Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.
We don’t allow questions seeking recommendations for books, tools, software libraries, and more. You can edit the question so it can be answered with facts and citations.
Closed 1 year ago.
The community reviewed whether to reopen this question 1 year ago and left it closed:
Original close reason(s) were not resolved
Improve this question
I have tried to puzzle out an answer to this question for many months while learning pandas. I use SAS for my day-to-day work and it is great for it's out-of-core support. However, SAS is horrible as a piece of software for numerous other reasons.
One day I hope to replace my use of SAS with python and pandas, but I currently lack an out-of-core workflow for large datasets. I'm not talking about "big data" that requires a distributed network, but rather files too large to fit in memory but small enough to fit on a hard-drive.
My first thought is to use HDFStore to hold large datasets on disk and pull only the pieces I need into dataframes for analysis. Others have mentioned MongoDB as an easier to use alternative. My question is this:
What are some best-practice workflows for accomplishing the following:
Loading flat files into a permanent, on-disk database structure
Querying that database to retrieve data to feed into a pandas data structure
Updating the database after manipulating pieces in pandas
Real-world examples would be much appreciated, especially from anyone who uses pandas on "large data".
Edit -- an example of how I would like this to work:
Iteratively import a large flat-file and store it in a permanent, on-disk database structure. These files are typically too large to fit in memory.
In order to use Pandas, I would like to read subsets of this data (usually just a few columns at a time) that can fit in memory.
I would create new columns by performing various operations on the selected columns.
I would then have to append these new columns into the database structure.
I am trying to find a best-practice way of performing these steps. Reading links about pandas and pytables it seems that appending a new column could be a problem.
Edit -- Responding to Jeff's questions specifically:
I am building consumer credit risk models. The kinds of data include phone, SSN and address characteristics; property values; derogatory information like criminal records, bankruptcies, etc... The datasets I use every day have nearly 1,000 to 2,000 fields on average of mixed data types: continuous, nominal and ordinal variables of both numeric and character data. I rarely append rows, but I do perform many operations that create new columns.
Typical operations involve combining several columns using conditional logic into a new, compound column. For example, if var1 > 2 then newvar = 'A' elif var2 = 4 then newvar = 'B'. The result of these operations is a new column for every record in my dataset.
Finally, I would like to append these new columns into the on-disk data structure. I would repeat step 2, exploring the data with crosstabs and descriptive statistics trying to find interesting, intuitive relationships to model.
A typical project file is usually about 1GB. Files are organized into such a manner where a row consists of a record of consumer data. Each row has the same number of columns for every record. This will always be the case.
It's pretty rare that I would subset by rows when creating a new column. However, it's pretty common for me to subset on rows when creating reports or generating descriptive statistics. For example, I might want to create a simple frequency for a specific line of business, say Retail credit cards. To do this, I would select only those records where the line of business = retail in addition to whichever columns I want to report on. When creating new columns, however, I would pull all rows of data and only the columns I need for the operations.
The modeling process requires that I analyze every column, look for interesting relationships with some outcome variable, and create new compound columns that describe those relationships. The columns that I explore are usually done in small sets. For example, I will focus on a set of say 20 columns just dealing with property values and observe how they relate to defaulting on a loan. Once those are explored and new columns are created, I then move on to another group of columns, say college education, and repeat the process. What I'm doing is creating candidate variables that explain the relationship between my data and some outcome. At the very end of this process, I apply some learning techniques that create an equation out of those compound columns.
It is rare that I would ever add rows to the dataset. I will nearly always be creating new columns (variables or features in statistics/machine learning parlance).
I routinely use tens of gigabytes of data in just this fashion
e.g. I have tables on disk that I read via queries, create data and append back.
It's worth reading the docs and late in this thread for several suggestions for how to store your data.
Details which will affect how you store your data, like:
Give as much detail as you can; and I can help you develop a structure.
