Develop Reference

Continuous aggregates over large datasets

I'm trying to think of an algorithm to solve this problem I have. It's not a HW problem, but for a side project I'm working on.
There's a table A that has about (order of) 10^5 rows and adds new in the order of 10^2 every day.
Table B has on the order of 10^6 rows and adds new at 10^3 every day. There's a one to many relation from A to B (many B rows for some row in A).
I was wondering how I could do continuous aggregates for this kind of data. I would like to have a job that runs every ~10mins and does this: For every row in A, find every row in B related to it that were created in the last day, week and month (and then sort by count) and save them in a different DB or cache them.
If this is confusing, here's a practical example: Say table A has Amazon products and table B has product reviews. We would like to show a sorted list of products with highest reviews in the last 4hrs, day, week etc. New products and reviews are added at a fast pace, and we'd like the said list to be as up-to-date as possible.
Current implementation I have is just a for loop (pseudo-code):
result = []
for product in db_products:
reviews = db_reviews(product_id=product.id, create>=some_time)
reviews_count = len(reviews)
result[product]['reviews'] = reviews
result[product]['reviews_count'] = reviews_count
sort(result, by=reviews_count)
return result
I do this every hour, and save the result in a json file to serve. The problem is that this doesn't really scale well, and takes a long time to compute.
So, where could I look to solve this problem?
UPDATE:
Thank you for your answers. But I ended up learning and using Apache Storm.

Summary of requirements
Having two bigger tables in a database, you need regularly creating some aggregates for past time periods (hour, day, week etc.) and store the results in another database.
I will assume, that once a time period is past, there are no changes to related records, in other words, the aggregate for past period has always the same result.
Proposed solution: Luigi
Luigi is framework for plumbing dependent tasks and one of typical uses is calculating aggregates for past periods.
The concept is as follows:
write simple Task instance, which defines required input data, output data (called Target) and process to create the target output.
Tasks can be parametrized, typical parameter is time period (specific day, hour, week etc.)
Luigi can stop tasks in the middle and start later. It will consider any task, for which is target already existing to be completed and will not rerun it (you would have to delete the target content to let it rerun).
In short: if the target exists, the task is done.
This works for multiple types of targets like files in local file system, on hadoop, at AWS S3, and also in database.
To prevent half done results, target implementations take care of atomicity, so e.g. files are first created in temporary location and are moved to final destination just after they are completed.
In databases there are structures to denote, that some database import is completed.
You are free to create your own target implementations (it has to create something and provide method exists to check, the result exists.
Using Luigi for your task
For the task you describe you will probably find everything you need already present. Just few tips:
class luigi.postgres.CopyToTable allowing to store records into Postgres database. The target will automatically create so called "marker table" where it will mark all completed tasks.
There are similar classes for other types of databases, one of them using SqlAlchemy which shall probably cover the database you use, see class luigi.contrib.sqla.CopyToTable
At Luigi doc is working example of importing data into sqlite database
Complete implementation is beyond extend feasible in StackOverflow answer, but I am sure, you will experience following:
The code to do the task is really clear - no boilerplate coding, just write only what has to be done.
nice support for working with time periods - even from command line, see e.g. Efficiently triggering recurring tasks. It even takes care of not going too far in past, to prevent generating too many tasks possibly overloading your servers (default values are very reasonably set and can be changed).
Option to run the task on multiple servers (using central scheduler, which is provided with Luigi implementation).
I have processed huge amounts of XML files with Luigi and also made some tasks, importing aggregated data into database and can recommend it (I am not author of Luigi, I am just happy user).
Speeding up database operations (queries)
If your task suffers from too long execution time to perform the database query, you have few options:
if you are counting reviews per product by Python, consider trying SQL query - it is often much faster. It shall be possible to create SQL query which uses count on proper records and returns directly the number you need. With group by you shall even get summary information for all products in one run.
set up proper index, probably on "reviews" table on "product" and "time period" column. This shall speed up the query, but make sure, it does not slow down inserting new records too much (too many indexes can cause that).
It might happen, that with optimized SQL query you will get working solution even without using Luigi.

