Using HDFS for Nuodb as storage. Would this have a performance impact?
If I understand correctly, HDFS is better suited for batch mode or write once and read many times, types of application. Would it not increase the latency for record to be fetch in case it needs to read from storage?
On top of that HDFS block size concept, keep the file size small that would increase the network traffic while data is being fetch. Am I missing something here? Please point out the same.
How would Nuodb manage these kind of latency gotchas?
Good afternoon,
My name is Elisabete and I am the Technical Support Engineer over at NuoDB. I believe that I may have just answered this via your post on our own forum, but I'm responding here as well for anyone else who's curious.
First... a mini lesson on NuoDB architecture/layout:
The most basic NuoDB set-up includes:
Broker Agent
Transaction Engine (TE)
Storage Manager (SM) connected to an Archive Directory
Broker Agents keep track of all the moving parts in the domain (collection of machines hosting NuoDB processes) and provide client applications with connection information for the next available Transaction Engine.
Transaction Engines process incoming SQL requests and manage transactions.
Storage Managers read and write data to and from "disk" (Archive Directory)
All of these components can reside on a single machine, but an optimal set up would have them spread across multiple host machines (allowing each process to take full advantage of the host's available CPU/RAM). Also, while it's possible to run with just one of each component, this is a case where more is definitely more. Additional Brokers provide resiliency, additional TE's increase performance/speed and additional SM's ensure durability.
Ok, so now lets talk about Storage:
This is the "Archive Directory" that your storage manager is writing to. Currently, we support three modes of storage:
Local Files System
Amazon Web Services: Simple Storage volume (S3), Elastic Block Storage (EBS)
Hadoop Distributed Files System (HDFS)
So, to elaborate on how NuoDB works with HDFS... it doesn't know about the multiple machines that the HDFS layer is writing to. As far as the SM is concerned, it is reading and writing data atoms to a single directory. The HDFS layer decides how to then distribute and retrieve data to and from the cluster of machines it resides over.
And now to finally address the question of latency:
Here's the thing, whenever we introduce a remote storage device, we inevitably introduce some amount of additional latency because the SM now has further to go when reading/writing atoms to/from memory. HDFS likely adds a bit more, because now it needs to do it's magic divvying up, distributing, retrieving and reassembling data. Add to that discrepancy in network speed, etc.
I imagine that the gained disk space outweighs the cost in travel time, but this is something you'd have to decide on a case by case basis.
Now, all of that said... I haven't mentioned that TE and SM's both have the ability to cache data to local memory. The size of this cache is something you can set, when starting up each process. NuoDB uses a combination of Multi-Version Concurrency Control (MVCC) and a near constant stream of communication between all of the processes, to ensure that data held in cache is kept up to date with all of the changes happening within the system. Garbage Collection also kicks in and clears out atoms in a Least Recently Used order, when the cache grows close to hitting its limit.
All of this helps reduce latency, because the TE's can hold onto the data they reference most often and grab copies of data they don't have from sibling TE's. When they do resort to asking the SM's for data, there's a chance that the SM (or one of its sibling SM's) has a copy of the requested data in local cache, saving itself the trip out to the Archive Directory.
Whew.. that was a lot and I absolutely glossed over more than a few concepts. These topics are covered in greater depth via the new suite of white papers (and the new "green book") available on our main website. I'm currently also working on some visual guides, to help explain all of this.
If you'd like to know more about NuoDB or if I didn't quite answer your question.... please reach out to me directly via the NuoDB Community Forums (I respond to posts there, a bit faster).
Thank you,
Elisabete
Technical Support Engineer at NuoDB
Related
Recently in a system design interview I was asked a question where cities were divided into zones and data of around 100 zones was available. An api took the zoneid as input and returned all the restaurants for that zone in response. The response time for the api was 50ms so the zone data was kept in memory to avoid delays.
If the zone data is approximately 25GB, then if the service is scaled to say 5 instances, it would need 125GB ram.
Now the requirement is to run 5 instances but use only 25 GB ram with the data split between instances.