Size of data, # of rows, columns, types of columns; are you appending
rows, or just columns?
What will typical operations look like. E.g. do a query on columns to select a bunch of rows and specific columns, then do an operation (in-memory), create new columns, save these.
(Giving a toy example could enable us to offer more specific recommendations.)
After that processing, then what do you do? Is step 2 ad hoc, or repeatable?
Input flat files: how many, rough total size in Gb. How are these organized e.g. by records? Does each one contains different fields, or do they have some records per file with all of the fields in each file?
Do you ever select subsets of rows (records) based on criteria (e.g. select the rows with field A > 5)? and then do something, or do you just select fields A, B, C with all of the records (and then do something)?
Do you 'work on' all of your columns (in groups), or are there a good proportion that you may only use for reports (e.g. you want to keep the data around, but don't need to pull in that column explicity until final results time)?
Solution
Ensure you have pandas at least 0.10.1 installed.
Read iterating files chunk-by-chunk and multiple table queries.
Since pytables is optimized to operate on row-wise (which is what you query on), we will create a table for each group of fields. This way it's easy to select a small group of fields (which will work with a big table, but it's more efficient to do it this way... I think I may be able to fix this limitation in the future... this is more intuitive anyhow):
(The following is pseudocode.)
import numpy as np
import pandas as pd
# create a store
store = pd.HDFStore('mystore.h5')
# this is the key to your storage:
# this maps your fields to a specific group, and defines
# what you want to have as data_columns.
# you might want to create a nice class wrapping this
# (as you will want to have this map and its inversion)
group_map = dict(
A = dict(fields = ['field_1','field_2',.....], dc = ['field_1',....,'field_5']),
B = dict(fields = ['field_10',...... ], dc = ['field_10']),
.....
REPORTING_ONLY = dict(fields = ['field_1000','field_1001',...], dc = []),
)
group_map_inverted = dict()
for g, v in group_map.items():
group_map_inverted.update(dict([ (f,g) for f in v['fields'] ]))
Reading in the files and creating the storage (essentially doing what append_to_multiple does):
for f in files:
# read in the file, additional options may be necessary here
# the chunksize is not strictly necessary, you may be able to slurp each
# file into memory in which case just eliminate this part of the loop
# (you can also change chunksize if necessary)
for chunk in pd.read_table(f, chunksize=50000):
# we are going to append to each table by group
# we are not going to create indexes at this time
# but we *ARE* going to create (some) data_columns
# figure out the field groupings
for g, v in group_map.items():
# create the frame for this group
frame = chunk.reindex(columns = v['fields'], copy = False)
# append it
store.append(g, frame, index=False, data_columns = v['dc'])
Now you have all of the tables in the file (actually you could store them in separate files if you wish, you would prob have to add the filename to the group_map, but probably this isn't necessary).
This is how you get columns and create new ones:
frame = store.select(group_that_I_want)
# you can optionally specify:
# columns = a list of the columns IN THAT GROUP (if you wanted to
# select only say 3 out of the 20 columns in this sub-table)
# and a where clause if you want a subset of the rows
# do calculations on this frame
new_frame = cool_function_on_frame(frame)
# to 'add columns', create a new group (you probably want to
# limit the columns in this new_group to be only NEW ones
# (e.g. so you don't overlap from the other tables)
# add this info to the group_map
store.append(new_group, new_frame.reindex(columns = new_columns_created, copy = False), data_columns = new_columns_created)
When you are ready for post_processing:
# This may be a bit tricky; and depends what you are actually doing.
# I may need to modify this function to be a bit more general:
report_data = store.select_as_multiple([groups_1,groups_2,.....], where =['field_1>0', 'field_1000=foo'], selector = group_1)
About data_columns, you don't actually need to define ANY data_columns; they allow you to sub-select rows based on the column. E.g. something like:
store.select(group, where = ['field_1000=foo', 'field_1001>0'])
They may be most interesting to you in the final report generation stage (essentially a data column is segregated from other columns, which might impact efficiency somewhat if you define a lot).