Data Warehousing? Summary tables are the right way to go.
Does the data change (once it is written)? If it does, then incrementally updating Summary Tables becomes a challenge. Most DW applications do not have that problem
Update the summary table (day + dimension(s) + count(s) + sum(s)) as you insert into the raw data table(s). Since you are getting only one insert per minute, INSERT INTO SummaryTable ... ON DUPLICATE KEY UPDATE ... would be quite adequate, and simpler than running a script every 10 minutes.
Do any reporting from a summary table, not the raw data (the Fact table). It will be a lot faster.
My Blog on Summary Tables discusses details. (It is aimed at bigger DW applications, but should be useful reading.)

I agree with Rick, summary tables make the most sense for you. Update the summary tables every 10 minutes and just pull data from it, as user's request summaries.
Also, make sure that your DB is indexed properly for performance. I'm sure db_products.id set as a unique index. but, also make sure that db_products.create is defined as a DATE or DATETIME and also indexed since you are using it in your WHERE statement.

Poor MySQL performance on EC2 Micro instance

I have one small webapp, which uses Pyhon/Flask and a MySQL db for storage of data. I have a studentsdatabase, which has around 3 thousand rows. When trying to load that page, the loading takes very much time, sometimes even a minute or so. Its around 20 seconds, which is really slow and I am wondering what is causing this. This is the state of the server before any request is made, and this happens when I try to load that site.
As I said, this is not too much records, and I am puzzled by why this is so ineffective. I am using Ubuntu 12.04, with Ver 14.14 Distrib 5.5.32, for debian-linux-gnu (x86_64) using readline 6.2 mysql version. Other queries run fine, for example listing students whose name starts with some letter takes around 2-3 seconds, which is acceptable. That shows the portion of the table, so I am guessing something is not optimized right.
My.cnf file is located here. I tried some stuff, added some lines at the bottom, but without too much success.
The actual queries are done by sqlalchemy, and this is the specific code used to load this:
score = db.session.query(Scores.id).order_by(Scores.date.desc()).correlate(Students).filter(Students.email == Scores.email).limit(1)
students = db.session.query(Students, score.as_scalar()).filter_by(archive=0).order_by(Students.exam_date)
return render_template("students.html", students=students.all())
This appears to be the sql generated:
SELECT student.id AS student_id, student.first_name AS student_first_name, student.middle_name AS student_middle_name, student.last_name AS student_last_name, student.email AS student_email, student.password AS student_password, student.address1 AS student_address1, student.address2 AS student_address2, student.city AS student_city, student.state AS student_state, student.zip AS student_zip, student.country AS student_country, student.phone AS student_phone, student.cell_phone AS student_cell_phone, student.active AS student_active, student.archive AS student_archive, student.imported AS student_imported, student.security_pin AS student_security_pin, (SELECT scores.id \nFROM scores \nWHERE student.email = scores.email ORDER BY scores.date DESC \n LIMIT 1) AS anon_1 \nFROM student \nWHERE student.archive = 0"
Thanks in advance for your time and help!

#datasage is right - the micro instance can only do so much. You might try starting a second micro-instance for your mysql database. Running both apache and mysql on a single micro instance will be slow.
From my experience, when using AWS's RDS service (mysql)- you can get reasonable performance on the micro-instance for testing. Depending on how long the instance has been on, sometimes you can get crawlers pinging your site, so it can help to IP restrict it to your computer in the security policy.
It doesn't look like your database structure is that complex - you might add an index on your email fields, but I suspect unless your dataset is over 5000 rows it won't make much difference. If you're using the sqlalchemy ORM, this would look like:
class Scores(base):
__tablename__ = 'center_master'
id = Column(Integer(), primary_key=True)
email = Column(String(255), index=True)

Micro instances are pretty slow performance wise. They are designed with burstable CPU profiles and will be heavily restricted when the burstable time is exceeded.
That said, your problem here is likely with your database design. Any time you want to join two tables, you want to have indexes on columns of the right and left side of the join. In this case you are using the email field.
Using strings to join on isn't quite as optimal as using an integer id. Also using the Explain keyword will running the query directly in mysql will show you an execution plan and can help you quickly identify where you may be missing indexes or have other problems.