I believe to achieve this we would need a second application which would act as a config manager to manage which instance holds which zone data. The instances can get which zones to track on startup from the config manager service. But the thing I am not able to figure out is how we redirect the request for a zone to the correct instance which holds its data especially if we use kubernetes. Also if the instance holding partial data restarts then how do we track which zone data it was holding
Splitting dataset over several nodes: sounds like sharding.
In-memory: the interviewer might be asking about redis or something similar.
Maybe this: https://redis.io/topics/partitioning#different-implementations-of-partitioning
Redis cluster might fit -- keep in mind that when the docs mention "client-side partitioning": the client is some redis client library, loaded by your backends, responding to HTTP client/end-user requests
Answering your comment: then, I'm not sure what they were looking for.
Comparing Java hashmaps to a redis cluster isn't completely fair, considering one is bound to your JVM, while the other is actually distributed / sharded, implying at least inter-process communications and most likely network/non-local queries.
Then again, if the question is to scale an ever-growing JVM: at some point, we need to address the elephant in the room: how do you guarantee data consistency, proper replication/sharding, what do you do when a member goes down, ...?
Distributed hashmap, using Hazelcast, may be more relevant. Some (hazelcast) would make the argument it is safer under heavy write load. Others that migrating from Hazelcast to Redis helped them improve service reliability. I don't have enough background in Java myself, I wouldn't know.
As a general rule: when asked about Java, you could argue that speed and reliability very much rely on your developers understanding of what they're doing. Which, in Java, implies a large margin of error. While we could suppose: if they're asking such questions, they probably have some good devs on their payroll.
Whereas distributed databases (in-memory, on disk, SQL or noSQL), ... is quite a complicated topic, that you would need to master (on top of java), to get it right.
The broad approach they're describing was described by Adya in 2019 as a LInK store. Linked In-memory Key-value stores allow for application objects supporting rich operations to be sharded across a cluster of instances.
I would tend to approach this by implementing a stateful application using Akka (disclaimer: I am at this writing employed by Lightbend, which employs the majority of the developers of Akka and offers support and consulting services to clients using Akka; as my SO history indicates, I would have the same approach even multiple years before I was employed by Lightbend) along these lines.
Akka Cluster to allow a set of JVMs running an application to form a cluster in a peer-to-peer manner and manage/track changes in the membership (including detecting instances which have crashed or are isolated by a network partition)
Akka Cluster Sharding to allow stateful objects keyed by ID to be distributed approximately evenly across a cluster and rebalanced in response to membership changes
These stateful objects are implemented as actors: they can update their state in response to messages and (since they process messages one at a time) without needing elaborate synchronization.
Cluster sharding implies that the actor responsible for an ID might exist on different instances, so that implies some persistence of the state of the zone outside of the cluster. For simplicity*, when an actor responsible for a given zone starts, it initializes itself from datastore (could be S3, could be Dynamo or Cassandra or whatever): after this its state is in memory so reads can be served directly from the actor's state instead of going to an underlying datastore.
By directing all writes through cluster sharding, the in-memory representation is, by definition, kept in sync with the writes. To some extent, we can say that the application is the cache: the backing datastore only exists to allow the cache to survive operational issues (and because it's only in response to issues of that sort that the datastore needs to be read, we can optimize the data store for writes vs. reads).
Cluster sharding relies on a conflict-free replicated data type (CRDT) to broadcast changes in the shard allocation to the nodes of the cluster. This allows, for instance, any instance to handle an HTTP request for any shard: it simply forwards a representation of the important parts of the request as a message to the shard which will distribute it to the correct actor.
From Kubernetes' perspective, the instances are stateless: no StatefulSet or similar is needed. The pods can query the Kubernetes API to find the other pods and attempt to join the cluster.
*: I have a fairly strong prior that event sourcing would be a better persistence approach, but I'll set that aside for now.
In Dask distributed documentation, they have the following information:
For example Dask developers use this ability to build in data locality
when we communicate to data-local storage systems like the Hadoop File
System. When users use high-level functions like
dask.dataframe.read_csv('hdfs:///path/to/files.*.csv') Dask talks to
the HDFS name node, finds the locations of all of the blocks of data,
and sends that information to the scheduler so that it can make
smarter decisions and improve load times for users.