You also might want to:
create a function which takes a list of fields, looks up the groups in the groups_map, then selects these and concatenates the results so you get the resulting frame (this is essentially what select_as_multiple does). This way the structure would be pretty transparent to you.
indexes on certain data columns (makes row-subsetting much faster).
enable compression.
Let me know when you have questions!
I think the answers above are missing a simple approach that I've found very useful.
When I have a file that is too large to load in memory, I break up the file into multiple smaller files (either by row or cols)
Example: In case of 30 days worth of trading data of ~30GB size, I break it into a file per day of ~1GB size. I subsequently process each file separately and aggregate results at the end
One of the biggest advantages is that it allows parallel processing of the files (either multiple threads or processes)
The other advantage is that file manipulation (like adding/removing dates in the example) can be accomplished by regular shell commands, which is not be possible in more advanced/complicated file formats
This approach doesn't cover all scenarios, but is very useful in a lot of them
There is now, two years after the question, an 'out-of-core' pandas equivalent: dask. It is excellent! Though it does not support all of pandas functionality, you can get really far with it. Update: in the past two years it has been consistently maintained and there is substantial user community working with Dask.
And now, four years after the question, there is another high-performance 'out-of-core' pandas equivalent in Vaex. It "uses memory mapping, zero memory copy policy and lazy computations for best performance (no memory wasted)." It can handle data sets of billions of rows and does not store them into memory (making it even possible to do analysis on suboptimal hardware).
If your datasets are between 1 and 20GB, you should get a workstation with 48GB of RAM. Then Pandas can hold the entire dataset in RAM. I know its not the answer you're looking for here, but doing scientific computing on a notebook with 4GB of RAM isn't reasonable.
I know this is an old thread but I think the Blaze library is worth checking out. It's built for these types of situations.
From the docs:
Blaze extends the usability of NumPy and Pandas to distributed and out-of-core computing. Blaze provides an interface similar to that of the NumPy ND-Array or Pandas DataFrame but maps these familiar interfaces onto a variety of other computational engines like Postgres or Spark.
Edit: By the way, it's supported by ContinuumIO and Travis Oliphant, author of NumPy.
This is the case for pymongo. I have also prototyped using sql server, sqlite, HDF, ORM (SQLAlchemy) in python. First and foremost pymongo is a document based DB, so each person would be a document (dict of attributes). Many people form a collection and you can have many collections (people, stock market, income).
pd.dateframe -> pymongo Note: I use the chunksize in read_csv to keep it to 5 to 10k records(pymongo drops the socket if larger)
aCollection.insert((a[1].to_dict() for a in df.iterrows()))
querying: gt = greater than...
pd.DataFrame(list(mongoCollection.find({'anAttribute':{'$gt':2887000, '$lt':2889000}})))
.find() returns an iterator so I commonly use ichunked to chop into smaller iterators.
How about a join since I normally get 10 data sources to paste together:
aJoinDF = pandas.DataFrame(list(mongoCollection.find({'anAttribute':{'$in':Att_Keys}})))
then (in my case sometimes I have to agg on aJoinDF first before its "mergeable".)
df = pandas.merge(df, aJoinDF, on=aKey, how='left')
And you can then write the new info to your main collection via the update method below. (logical collection vs physical datasources).
collection.update({primarykey:foo},{key:change})
On smaller lookups, just denormalize. For example, you have code in the document and you just add the field code text and do a dict lookup as you create documents.
Now you have a nice dataset based around a person, you can unleash your logic on each case and make more attributes. Finally you can read into pandas your 3 to memory max key indicators and do pivots/agg/data exploration. This works for me for 3 million records with numbers/big text/categories/codes/floats/...
