I'm storing the output of a sensor pod in a Postgres db. There are many sensor pods (dozens), each generating a reading every 100ms, 24 hours per day - that's 86x,xxx records per pod per day. The sensor pod is relatively dumb, and lives in an environment of unreliable connectivity, so it produces n(tbd) line CSV files locally, and then ships them all off when it has network access.
The sensor pod knows its own name, and the data it produces but that's about it. I'm trying to decide how to load the data efficiently into the DB. There are two options I'm considering:
Use the COPY syntax, and give it the file directly
Do a mass insert
Can someone speak to the performance ramifications of both of those?
My hesitation with option 1 is that I need to supply some ancillary data (some foreign keys, etc.) that won't be in the file. The only way to do that, without making the sensor pod aware of the random crap that's also going into that table is to generate temporary tables for the load, and then move them into their final destination with an insert .. select which seems wasteful.
This is a high write / low read environment.
Related
My application produces rotating log files containing multiple application metrics. The log file is rotated once a minute, but each file is still relatively large (over 30MB, with 100ks of rows)
I'd like to feed the logs into PipelineDB (running on the same single machine) which Countiuous View can create for me exactly the aggregations I need over the metrics.
I can easily ship the logs to PipelineDB using copy from stdin, which works great.
However, a machine might occasionally power off unexpectedly (e.g. due to power shortage) during the copy of a log file. Which means that once back online there is uncertainty how much of the file has been inserted into PipelineDB.
How could I ensure that each row in my logs is inserted exactly once in such cases? (It's very important that I get complete and accurate aggregations)
Notice each row in the log file has a unique identifier (serial number created by my application), but I can't find in the docs the option to define a unique field in the stream. I assume that PipelineDB's design is not meant to handle unique fields in stream rows
Nonetheless, are there any alternative solutions to this issue?
Exactly once semantics in a streaming (infinite rows) context is a very complex problem. Most large PipelineDB deployments use some kind of message bus infrastructure (e.g. Kafka) in front of PipelineDB for delivery semantics and reliability, as that's not PipelineDB's core focus.
That being said, there are a couple of approaches you could use here that may be worth thinking about.
First, you could maintain a regular table in PipelineDB that keeps track of each logfile and the line number that it has successfully written to PipelineDB. When beginning to ship a new logfile, check it against this table to determine which line number to start at.
Secondly, you could separate your aggregations by logfile (by including a path or something in the grouping) and simply DELETE any existing rows for that logfile before sending it. Then use combine to aggregate over all logfiles at read time, possibly with a VIEW.
I'm using MongoDB with approximately 4 million documents and around 5-6GB database size. The machine has 10GB of RAM, and free only reports around 3.7GB in use. The database is used for a video game related ladder (rankings) website, separated by region.
It's a fairly write heavy operation, but still gets a significant number of reads as well. We use an updater which queries an outside source every hour or two. This updater then processes the records and updates documents on the database. The updater only processes one region at a time (see previous paragraph), so approximately 33% of the database is updated.
When the updater runs, and for the duration that it runs, the average flush time spikes up to around 35-40 seconds, and we experience general slowdowns with other queries. The updater is RAN on a SEPARATE MACHINE and only queries MongoDB at the end, when all the data has been retrieved and processed from the third party.
Some people have suggested slowing down the number of updates, or only updating players who have changed, but the problem comes down to rankings. Since we support ties between players, we need to pre-calculate the ranks - so if only a few users have actually changed ranks, we still need to update the rest of the users ranks accordingly. At least, that was the case with MySQL - I'm not sure if there is a good solution with MongoDB for ranking ~800K->1.2 million documents while supporting ties.
My question is: how can we improve the flush and slowdown we're experiencing? Why is it spiking so high? Would disabling journaling (to take some load off the i/o) help, as data loss isn't something I'm worried about as the database is updated frequently regardless?
Server status: http://pastebin.com/w1ETfPWs
You are using the wrong tool for the job. MongoDB isn't designed for ranking large ladders in real time, at least not quickly.
Use something like Redis, Redis have something called a "Sorted List" designed just for this job, with it you can have 100 millions entries and still fetch the 5000000th to 5001000th at sub millisecond speed.
From the official site (Redis - Sorted sets):
Sorted sets
With sorted sets you can add, remove, or update elements in a very fast way (in a time proportional to the logarithm of the number of elements). Since elements are taken in order and not ordered afterwards, you can also get ranges by score or by rank (position) in a very fast way. Accessing the middle of a sorted set is also very fast, so you can use Sorted Sets as a smart list of non repeating elements where you can quickly access everything you need: elements in order, fast existence test, fast access to elements in the middle!
In short with sorted sets you can do a lot of tasks with great performance that are really hard to model in other kind of databases.
