework and we can define some sort of pipelines for doing aggregations. It supports capped collections and we can use it to store the last 10 events for each ID if the cardinality of the ID column is not high. It seems this tool can cover all of our requirements.
ElasticSearch: It's great on random access, maybe the greatest. With some kind of filter aggregations, we can have a type of window function. It can handle a large amount of data with sharding. But its query language is hard. I can imagine we can answer the first and second questions with ES, but for now, I can't make a query in my mind. It takes time to find the right solution with it.
So it seems MongoDB and ElasticSearch can answer our requirements, but there is a lot of 'if's on the way. I think we can't find a straightforward solution with a single tool. Maybe we should choose multiple tools and use techniques like duplicating data to find an optimal solution.
I've been trying to think about what the ideal table structure would be for the fastest Spark queries.
I'll try and provide a use case: Let's say your gathering stats for every car in the world and you want to use calculate various metrics with basic math (i.e. add, sub, mult, div).
Would be better to structure the data in a tall table with minimal fields like: day, metric, type, value?
Or would it be better to build a wide tables, that may store metrics independently. With more fields like: day, emmision_value, tire_pressure_value, speed_value, weight_value, heat_value, radio_value, etc .
Is it right to say that tall tables are better for spark? I assume it would be less memory intensive with a taller table.
As mentioned in the comments, this is a subjective question not exactly related to spark, but I'll try and answer none the less.
I assume it would be less memory intensive with a taller table.
Not really, the amount of storage required should be the same in either case based on the use case you have mentioned so let's get this out of the way. In case of taller tables there more rows and lesser columns and in case of wide tables the opposite. So on a cell level it should roughly be the same. I'm considering un compressed data independent of storage format.
Now lets talk about the mentioned use case. Simply put, it's aggregations. This may be fed downstream or may be used for reporting. Generally keeping this is mind, wider tables/views are better simply because - Lesser rows per day = less I/O as less shuffle.
Having said that, look through the cons below as well,
Schema evolution problems due to fixed schema
more suited for batch processing
Taller tables will be more streaming friendly, easier to extend for additional metrics and if its used with a source that supports push down, can result in quick partial scans.
in short, it very much depends on your operations.
I'm using intake to access the catalog catalog.ocean.GFDL_CM2_6.GFDL_CM2_6_control_ocean_surface.
At the moment I only work with small patches of that data, but accessing that data every single time is still quite costly (it's on Google Cloud Storage). So I want to use the persist option of intake to store that data locally. However as far as I've understood from the docs, it looks like one can only persist the whole dataset. For that specific dataset that would amount to almost 400 dollars if I take a cost of 0.1$ per GB, since the total data is 3976GB.
Hence my questions:
Is there a way (especially for a zarr file which in theory should make this quite easy) to persist only parts of the data (for instance only a subset of the variables)
This is probably more complicated, but can I push things further, by persisting regions of data I'm interested in (in terms of coordinates values for instance)?
There is no direct Intake way to do what you are asking for. Intake was conceived as a way to get your data into a format that you can then manipulate as you normally do, i.e., deal with only the loading part, so that a persisted data-set is the same as the original.
However, it is not hard to accomplish manually: you should grab the xarray, filter for the region you need, and call to_zarr to save the new dataset. You can then point a simple catalogue entry like the old one at the new location.
You could have done this manipulation in a driver directly if this was a specific pattern that would repeat a lot. In fact, we have mooted the idea of whether/how to implement such processing steps in Intake, but there is no plan yet. In the end, we may take the work on pipelines in Holoviews to describe processing steps.
We're investigating options to store and read a lot of immutable data (events) and I'd like some feedback on whether Cassandra would be a good fit.
Requirements:
We need to store about 10 events per seconds (but the rate will increase). Each event is small, about 1 Kb.
A really important requirement is that we need to be able to replay all events in order. For us it would be fine to read all data in insertion order (like a table scan) so an explicit sort might not be necessary.
Querying the data in any other way is not a prime concern and since Cassandra is a schema db I don't suppose it's possible when the events come in many different forms? Would Cassandra be a good fit for this? If so is there something one should be aware of?
I've had the exact same requirements for a "project" (rather a tool) a year ago, and I used Cassandra and I didn't regret. In general it fits very well. You can fit quite a lot of data in a Cassandra cluster and the performance is impressive (although you might need tweaking) and the natural ordering is a nice thing to have.
Rather than expressing the benefits of using it, I'll rather concentrate on possible pitfalls you might not consider before starting.
You have to think about your schema. The data is naturally ordered within one row by the clustering key, in your case it will be the timestamp. However, you cannot order data between different rows. They might be ordered after the query, but it is not guaranteed in any way so don't think about it. There was some kind of way to write a query before 2.1 I believe (using order by and disabling paging and allowing filtering) but that introduced bad performance and I don't think it is even possible now. So you should order data between rows on your querying side.
This might be an issue if you have multiple variable types (such as temperature and pressure) that have to be replayed at the same time, and you put them in different rows. You have to get those rows with different variable types, then do your resorting on the querying side. Another way to do it is to put all variable types in one row, but than filtering for only a subset is an issue to solve.
Rowlength is limited to 2 billion elements, and although that seems a lot, it really is not unreachable with time series data. Especially because you don't want to get near those two billions, keep it lower in hundreds of millions maximum. If you put some parameter on which you will split the rows (some increasing index or rounding by day/month/year) you will have to implement that in your query logic as well.
Experiment with your queries first on a dummy example. You cannot arbitrarily use <, > or = in queries. There are specific rules in SQL with filtering, or using the WHERE clause..