You can also use the two methods built into MongoDB (MapReduce and aggregate framework). See here for more info about the aggregate framework, as it seems to be easier than MapReduce and looks handy for quick aggregate work. Notice I didn't need to define my fields or relations, and I can add items to a document. At the current state of the rapidly changing numpy, pandas, python toolset, MongoDB helps me just get to work :)
One trick I found helpful for large data use cases is to reduce the volume of the data by reducing float precision to 32-bit. It's not applicable in all cases, but in many applications 64-bit precision is overkill and the 2x memory savings are worth it. To make an obvious point even more obvious:
>>> df = pd.DataFrame(np.random.randn(int(1e8), 5))
>>> df.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 100000000 entries, 0 to 99999999
Data columns (total 5 columns):
...
dtypes: float64(5)
memory usage: 3.7 GB
>>> df.astype(np.float32).info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 100000000 entries, 0 to 99999999
Data columns (total 5 columns):
...
dtypes: float32(5)
memory usage: 1.9 GB
I spotted this a little late, but I work with a similar problem (mortgage prepayment models). My solution has been to skip the pandas HDFStore layer and use straight pytables. I save each column as an individual HDF5 array in my final file.
My basic workflow is to first get a CSV file from the database. I gzip it, so it's not as huge. Then I convert that to a row-oriented HDF5 file, by iterating over it in python, converting each row to a real data type, and writing it to a HDF5 file. That takes some tens of minutes, but it doesn't use any memory, since it's only operating row-by-row. Then I "transpose" the row-oriented HDF5 file into a column-oriented HDF5 file.
The table transpose looks like:
def transpose_table(h_in, table_path, h_out, group_name="data", group_path="/"):
# Get a reference to the input data.
tb = h_in.getNode(table_path)
# Create the output group to hold the columns.
grp = h_out.createGroup(group_path, group_name, filters=tables.Filters(complevel=1))
for col_name in tb.colnames:
logger.debug("Processing %s", col_name)
# Get the data.
col_data = tb.col(col_name)
# Create the output array.
arr = h_out.createCArray(grp,
col_name,
tables.Atom.from_dtype(col_data.dtype),
col_data.shape)
# Store the data.
arr[:] = col_data
h_out.flush()
Reading it back in then looks like:
def read_hdf5(hdf5_path, group_path="/data", columns=None):
"""Read a transposed data set from a HDF5 file."""
if isinstance(hdf5_path, tables.file.File):
hf = hdf5_path
else:
hf = tables.openFile(hdf5_path)
grp = hf.getNode(group_path)
if columns is None:
data = [(child.name, child[:]) for child in grp]
else:
data = [(child.name, child[:]) for child in grp if child.name in columns]
# Convert any float32 columns to float64 for processing.
for i in range(len(data)):
name, vec = data[i]
if vec.dtype == np.float32:
data[i] = (name, vec.astype(np.float64))
if not isinstance(hdf5_path, tables.file.File):
hf.close()
return pd.DataFrame.from_items(data)
Now, I generally run this on a machine with a ton of memory, so I may not be careful enough with my memory usage. For example, by default the load operation reads the whole data set.
This generally works for me, but it's a bit clunky, and I can't use the fancy pytables magic.
Edit: The real advantage of this approach, over the array-of-records pytables default, is that I can then load the data into R using h5r, which can't handle tables. Or, at least, I've been unable to get it to load heterogeneous tables.
As noted by others, after some years an 'out-of-core' pandas equivalent has emerged: dask. Though dask is not a drop-in replacement of pandas and all of its functionality it stands out for several reasons:
Dask is a flexible parallel computing library for analytic computing that is optimized for dynamic task scheduling for interactive computational workloads of
“Big Data” collections like parallel arrays, dataframes, and lists that extend common interfaces like NumPy, Pandas, or Python iterators to larger-than-memory or distributed environments and scales from laptops to clusters.