With Sorted Sets you can:
Take a leader board in a massive online game, where every time a new score is submitted you update it using ZADD. You can easily take the top users using ZRANGE, you can also, given an user name, return its rank in the listing using ZRANK. Using ZRANK and ZRANGE together you can show users with a score similar to a given user. All very quickly.
Sorted Sets are often used in order to index data that is stored inside Redis. For instance if you have many hashes representing users, you can use a sorted set with elements having the age of the user as the score and the ID of the user as the value. So using ZRANGEBYSCORE it will be trivial and fast to retrieve all the users with a given interval of ages.
Sorted Sets are probably the most advanced Redis data types, so take some time to check the full list of Sorted Set commands to discover what you can do with Redis!
Without seeing any disk statistics, I am of the opinion that you are saturating your disks.
This can be checked with iostat -xmt 2, and checking the %util column.
Please don't disable journalling - you will only cause more issues later down the line when your machine crashes.
Separating collections will have no effect. Separating databases may, but if you're IO bound, this will do nothing to help you.
Options
If I am correct, and your disks are saturated, adding more disks in a RAID 10 configuration will vastly help performance and durability - more so if you separate the journal off to an SSD.
Assuming that this machine is a single server, you can setup a replicaset and send your read queries there. This should help you a fair bit, but not as much as the disks.
Let's say that I'm building a stack exchange clone, and every time I examine a question, I also load each and every answer. The table might look like:
id integer
question_id FOREIGN KEY
answer bool
date timestamp
How can I tell django to tell postgres to keep all the answers together for fast access? Postgres has the underlying feature CLUSTER USING.
(CLUSTER USING is 'defragmenting' feature for tables. This works especially well for small records, since they may all end up in the same disk block and greatly reduce load time. The defragmenting is typically done as a batch job at times of low load).
As far as I can tell, you can't. But you can treat this as a database administration task, and do it from the psql command line:
# CLUSTER table USING index_name;
# ANALYZE VERBOSE table;
# CLUSTER VERBOSE;
This will be remembered. Each time you run CLUSTER VERBOSE it will lock all the tables and sort the data. All your answers (in the example above) will be gathered together on disk. This makes sense even for solid state storage, since the eventual database read will cover fewer sectors, meaning fewer I/O operations to retrieve the group.
Obviously you must pick your index well: the wrong choice can scatter the data you actually access. The performance benefit is the best for sparse datasets, and becomes less relevant if most everything is frequently accessed.
A better name for the CLUSTER feature might be "DEFRAG", as this is an operation analogous defragmenting a filesystem.
We're still evaluating Cassandra for our data store. As a very simple test, I inserted a value for 4 columns into the Keyspace1/Standard1 column family on my local machine amounting to about 100 bytes of data. Then I read it back as fast as I could by row key. I can read it back at 160,000/second. Great.
Then I put in a million similar records all with keys in the form of X.Y where X in (1..10) and Y in (1..100,000) and I queried for a random record. Performance fell to 26,000 queries per second. This is still well above the number of queries we need to support (about 1,500/sec)
Finally I put ten million records in from 1.1 up through 10.1000000 and randomly queried for one of the 10 million records. Performance is abysmal at 60 queries per second and my disk is thrashing around like crazy.
I also verified that if I ask for a subset of the data, say the 1,000 records between 3,000,000 and 3,001,000, it returns slowly at first and then as they cache, it speeds right up to 20,000 queries per second and my disk stops going crazy.
I've read all over that people are storing billions of records in Cassandra and fetching them at 5-6k per second, but I can't get anywhere near that with only 10mil records. Any idea what I'm doing wrong? Is there some setting I need to change from the defaults? I'm on an overclocked Core i7 box with 6gigs of ram so I don't think it's the machine.
Here's my code to fetch records which I'm spawning into 8 threads to ask for one value from one column via row key:
ColumnPath cp = new ColumnPath();
cp.Column_family = "Standard1";
cp.Column = utf8Encoding.GetBytes("site");
string key = (1+sRand.Next(9)) + "." + (1+sRand.Next(1000000));
ColumnOrSuperColumn logline = client.get("Keyspace1", key, cp, ConsistencyLevel.ONE);
Thanks for any insights
purely random reads is about worst-case behavior for the caching that your OS (and Cassandra if you set up key or row cache) tries to do.
if you look at contrib/py_stress in the Cassandra source distribution, it has a configurable stdev to perform random reads but with some keys hotter than others. this will be more representative of most real-world workloads.
Add more Cassandra nodes and give them lots of memory (-Xms / -Xmx). The more Cassandra instances you have, the data will be partitioned across the nodes and much more likely to be in memory or more easily accessed from disk. You'll be very limited with trying to scale a single workstation class CPU. Also, check the default -Xms/-Xmx setting. I think the default is 1GB.
It looks like you haven't got enough RAM to store all the records in memory.