Collecting keys vs automatic indexing in Google App Engine

After enabling Appstats and profiling my application, I went on a panic rage trying to figure out how to reduce costs by any means. A lot of my costs per request came from queries, so I sought out to eliminate querying as much as possible.
For example, I had one query where I wanted to get a User's StatusUpdates after a certain date X. I used a query to fetch: statusUpdates = StatusUpdates.query(StatusUpdates.date > X).
So I thought I might outsmart the system and avoid a query, but incur higher write costs for the sake of lower read costs. I thought that every time a user writes a Status, I store the key to that status in a list property of the user. So instead of querying, I would just do ndb.get_multi(user.list_of_status_keys).
The question is, what is the difference for the system between these two approaches? Sure I avoid a query with the second case, but what is happening behind the scenes here? Is what I'm doing in the second case, where I'm collecting keys, just me doing a manual indexing that GAE would have done for me with queries?
In general, what is the difference between get_multi(keys) and a query? Which is more efficient? Which is less costly?

Check the docs on billing:
https://developers.google.com/appengine/docs/billing
It's pretty straightforward. Reads are $0.07/100k, smalls are $0.01/100k, so you want to do smalls.
A query is 1 read + 1 small / entity
A get is 1 read. If you are getting more than 1 entity back with a query, it's cheaper to do a query than reading entities from keys.
Query is likely more efficient too. The only benefit from doing the gets is that they'll be fully consistent (whereas a query is eventually consistent).

Storing the keys does not query, as you cannot do anything with just the keys. You will still have to fetch the Status objects from memory. Also, since you want to query on the date of the Status object, you will need to fetch all the Status objects into memory and compare their dates yourself. If you use a Query, appengine will fetch only the Status with the required date. Since you fetch less, your read costs will be lower.
As this is basically the same question as you have posed here, I suggest that you look at the answer I gave there.

Reverse Search Best Practices?

I'm making an app that has a need for reverse searches. By this, I mean that users of the app will enter search parameters and save them; then, when any new objects get entered onto the system, if they match the existing search parameters that a user has saved, a notification will be sent, etc.
I am having a hard time finding solutions for this type of problem.
I am using Django and thinking of building the searches and pickling them using Q objects as outlined here: http://www.djangozen.com/blog/the-power-of-q
The way I see it, when a new object is entered into the database, I will have to load every single saved query from the db and somehow run it against this one new object to see if it would match that search query... This doesn't seem ideal - has anyone tackled such a problem before?

At the database level, many databases offer 'triggers'.
Another approach is to have timed jobs that periodically fetch all items from the database that have a last-modified date since the last run; then these get filtered and alerts issued. You can perhaps put some of the filtering into the query statement in the database. However, this is a bit trickier if notifications need to be sent if items get deleted.
You can also put triggers manually into the code that submits data to the database, which is perhaps more flexible and certainly doesn't rely on specific features of the database.
A nice way for the triggers and the alerts to communicate is through message queues - queues such as RabbitMQ and other AMQP implementations will scale with your site.