Dask emphasizes the following virtues:
Familiar: Provides parallelized NumPy array and Pandas DataFrame objects
Flexible: Provides a task scheduling interface for more custom workloads and integration with other projects.
Native: Enables distributed computing in Pure Python with access to the PyData stack.
Fast: Operates with low overhead, low latency, and minimal serialization necessary for fast numerical algorithms
Scales up: Runs resiliently on clusters with 1000s of cores Scales down: Trivial to set up and run on a laptop in a single process
Responsive: Designed with interactive computing in mind it provides rapid feedback and diagnostics to aid humans
and to add a simple code sample:
import dask.dataframe as dd
df = dd.read_csv('2015-*-*.csv')
df.groupby(df.user_id).value.mean().compute()
replaces some pandas code like this:
import pandas as pd
df = pd.read_csv('2015-01-01.csv')
df.groupby(df.user_id).value.mean()
and, especially noteworthy, provides through the concurrent.futures interface a general infrastructure for the submission of custom tasks:
from dask.distributed import Client
client = Client('scheduler:port')
futures = []
for fn in filenames:
future = client.submit(load, fn)
futures.append(future)
summary = client.submit(summarize, futures)
summary.result()
It is worth mentioning here Ray as well,
it's a distributed computation framework, that has it's own implementation for pandas in a distributed way.
Just replace the pandas import, and the code should work as is:
# import pandas as pd
import ray.dataframe as pd
# use pd as usual
can read more details here:
https://rise.cs.berkeley.edu/blog/pandas-on-ray/
Update:
the part that handles the pandas distribution, has been extracted to the modin project.
the proper way to use it is now is:
# import pandas as pd
import modin.pandas as pd
One more variation
Many of the operations done in pandas can also be done as a db query (sql, mongo)
Using a RDBMS or mongodb allows you to perform some of the aggregations in the DB Query (which is optimized for large data, and uses cache and indexes efficiently)
Later, you can perform post processing using pandas.
The advantage of this method is that you gain the DB optimizations for working with large data, while still defining the logic in a high level declarative syntax - and not having to deal with the details of deciding what to do in memory and what to do out of core.
And although the query language and pandas are different, it's usually not complicated to translate part of the logic from one to another.
Consider Ruffus if you go the simple path of creating a data pipeline which is broken down into multiple smaller files.
I'd like to point out the Vaex package.
Vaex is a python library for lazy Out-of-Core DataFrames (similar to Pandas), to visualize and explore big tabular datasets. It can calculate statistics such as mean, sum, count, standard deviation etc, on an N-dimensional grid up to a billion (109) objects/rows per second. Visualization is done using histograms, density plots and 3d volume rendering, allowing interactive exploration of big data. Vaex uses memory mapping, zero memory copy policy and lazy computations for best performance (no memory wasted).
Have a look at the documentation: https://vaex.readthedocs.io/en/latest/
The API is very close to the API of pandas.
I recently came across a similar issue. I found simply reading the data in chunks and appending it as I write it in chunks to the same csv works well. My problem was adding a date column based on information in another table, using the value of certain columns as follows. This may help those confused by dask and hdf5 but more familiar with pandas like myself.
def addDateColumn():
"""Adds time to the daily rainfall data. Reads the csv as chunks of 100k
rows at a time and outputs them, appending as needed, to a single csv.
Uses the column of the raster names to get the date.
"""
df = pd.read_csv(pathlist[1]+"CHIRPS_tanz.csv", iterator=True,
chunksize=100000) #read csv file as 100k chunks
'''Do some stuff'''
count = 1 #for indexing item in time list
for chunk in df: #for each 100k rows
newtime = [] #empty list to append repeating times for different rows
toiterate = chunk[chunk.columns[2]] #ID of raster nums to base time
while count <= toiterate.max():
for i in toiterate:
if i ==count:
newtime.append(newyears[count])
count+=1
print "Finished", str(chunknum), "chunks"
chunk["time"] = newtime #create new column in dataframe based on time
outname = "CHIRPS_tanz_time2.csv"
#append each output to same csv, using no header
chunk.to_csv(pathlist[2]+outname, mode='a', header=None, index=None)
The parquet file format is ideal for the use case you described. You can efficiently read in a specific subset of columns with pd.read_parquet(path_to_file, columns=["foo", "bar"])
https://pandas.pydata.org/docs/reference/api/pandas.read_parquet.html
At the moment I am working "like" you, just on a lower scale, which is why I don't have a PoC for my suggestion.
