In the last days I played a bit with riak. The initial setup was easier then I thought. Now I have a 3 node cluster, all nodes running on the same vm for the sake of testing.
I admit, the hardware settings of my virtual machine are very much downgraded (1 CPU, 512 MB RAM) but still I am a quite surprised by the slow performance of riak.
Map Reduce
Playing a bit with map reduce I had around 2000 objects in one bucket, each about 1k - 2k in size as json. I used this map function:
function(value, keyData, arg) {
var data = Riak.mapValuesJson(value)[0];
if (data.displayname.indexOf("max") !== -1) return [data];
return [];
}
And it took over 2 seconds just for performing the http request returning its result, not counting the time it took in my client code to deserialze the results from json. Removing 2 of 3 nodes seemed to slightly improve the performance to just below 2 seconds, but this still seems really slow to me.
Is this to be expected? The objects were not that large in bytesize and 2000 objects in one bucket isnt that much, either.
Insert
Batch inserting of around 60.000 objects in the same size as above took rather long and actually didnt really work.
My script which inserted the objects in riak died at around 40.000 or so and said it couldnt connect to the riak node anymore. In the riak logs I found an error message which indicated that the node ran out of memory and died.
Question
This is really my first shot at riak, so there is definately the chance that I screwed something up.
Are there any settings I could tweak?
Are the hardware settings too constrained?
Maybe the PHP client library I used for interacting with riak is the limiting factor here?
Running all nodes on the same physical machine is rather stupid, but if this is a problem - how can i better test the performance of riak?
Is map reduce really that slow? I read about the performance hit that map reduce has on the riak mailing list, but if Map Reduce is slow, how are you supposed to perform "queries" for data needed nearly in realtime? I know that riak is not as fast as redis.
It would really help me a lot if anyone with more experience in riak could help me out with some of these questions.
This answer is a bit late, but I want to point out that Riak's mapreduce implementation is designed primarily to work with links, not entire buckets.
Riak's internal design is actually pretty much optimized against working with entire buckets. That's because buckets are not considered to be sequential tables but a keyspace distributed across a cluster of nodes. This means that random access is very fast — probably O(log n), but don't quote me on that — whereas serial access is very, very, very slow. Serial access, the way Riak is currently designed, necessarily means asking all nodes for their data.
Incidentally, "buckets" in Riak terminology are, confusingly and disappointingly, not implemented the way you probably think. What Riak calls a bucket is in reality just a namespace. Internally, there is only one bucket, and keys are stored with the bucket name as a prefix. This means that no matter how small or large you bucket is, enumerating the keys in a single bucket of size n will take m time, where m is the total number of keys in all buckets.
These limitations are implementation choices by Basho, not necessarily design flaws. Cassandra implements the exact same partitioning model as Riak, but supports efficient sequential range scans and mapreduce across large amounts of keys. Cassandra also implements true buckets.
A recommendation I'd have now that some time has passed and several new versions of Riak have come about is this. Never rely on full bucket map/reduce, that's not an optimized operation, and chances are very good there are other ways to optimize your map/reduce so you don't have to look through so much data to pull out the singlets you need.
Secondary indices now available in newer versions of Riak are definitely the way to go in this regard. Put an index on the objects you want to find (perhaps named 'ismax_int' with a value of 0 or 1). You can map/reduce a secondary index with hundreds of thousands of keys in microseconds which a full bucket scan would have taken multiple seconds to consider.
I don't have direct experience of Riak, but have worked with Cassandra a little, which is similar.
Firstly, performance will probably depend a lot on the number of cores available, and the memory. These systems are usually heavily pipelined and concurrent and benefit from a lot of cores. 4+ cores and 4GB+ of RAM would be a good starting point.
Secondly, MapReduce is designed for batch processing, not realtime queries.
Riak and all similar Key-Value stores are designed for high write performance, high read performance for simple lookups, no complex querying at all.
Just for comparison, Cassandra on a single node (6 core, 6GB) can do 20,000 individual inserts per second.
Related
Can someone point me to cassandra client code that can achieve a read throughput of at least hundreds of thousands of reads/s if I keep reading the same record (or even a small number of records) over and over? I believe row_cache_size_in_mb is supposed to cache frequently used records in memory, but setting it to say 10MB seems to make no difference.
I tried cassandra-stress of course, but the highest read throughput it achieves with 1KB records (-col size=UNIFORM\(1000..1000\)) is ~15K/s.
With low numbers like above, I can easily write an in-memory hashmap based cache that will give me at least a million reads per second for a small working set size. How do I make cassandra do this automatically for me? Or is it not supposed to achieve performance close to an in-memory map even for a tiny working set size?