The amount of effort you use to solve this problem is directly related to the number of stored queries you are dealing with.
Over 20 years ago we handled stored queries by treating them as minidocs and indexing them based on all of the must have and may have terms. A new doc's term list was used as a sort of query against this "database of queries" and that built a list of possibly interesting searches to run, and then only those searches were run against the new docs. This may sound convoluted, but when there are more than a few stored queries (say anywhere from 10,000 to 1,000,000 or more) and you have a complex query language that supports a hybrid of Boolean and similarity-based searching, it substantially reduced the number we had to execute as full-on queries -- often no more that 10 or 15 queries.
One thing that helped was that we were in control of the horizontal and the vertical of the whole thing. We used our query parser to build a parse tree and that was used to build the list of must/may have terms we indexed the query under. We warned the customer away from using certain types of wildcards in the stored queries because it could cause an explosion in the number of queries selected.
Update for comment:
Short answer: I don't know for sure.
Longer answer: We were dealing with a custom built text search engine and part of it's query syntax allowed slicing the doc collection in certain ways very efficiently, with special emphasis on date_added. We played a lot of games because we were ingesting 4-10,000,000 new docs a day and running them against up to 1,000,000+ stored queries on a DEC Alphas with 64MB of main memory. (This was in the late 80's/early 90's.)
I'm guessing that filtering on something equivalent to date_added could be done used in combination the date of the last time you ran your queries, or maybe the highest id at last query run time. If you need to re-run the queries against a modified record you could use its id as part of the query.
For me to get any more specific, you're going to have to get a lot more specific about exactly what problem you are trying to solve and the scale of the solution you are trying accomplishing.

If you stored the type(s) of object(s) involved in each stored search as a generic relation, you could add a post-save signal to all involved objects. When the signal fires, it looks up only the searches that involve its object type and runs those. That probably will still run into scaling issues if you have a ton of writes to the db and a lot of saved searches, but it would be a straightforward Django approach.

Why does not postgresql start returning rows immediately?

The following query returns data right away:
SELECT time, value from data order by time limit 100;
Without the limit clause, it takes a long time before the server starts returning rows:
SELECT time, value from data order by time;
I observe this both by using the query tool (psql) and when querying using an API.
Questions/issues:
The amount of work the server has to do before starting to return rows should be the same for both select statements. Correct?
If so, why is there a delay in case 2?
Is there some fundamental RDBMS issue that I do not understand?
Is there a way I can make postgresql start returning result rows to the client without pause, also for case 2?
EDIT (see below). It looks like setFetchSize is the key to solving this. In my case I execute the query from python, using SQLAlchemy. How can I set that option for a single query (executed by session.execute)? I use the psycopg2 driver.
The column time is the primary key, BTW.
EDIT:
I believe this excerpt from the JDBC driver documentation describes the problem and hints at a solution (I still need help - see the last bullet list item above):
By default the driver collects all the results for the query at once. This can be inconvenient for large data sets so the JDBC driver provides a means of basing a ResultSet on a database cursor and only fetching a small number of rows.
and
Changing code to cursor mode is as simple as setting the fetch size of the Statement to the appropriate size. Setting the fetch size back to 0 will cause all rows to be cached (the default behaviour).
// make sure autocommit is off
conn.setAutoCommit(false);
Statement st = conn.createStatement();
// Turn use of the cursor on.
st.setFetchSize(50);

The psycopg2 dbapi driver buffers the whole query result before returning any rows. You'll need to use server side cursor to incrementally fetch results. For SQLAlchemy see server_side_cursors in the docs and if you're using the ORM the Query.yield_per() method.
SQLAlchemy currently doesn't have an option to set that per single query, but there is a ticket with a patch for implementing that.

In theory, because your ORDER BY is by primary key, a sort of the results should not be necessary, and the DB could indeed return data right away in key order.
I would expect a capable DB of noticing this, and optimizing for it. It seems that PGSQL is not. * shrug *
You don't notice any impact if you have LIMIT 100 because it's very quick to pull those 100 results out of the DB, and you won't notice any delay if they're first gathered up and sorted before being shipped out to your client.
I suggest trying to drop the ORDER BY. Chances are, your results will be correctly ordered by time anyway (there may even be a standard or specification that mandates this, given your PK), and you might get your results more quickly.