However, I seem to find success in using pickle as caching system and outsourcing execution of various functions into files - executing these files from my commando / main file; For example i use a prepare_use.py to convert object types, split a data set into test, validating and prediction data set.
How does your caching with pickle work?
I use strings in order to access pickle-files that are dynamically created, depending on which parameters and data sets were passed (with that i try to capture and determine if the program was already run, using .shape for data set, dict for passed parameters).
Respecting these measures, i get a String to try to find and read a .pickle-file and can, if found, skip processing time in order to jump to the execution i am working on right now.
Using databases I encountered similar problems, which is why i found joy in using this solution, however - there are many constraints for sure - for example storing huge pickle sets due to redundancy.
Updating a table from before to after a transformation can be done with proper indexing - validating information opens up a whole other book (I tried consolidating crawled rent data and stopped using a database after 2 hours basically - as I would have liked to jump back after every transformation process)
I hope my 2 cents help you in some way.
Greetings.
Long story short, I'm rewriting a piece of a system and am looking for a way to store some hit counters in AWS SimpleDB.
For those of you not familiar with SimpleDB, the (main) problem with storing counters is that the cloud propagation delay is often over a second. Our application currently gets ~1,500 hits per second. Not all those hits will map to the same key, but a ballpark figure might be around 5-10 updates to a key every second. This means that if we were to use a traditional update mechanism (read, increment, store), we would end up inadvertently dropping a significant number of hits.
One potential solution is to keep the counters in memcache, and using a cron task to push the data. The big problem with this is that it isn't the "right" way to do it. Memcache shouldn't really be used for persistent storage... after all, it's a caching layer. In addition, then we'll end up with issues when we do the push, making sure we delete the correct elements, and hoping that there is no contention for them as we're deleting them (which is very likely).
Another potential solution is to keep a local SQL database and write the counters there, updating our SimpleDB out-of-band every so many requests or running a cron task to push the data. This solves the syncing problem, as we can include timestamps to easily set boundaries for the SimpleDB pushes. Of course, there are still other issues, and though this might work with a decent amount of hacking, it doesn't seem like the most elegant solution.
Has anyone encountered a similar issue in their experience, or have any novel approaches? Any advice or ideas would be appreciated, even if they're not completely flushed out. I've been thinking about this one for a while, and could use some new perspectives.
The existing SimpleDB API does not lend itself naturally to being a distributed counter. But it certainly can be done.
Working strictly within SimpleDB there are 2 ways to make it work. An easy method that requires something like a cron job to clean up. Or a much more complex technique that cleans as it goes.
The Easy Way
The easy way is to make a different item for each "hit". With a single attribute which is the key. Pump the domain(s) with counts quickly and easily. When you need to fetch the count (presumable much less often) you have to issue a query
SELECT count(*) FROM domain WHERE key='myKey'
Of course this will cause your domain(s) to grow unbounded and the queries will take longer and longer to execute over time. The solution is a summary record where you roll up all the counts collected so far for each key. It's just an item with attributes for the key {summary='myKey'} and a "Last-Updated" timestamp with granularity down to the millisecond. This also requires that you add the "timestamp" attribute to your "hit" items. The summary records don't need to be in the same domain. In fact, depending on your setup, they might best be kept in a separate domain. Either way you can use the key as the itemName and use GetAttributes instead of doing a SELECT.