Can someone point me to cassandra client code that can achieve a read throughput of at least hundreds of thousands of reads/s if I keep reading the same record (or even a small number of records) over and over?
There are some solution for this scenario
One idea is to use row cache but be careful, any update/delete to a single column will invalidate the whole partition from the cache so you loose all the benefit. Row cache best usage is for small dataset and are frequently read but almost never modified.
Are you sure that your cassandra-stress scenario never update or write to the same partition over and over again ?
Here are my findings: when I enable row_cache, counter_cache, and key_cache all to sizable values, I am able to verify using "top" that cassandra does no disk I/O at all; all three seem necessary to ensure no disk activity. Yet, despite zero disk I/O, the throughput is <20K/s even for reading a single record over and over. This likely confirms (as also alluded to in my comment) that cassandra incurs the cost of serialization and deserialization even if its operations are completely in-memory, i.e., it is not designed to compete with native hashmap performance. So, if you want get native hashmap speeds for a small-working-set workload but expand to disk if the map grows big, you would need to write your own cache on top of cassandra (or any of the other key-value stores like mongo, redis, etc. for that matter).
For those interested, I also verified that redis is the fastest among cassandra, mongo, and redis for a simple get/put small-working-set workload, but even redis gets at best ~35K/s read throughput (largely independent, by design, of the request size), which hardly comes anywhere close to native hashmap performance that simply returns pointers and can do so comfortably at over 2 million/s.
I was playing around with cassandra-stress tool on my own laptop (8 cores, 16GB) with Cassandra 2.2.3 installed out of the box with having its stock configuration. I was doing exactly what was described here:
http://www.datastax.com/dev/blog/improved-cassandra-2-1-stress-tool-benchmark-any-schema
And measuring its insert performance.
My observations were:
using the code from https://gist.github.com/tjake/fb166a659e8fe4c8d4a3 without any modifications I had ~7000 inserts/sec.
when modifying line 35 in the code above (cluster: fixed(1000)) to "cluster: fixed(100)", i. e. configuring my test data distribution to have 100 clustering keys instead of 1000, the performance was jumping up to ~11000 inserts/sec
when configuring it to have 5000 clustering keys per partition, the performance was reducing to just 700 inserts/sec
The documentation says however Cassandra can support up to 2 billion rows per partition. I don't need that much still I don't get how just 5000 records per partition can slow the writes 10 times down or am I missing something?
Supporting is a little different from "best performaning". You can have very wide partitions, but the rule-of-thumb is to try to keep them under 100mb for misc performance reasons. Some operations can be performed more efficiently when the entirety of the partition can be stored in memory.
As an example (this is old example, this is a complete non issue post 2.0 where everything is single pass) but in some versions when the size is >64mb compaction has a two pass process, that halves compaction throughput. It still worked with huge partitions. I've seen many multi gb ones that worked just fine. but the systems with huge partitions were difficult to work with operationally (managing compactions/repairs/gcs).
I would say target the rule of thumb initially of 100mb and test from there to find own optimal. Things will always behave differently based on use case, to get the most out of a node the best you can do is some benchmarks closest to what your gonna do (true of all systems). This seems like something your already doing so your definitely on the right path.
I know that a big part of the performance from Couchbase comes from serving in-memory documents and for many of my data types that seems like an entirely reasonable aspiration but considering how user-data scales and is used I'm wondering if it's reasonable to plan for only a small percentage of the user documents to be in memory all of the time. I'm thinking maybe only 10-15% at any given time. Is this a reasonable assumption considering:
At any given time period there will be a only a fractional number of users will be using the system.
In this case, users only access there own data (or predominantly so)
Recently entered data is exponentially more likely to be viewed than historical user documents
UPDATE:
Some additional context:
Let's assume there's a user base of a 1 million customers, that 20% rarely if ever access the site, 40% access it once a week, and 40% access it every day.
At any given moment, only 5-10% of the user population would be logged in
When a user logs in they are like to re-query for certain documents in a single session (although the client does do some object caching to minimise this)
For any user, the most recent records are very active, the very old records very inactive
In summary, I would say of a majority of user-triggered transactional documents are queried quite infrequently but there are a core set -- records produced in the last 24-48 hours and relevant to the currently "logged in" group -- that would have significant benefits to being in-memory.
Two sub-questions are:
Is there a way to indicate a timestamp on a per-document basis to indicate it's need to be kept in memory?
How does couchbase overcome the growing list of document id's in-memory. It is my understanding that all ID's must always be in memory? isn't this too memory intensive for some apps?
First,one of the major benefits to CB is the fact that it is spread across multiple nodes. This also means your queries are spread across multiple nodes and you have a performance gain as a result (I know several other similar nosql spread across nodes - so maybe not relevant for your comparison?).