All in all these things might seem important, but they are really not too much of a hassle when you get to know Cassandra a bit. I'm underlining them just to give you a heads up. If something is not logical at first just fall back to understanding why it is like that and the whole theory about data distribution and the ring topology.
Don't expect too much from the collections within the columns, their length is limited to ~65000 elements.
Don't fall into the misconception that batched statements are faster (this one is a classic :) )
Based on the requirements you expressed, Cassandra could be a good fit as it's a write-optimized data store. Timeseries are quite a common pattern and you can define a clustering order, for example, on the timestamp of the events in order to retrieve all the events in time order. I've found this article on Datastax Academy very useful when wanted to learn about time series.
Variable data structure it's not a problem: you can store the data in a BLOB, then parse it internally from your application (i.e. store it as JSON and read it in your model), or you could even store the data in a map, although collections in Cassandra have some caveats that it's good to be aware of. Here you can find docs about collections in Cassandra 2.0/2.1.
Cassandra is quite different from a SQL database, and although CQL has some similarities there are fundamental differences in usage patterns. It's very important to know how Cassandra works and how to model your data in order to pursue efficiency - a great article from Datastax explains the basics of data modelling.
In a nutshell: Cassandra may be a good fit for you, but before using it take some time to understand its internals as it could be a bad beast if you use it poorly.
I am trying to figure out exactly what these new fangled data stores such as bigtable, hbase and cassandra really are.
I work with massive amounts of stock market data, billions of rows of price/quote data that can add up to 100s of gigabytes every day (although these text files often compress by at least an order of magnitude). This data is basically a handful of numbers, two or three short strings and a timestamp (usually millisecond level). If I had to pick a unique identifier for each row, I would have to pick the whole row (since an exchange may generate multiple values for the same symbol in the same millisecond).
I suppose the simplest way to map this data to bigtable (I'm including its derivatives) is by symbol name and date (which may return a very large time series, more than million data points isn't unheard of). From reading their descriptions, it looks like multiple keys can be used with these systems. I'm also assuming that decimal numbers are not good candidates for keys.
Some of these systems (Cassandra, for example) claims to be able to do range queries. Would I be able to efficiently query, say, all values for MSFT, for a given day, between 11:00 am and 1:30 pm ?
What if I want to search across ALL symbols for a given day, and request all symbols that have a price between $10 and $10.25 (so I'm searching the values, and want keys returned as a result)?
What if I want to get two times series, subtract one from the other, and return the two times series and their result, will I have to do his logic in my own program?
Reading relevant papers seems to show that these systems are not a very good fit for massive time series systems. However, if systems such as google maps are based on them, I think time series should work as well. For example, think of time as the x-axis, prices as y-axis and symbols as named locations--all of a sudden it looks like bigtable should be the ideal store for time series (if the whole earth can be stored, retrieved, zoomed and annotated, stock market data should be trivial).
Can some expert point me in the right direction or clear up any misunderstandings.
Thanks
I am not an expert yet, but I've been playing with Cassandra for a few days now, and I have some answers for you:
Don't worry about amount of data, it's irrelevant with systems like Cassandra, if you have $$$ for a large hardware cluster.
Some of these systems (Cassandra, for example) claims to be able to do range queries. Would I be able to efficiently query, say, all values for MSFT, for a given day, between 11:00 am and 1:30 pm ?
Cassandra is very useful when you know how to work with keys. It can swift through keys very quickly. So to search for MSFT between 11:00 and 1:30pm, you'd have to key your rows like this:
MSFT-timestamp, GOOG-timestamp , ..etc
Then you can tell Cassandra to find all keys that start with MSFT-now and end with MSFT-now+1hour.
What if I want to search across ALL symbols for a given day, and request all symbols that have a price between $10 and $10.25 (so I'm searching the values, and want keys returned as a result)?
I am not an expert, but so far I realized that Cassandra doesn't' search by values at all. So if you want to do the above, you will have to make another table dedicated just to this problem and design your schema to fit the case. But it won't be much different from what I described above. It's all about naming your keys and columns. Cassandra can find them very quickly!
What if I want to get two times series, subtract one from the other, and return the two times series and their result, will I have to do his logic in my own program?
Correct, all logic is done inside your program. This is not MySQL. This is just a storage engine. (But I am sure the next versions will offer these sort of things)
Please remember, that I am a novice at this, if I am wrong, feel free to correct me.
If you're dealing with a massive time series database, then the standards are:
KDB: http://www.kx.com/
OneTick: http://www.onetick.com
Vhayu: http://www.vhayu.com
These aren't cheap, but they can handle your data very efficiently.
Someone whom I respect recommended the Open Time Series Database. In particular, that the schema was the nicest he had ever seen.
http://opentsdb.net/
'Am standing in front of the same mountain. My main problem with cassandra is that I cannot get a stream on the result set, for example in the form of an iterator.
I am looking already up and down the docs and the net, but nothing.
I can't fetch all the keys and then get the rows as billions of rows makes this impossible.
The DataStax Java Driver allows for automatic paging so that will stream the results just like an iterator and it's all built in. This is in Cassandra 2.0.1 by the way - http://www.datastax.com/dev/blog/client-side-improvements-in-cassandra-2-0
Just for the sake of completeness reading this in 2018, there is now a special database just for timeseries data called TimescaleDB
http://www.timescale.com/
This blog is worth reading, it explains why it´s superior to solutions like Cassandra for that special case and why they decided to build it on top of the relational PostgreSQL database
https://blog.timescale.com/time-series-data-why-and-how-to-use-a-relational-database-instead-of-nosql-d0cd6975e87c