Now getting the count is a two step process. You have to pull the summary record and also query for 'Timestamp' strictly greater than whatever the 'Last-Updated' time is in your summary record and add the two counts together.
SELECT count(*) FROM domain WHERE key='myKey' AND timestamp > '...'
You will also need a way to update your summary record periodically. You can do this on a schedule (every hour) or dynamically based on some other criteria (for example do it during regular processing whenever the query returns more than one page). Just make sure that when you update your summary record you base it on a time that is far enough in the past that you are past the eventual consistency window. 1 minute is more than safe.
This solution works in the face of concurrent updates because even if many summary records are written at the same time, they are all correct and whichever one wins will still be correct because the count and the 'Last-Updated' attribute will be consistent with each other.
This also works well across multiple domains even if you keep your summary records with the hit records, you can pull the summary records from all your domains simultaneously and then issue your queries to all domains in parallel. The reason to do this is if you need higher throughput for a key than what you can get from one domain.
This works well with caching. If your cache fails you have an authoritative backup.
The time will come where someone wants to go back and edit / remove / add a record that has an old 'Timestamp' value. You will have to update your summary record (for that domain) at that time or your counts will be off until you recompute that summary.
This will give you a count that is in sync with the data currently viewable within the consistency window. This won't give you a count that is accurate up to the millisecond.
The Hard Way
The other way way is to do the normal read - increment - store mechanism but also write a composite value that includes a version number along with your value. Where the version number you use is 1 greater than the version number of the value you are updating.
get(key) returns the attribute value="Ver015 Count089"
Here you retrieve a count of 89 that was stored as version 15. When you do an update you write a value like this:
put(key, value="Ver016 Count090")
The previous value is not removed and you end up with an audit trail of updates that are reminiscent of lamport clocks.
This requires you to do a few extra things.
the ability to identify and resolve conflicts whenever you do a GET
a simple version number isn't going to work you'll want to include a timestamp with resolution down to at least the millisecond and maybe a process ID as well.
in practice you'll want your value to include the current version number and the version number of the value your update is based on to more easily resolve conflicts.
you can't keep an infinite audit trail in one item so you'll need to issue delete's for older values as you go.
What you get with this technique is like a tree of divergent updates. you'll have one value and then all of a sudden multiple updates will occur and you will have a bunch of updates based off the same old value none of which know about each other.
When I say resolve conflicts at GET time I mean that if you read an item and the value looks like this:
11 --- 12
/
10 --- 11
\
11
You have to to be able to figure that the real value is 14. Which you can do if you include for each new value the version of the value(s) you are updating.
It shouldn't be rocket science
If all you want is a simple counter: this is way over-kill. It shouldn't be rocket science to make a simple counter. Which is why SimpleDB may not be the best choice for making simple counters.
That isn't the only way but most of those things will need to be done if you implement an SimpleDB solution in lieu of actually having a lock.
Don't get me wrong, I actually like this method precisely because there is no lock and the bound on the number of processes that can use this counter simultaneously is around 100. (because of the limit on the number of attributes in an item) And you can get beyond 100 with some changes.
Note
But if all these implementation details were hidden from you and you just had to call increment(key), it wouldn't be complex at all. With SimpleDB the client library is the key to making the complex things simple. But currently there are no publicly available libraries that implement this functionality (to my knowledge).
To anyone revisiting this issue, Amazon just added support for Conditional Puts, which makes implementing a counter much easier.
Now, to implement a counter - simply call GetAttributes, increment the count, and then call PutAttributes, with the Expected Value set correctly. If Amazon responds with an error ConditionalCheckFailed, then retry the whole operation.
Note that you can only have one expected value per PutAttributes call. So, if you want to have multiple counters in a single row, then use a version attribute.
pseudo-code:
begin
attributes = SimpleDB.GetAttributes
initial_version = attributes[:version]
attributes[:counter1] += 3
attributes[:counter2] += 7
attributes[:version] += 1
SimpleDB.PutAttributes(attributes, :expected => {:version => initial_version})
rescue ConditionalCheckFailed
retry
end
I see you've accepted an answer already, but this might count as a novel approach.