Next, I believe this question is a little bit too broad as I believe the answer will really depend on your usage. Does a given user only query his data one time, at random? If so, then according to you there will only be an in-memory benefit 10-15% of the time. If instead, once a user is on the site, they might query their data multiple times, there is a definite performance benefit.
Regardless, Couchbase has pretty fast disk-access performance, particularly on SSDs, so it probably doesn't make much difference either way, but again without specifics there is no way to be sure. If it's a relatively small document size, and if it involves a user waiting for one of them to load, then the user certainly will not notice a difference whether the document is loaded from RAM or disk.
Here is an interesting article on benchmarks for CB against similar nosql platforms.
Edit:
After reading your additional context, I think your scenario lines up pretty much exactly how Couchbase was designed to operate. From an eviction standpoint, CB keeps the newest and most-frequently accessed items in RAM. As RAM fills up with new and/or old items, oldest and least-frequently accessed are "evicted" to disk. This link from the Couchbase Manual explains more about how this works.
I think you are on the right track with Couchbase - in any regard, it's flexibility with scaling will easily allow you to tune the database to your application. I really don't think you can go wrong here.
Regarding your two questions:
Not in Couchbase 2.2
You should use relatively small document IDs. While it is true they are stored in RAM, if your document ids are small, your deployment is not "right-sized" if you are using a significant percentage of the available cluster RAM to store keys. This link talks about keys and gives details relevant to key size (e.g. 250-byte limit on size, metadata, etc.).
Basically what you are making a decision point on is sizing the Couchbase cluster for bucket RAM, and allowing a reduced residency ratio (% of document values in RAM), and using Cache Misses to pull from disk.
However, there are caveats in this scenario as well. You will basically also have relatively constant "cache eviction" where "not recently used" values are being removed from RAM cache as you pull cache missed documents from disk into RAM. This is because you will always be floating at the high water mark for the Bucket RAM quota. If you also simultaneously have a high write velocity (new/updated data) they will also need to be persisted. These two processes can compete for Disk I/O if the write velocity exceeds your capacity to evict/retrieve, and your SDK client will receive a Temporary OOM error if you actually cannot evict fast enough to open up RAM for new writes. As you scale horizontally, this becomes less likely as you have more Disk I/O capacity spread across more machines all simultaneously doing this process.
If when you say "queried" you mean querying indexes (i.e. Views), this is a separate data structure on disk that you would be querying and of course getting results back is not subject to eviction/NRU, but if you follow the View Query with a multi-get the above still applies. (Don't emit entire documents into your Index!)
I have a cluster application, which is divided into a controller and a bunch of workers. The controller runs on a dedicated host, the workers phone in over the network and get handed jobs, so far so normal. (Basically the "divide-and-conquer pipeline" from the zeromq manual, with job-specific wrinkles. That's not important right now.)
The controller's core data structure is unordered_map<string, queue<string>> in pseudo-C++ (the controller is actually implemented in Python, but I am open to the possibility of rewriting it in something else). The strings in the queues define jobs, and the keys of the map are a categorization of the jobs. The controller is seeded with a set of jobs; when a worker starts up, the controller removes one string from one of the queues and hands it out as the worker's first job. The worker may crash during the run, in which case the job gets put back on the appropriate queue (there is an ancillary table of outstanding jobs). If it completes the job successfully, it will send back a list of new job-strings, which the controller will sort into the appropriate queues. Then it will pull another string off some queue and send it to the worker as its next job; usually, but not always, it will pick the same queue as the previous job for that worker.
Now, the question. This data structure currently sits entirely in main memory, which was fine for small-scale test runs, but at full scale is eating all available RAM on the controller, all by itself. And the controller has several other tasks to accomplish, so that's no good.
What approach should I take? So far, I have considered:
a) to convert this to a primarily-on-disk data structure. It could be cached in RAM to some extent for efficiency, but jobs take tens of seconds to complete, so it's okay if it's not that efficient,
b) using a relational database - e.g. SQLite, (but SQL schemas are a very poor fit AFAICT),
c) using a NoSQL database with persistency support, e.g. Redis (data structure maps over trivially, but this still appears very RAM-centric to make me feel confident that the memory-hog problem will actually go away)
Concrete numbers: For a full-scale run, there will be between one and ten million keys in the hash, and less than 100 entries in each queue. String length varies wildly but is unlikely to be more than 250-ish bytes. So, a hypothetical (impossible) zero-overhead data structure would require 234 – 237 bytes of storage.