If you swap to disk then you are in trouble, and performance is expected to drop significantly, especially if you are random reading.
You could also try benchmarking some other popular alternatives, like Redis or VoltDB.
VoltDB can certainly handle this level of read performance as well as writes and operates using a cluster of servers. As an in-memory solution you need to build a large enough cluster to hold all of your data in RAM.
Background:
I have a PostgreSQL (v8.3) database that is heavily optimized for OLTP.
I need to extract data from it on a semi real-time basis (some-one is bound to ask what semi real-time means and the answer is as frequently as I reasonably can but I will be pragmatic, as a benchmark lets say we are hoping for every 15min) and feed it into a data-warehouse.
How much data? At peak times we are talking approx 80-100k rows per min hitting the OLTP side, off-peak this will drop significantly to 15-20k. The most frequently updated rows are ~64 bytes each but there are various tables etc so the data is quite diverse and can range up to 4000 bytes per row. The OLTP is active 24x5.5.
Best Solution?
From what I can piece together the most practical solution is as follows:
Create a TRIGGER to write all DML activity to a rotating CSV log file
Perform whatever transformations are required
Use the native DW data pump tool to efficiently pump the transformed CSV into the DW
Why this approach?
TRIGGERS allow selective tables to be targeted rather than being system wide + output is configurable (i.e. into a CSV) and are relatively easy to write and deploy. SLONY uses similar approach and overhead is acceptable
CSV easy and fast to transform
Easy to pump CSV into the DW
Alternatives considered ....
Using native logging (http://www.postgresql.org/docs/8.3/static/runtime-config-logging.html). Problem with this is it looked very verbose relative to what I needed and was a little trickier to parse and transform. However it could be faster as I presume there is less overhead compared to a TRIGGER. Certainly it would make the admin easier as it is system wide but again, I don't need some of the tables (some are used for persistent storage of JMS messages which I do not want to log)
Querying the data directly via an ETL tool such as Talend and pumping it into the DW ... problem is the OLTP schema would need tweaked to support this and that has many negative side-effects
Using a tweaked/hacked SLONY - SLONY does a good job of logging and migrating changes to a slave so the conceptual framework is there but the proposed solution just seems easier and cleaner
Using the WAL
Has anyone done this before? Want to share your thoughts?
Assuming that your tables of interest have (or can be augmented with) a unique, indexed, sequential key, then you will get much much better value out of simply issuing SELECT ... FROM table ... WHERE key > :last_max_key with output to a file, where last_max_key is the last key value from the last extraction (0 if first extraction.) This incremental, decoupled approach avoids introducing trigger latency in the insertion datapath (be it custom triggers or modified Slony), and depending on your setup could scale better with number of CPUs etc. (However, if you also have to track UPDATEs, and the sequential key was added by you, then your UPDATE statements should SET the key column to NULL so it gets a new value and gets picked by the next extraction. You would not be able to track DELETEs without a trigger.) Is this what you had in mind when you mentioned Talend?
I would not use the logging facility unless you cannot implement the solution above; logging most likely involves locking overhead to ensure log lines are written sequentially and do not overlap/overwrite each other when multiple backends write to the log (check the Postgres source.) The locking overhead may not be catastrophic, but you can do without it if you can use the incremental SELECT alternative. Moreover, statement logging would drown out any useful WARNING or ERROR messages, and the parsing itself will not be instantaneous.
Unless you are willing to parse WALs (including transaction state tracking, and being ready to rewrite the code everytime you upgrade Postgres) I would not necessarily use the WALs either -- that is, unless you have the extra hardware available, in which case you could ship WALs to another machine for extraction (on the second machine you can use triggers shamelessly -- or even statement logging -- since whatever happens there does not affect INSERT/UPDATE/DELETE performance on the primary machine.) Note that performance-wise (on the primary machine), unless you can write the logs to a SAN, you'd get a comparable performance hit (in terms of thrashing filesystem cache, mostly) from shipping WALs to a different machine as from running the incremental SELECT.
if you can think of a 'checksum table' that contains only the id's and the 'checksum' you can not only do a quick select of the new records but also the changed and deleted records.
the checksum could be a crc32 checksum function you like.
The new ON CONFLICT clause in PostgreSQL has changed the way I do many updates. I pull the new data (based on a row_update_timestamp) into a temp table then in one SQL statement INSERT into the target table with ON CONFLICT UPDATE. If your target table is partitioned then you need to jump through a couple of hoops (i.e. hit the partition table directly). The ETL can happen as you load the the Temp table (most likely) or in the ON CONFLICT SQL (if trivial). Compared to to other "UPSERT" systems (Update, insert if zero rows etc.) this shows a huge speed improvement. In our particular DW environment we don't need/want to accommodate DELETEs. Check out the ON CONFLICT docs - it gives Oracle's MERGE a run for it's money!