If you're building a web app then you can use Google's Analytics product to track page impressions (if the page to domain-item mapping fits) and then to use the Analytics API to periodically push that data up into the items themselves.
I haven't thought this through in detail so there may be holes. I'd actually be quite interested in your feedback on this approach given your experience in the area.
Thanks
Scott
For anyone interested in how I ended up dealing with this... (slightly Java-specific)
I ended up using an EhCache on each servlet instance. I used the UUID as a key, and a Java AtomicInteger as the value. Periodically a thread iterates through the cache and pushes rows to a simpledb temp stats domain, as well as writing a row with the key to an invalidation domain (which fails silently if the key already exists). The thread also decrements the counter with the previous value, ensuring that we don't miss any hits while it was updating. A separate thread pings the simpledb invalidation domain, and rolls up the stats in the temporary domains (there are multiple rows to each key, since we're using ec2 instances), pushing it to the actual stats domain.
I've done a little load testing, and it seems to scale well. Locally I was able to handle about 500 hits/second before the load tester broke (not the servlets - hah), so if anything I think running on ec2 should only improve performance.
Answer to feynmansbastard:
If you want to store huge amount of events i suggest you to use distributed commit log systems such as kafka or aws kinesis. They allow to consume stream of events cheap and simple (kinesis's pricing is 25$ per month for 1K events per seconds) – you just need to implement consumer (using any language), which bulk reads all events from previous checkpoint, aggregates counters in memory then flushes data into permanent storage (dynamodb or mysql) and commit checkpoint.
Events can be logged simply using nginx log and transfered to kafka/kinesis using fluentd. This is very cheap, performant and simple solution.
Also had similiar needs/challenges.
I looked at using google analytics and count.ly. the latter seemed too expensive to be worth it (plus they have a somewhat confusion definition of sessions). GA i would have loved to use, but I spent two days using their libraries and some 3rd party ones (gadotnet and one other from maybe codeproject). unfortunately I could only ever see counters post in GA realtime section, never in the normal dashboards even when the api reported success. we were probably doing something wrong but we exceeded our time budget for ga.
We already had an existing simpledb counter that updated using conditional updates as mentioned by previous commentor. This works well, but suffers when there is contention and conccurency where counts are missed (for example, our most updated counter lost several million counts over a period of 3 months, versus a backup system).
We implemented a newer solution which is somewhat similiar to the answer for this question, except much simpler.
We just sharded/partitioned the counters. When you create a counter you specify the # of shards which is a function of how many simulatenous updates you expect. this creates a number of sub counters, each which has the shard count started with it as an attribute :
COUNTER (w/5shards) creates :
shard0 { numshards = 5 } (informational only)
shard1 { count = 0, numshards = 5, timestamp = 0 }
shard2 { count = 0, numshards = 5, timestamp = 0 }
shard3 { count = 0, numshards = 5, timestamp = 0 }
shard4 { count = 0, numshards = 5, timestamp = 0 }
shard5 { count = 0, numshards = 5, timestamp = 0 }
Sharded Writes
Knowing the shard count, just randomly pick a shard and try to write to it conditionally. If it fails because of contention, choose another shard and retry.
If you don't know the shard count, get it from the root shard which is present regardless of how many shards exist. Because it supports multiple writes per counter, it lessens the contention issue to whatever your needs are.
Sharded Reads
if you know the shard count, read every shard and sum them.
If you don't know the shard count, get it from the root shard and then read all and sum.
Because of slow update propogation, you can still miss counts in reading but they should get picked up later. This is sufficient for our needs, although if you wanted more control over this you could ensure that- when reading- the last timestamp was as you expect and retry.