Ultimately, it all boils down on how you define efficiency needed on part of the controller -- e.g. response times, throughput, memory consumption, disk consumption, scalability... These properties are directly or indirectly related to:
number of requests the controller needs to handle per second (throughput)
acceptable response times
future growth expectations
From your options, here's how I'd evaluate each option:
a) to convert this to a primarily-on-disk data structure. It could be
cached in RAM to some extent for efficiency, but jobs take tens of
seconds to complete, so it's okay if it's not that efficient,
Given the current memory hog requirement, some form of persistent storage seems a reaonsable choice. Caching comes into play if there is a repeatable access pattern, say the same queue is accessed over and over again -- otherwise, caching is likely not to help.
This option makes sense if 1) you cannot find a database that maps trivially to your data structure (unlikely), 2) for some other reason you want to have your own on-disk format, e.g. you find that converting to a database is too much overhead (again, unlikely).
One alternative to databases is to look at persistent queues (e.g. using a RabbitMQ backing store), but I'm not sure what the per-queue or overall size limits are.
b) using a relational database - e.g. SQLite, (but SQL schemas are a
very poor fit AFAICT),
As you mention, SQL is probably not a good fit for your requirements, even though you could surely map your data structure to a relational model somehow.
However, NoSQL databases like MongoDB or CouchDB seem much more appropriate. Either way, a database of some sort seems viable as long as they can meet your throughput requirement. Many if not most NoSQL databases are also a good choice from a scalability perspective, as they include support for sharding data across multiple machines.
c) using a NoSQL database with persistency support, e.g. Redis (data
structure maps over trivially, but this still appears very RAM-centric
to make me feel confident that the memory-hog problem will actually go
away)
An in-memory database like Redis doesn't solve the memory hog problem, unless you set up a cluster of machines that each holds a part of the overall data. This makes sense only if keeping all data in-memory is needed due to low response times requirements. Yet, given the nature of your jobs, taking tens of seconds to complete, response times, respective to workers, hardly matter.
If you find, however, that response times do matter, Redis would be a good choice, as it handles partitioning trivially using either client-side consistent-hashing or at the cluster level, thus also supporting scalability scenarios.
In any case
Before you choose a solution, be sure to clarify your requirements. You mention you want an efficient solution. Since efficiency can only be gauged against some set of requirements, here's the list of questions I would try to answer first:
*Requirements
how many jobs are expected to complete, say per minute or per hour?
how many workers are needed to do so?
concluding from that:
what is the expected load in requestes/per second, and
what response times are expected on part of the controller (handing out jobs, receiving results)?
And looking into the future:
will the workload increase, i.e. does your solution need to scale up (more jobs per time unit, more more data per job?)
will there be a need for persistency of jobs and results, e.g. for auditing purposes?
Again, concluding from that,
how will this influence the number of workers?
what effect will it have on the number of requests/second on part of the controller?
With these answers, you will find yourself in a better position to choose a solution.
I would look into a message queue like RabbitMQ. This way it will first fill up the RAM and then use the disk. I have up to 500,000,000 objects in queues on a single server and it's just plugging away.
RabbitMQ works on Windows and Linux and has simple connectors/SDKs to about any kind of language.
https://www.rabbitmq.com/
I had always thought that Mongo had excellent performance with it's mapreduce functionality, but am now reading that it is a slow implementation of it. So if I had to pick an alternative to benchmark against, what should it be?
My software will be such that users will often have millions of records, and often be sorting and crunching through unpredictable subsets that are 10s or 100s of thousands. Most of the analysis of data that uses the full millions of records can be done in summary tables and the like. I'd originally thought Hypertable was a viable alternative, but in doing research I saw in their documents their mention that Mongo would be a more performant option, while Hypertable had other benefits. But for my application speed is my number one initial priority.
First of all, it's important to decide on what is "fast enough". Undoubtedly there are faster solutions than MongoDB's map/reduce but in most cases you may be looking at significantly higher development cost.
That said MongoDB's map/reduce runs, at time of writing, on a single thread which means it will not utilize all the cpu available to it. Also, MongoDB has very little in the way of native aggregation functionality. This will change fixed with version 2.1 onwards that should improve performance though (see https://jira.mongodb.org/browse/SERVER-447 and http://www.slideshare.net/cwestin63/mongodb-aggregation-mongosf-may-2011).
Now, what MongoDB is good at is scaling up easily, especially when it comes to reads. And this is important because the best solution for number crunching on large datasets is definitely a map/reduce cloud like Augusto suggested. Let such an m/r do the number crunching while MongoDB makes the required data available at high speeds. Database query throughput too low is easily solved by adding more mongo shards. Number crunching/aggregation performance too slow is solved by adding more m/r boxes. Basically performance becomes a function of number of instances you reserve for the problem, and thus cost.