However, it seems that the get_block_locations() was removed from the HDFS fs backend, so my question is: what is the current state of Dask regarding to HDFS ? Is it sending computation to nodes where data is local ? Is it optimizing the scheduler to take into account data locality on HDFS ?
Quite right, with the appearance of arrow's HDFS interface, which is now preferred over hdfs3, the consideration of block locations is no longer part of workloads accessing HDFS, since arrow's implementation doesn't include the get_block_locations() method.
However, we already wanted to remove the somewhat convoluted code which made this work, because we found that the inter-node bandwidth on test HDFS deployments was perfectly adequate that it made little practical difference in most workloads. The extra constrains on the size of the blocks versus the size of the partitions you would like in-memory created an additional layer of complexity.
By removing the specialised code, we could avoid the very special case that was being made for HDFS as opposed to external cloud storage (s3, gcs, azure) where it didn't matter which worker accessed which part of the data.
In short, yes the docs should be updated.
My use case is as follow:
We have about 500 servers running in an autoscaling EC2 cluster that need to access the same configuration data (layed out in a key/value fashion) several million times per second.
The configuration data isn't very large (1 or 2 GBs) and doesn't change much (a few dozen updates/deletes/inserts per minute during peak time).
Latency is critical for us, so the data needs to be replicated and kept in memory on every single instance running our application.
Eventual consistency is fine. However we need to make sure that every update will be propagated at some point. (knowing that the servers can be shutdown at any time)
The update propagation across the servers should be reliable and easy to setup (we can't have static IPs for our servers, or we don't wanna go the route of "faking" multicast on AWS etc...)
Here are the solutions we've explored in the past:
Using regular java maps and use our custom built system to propagate updates across the cluster. (obviously, it doesn't scale that well)
Using EhCache and its replication feature. But setting it up on EC2 is very painful and somehow unreliable.
Here are the solutions we're thinking of trying out:
Apache Ignite (https://ignite.apache.org/) with a REPLICATED strategy.
Hazelcast's Replicated Map feature. (http://docs.hazelcast.org/docs/latest/manual/html-single/index.html#replicated-map)
Apache Geode on every application node. (http://geode.apache.org/)
I would like to know if each of those solutions would work for our use case. And eventually, what issues I'm likely to face with each of them.
Here is what I found so far:
Hazelcast's Replicated Map is somehow recent and still a bit unreliable (async updates can be lost in case of scaling down)
It seems like Geode became "stable" fairly recently (even though it's supposedly in development since the early 2000s)
Ignite looks like it could be a good fit, but I'm not too sure how their S3 discovery based system will work out if we keep adding / removing node regularly.
Thanks!
Geode should work for your use case. You should be able to use a Geode Replicated region on each node. You can choose to do synchronous OR asynchronous replication. In case of failures, the replicated region gets an initial copy of the data from an existing member in the system, while making sure that no in-flight operations are lost.
In terms of configuration, you will have to start a couple/few member discovery processes (Geode locators) and point each member to these locators. (We recommend that you start one locator/AZ and use 3 AZs to protect against network partitioning).
Geode/GemFire has been stable for a while; powering low latency high scalability requirements for reservation systems at Indian and Chinese railways among other users for a very long time.
Disclosure: I am a committer on Geode.
Ignite provides native AWS integration for discovery over S3 storage: https://apacheignite-mix.readme.io/docs/amazon-aws. It solves main issue - you don't need to change configuration when instances are restarted. In a nutshell, any nodes that successfully joins topology writes its coordinates to a bucket (and removes them when fails or leaves). When you start a new node, it reads this bucket and connects to one the listed addresses.
Hazelcast's Replicated Map will not work for your use-case. Note that it is a map that is replicated across all it's nodes not on the client nodes/servers. Also, as you said, it is not fully reliable yet.
Here is the Hazelcast solution:
Create a Hazelcast cluster with a set of nodes depending upon the size of data.
Create a Distributed map(IMap) and tweak the count & eviction configurations based on size/number of key/value pairs. The data gets partitioned across all the nodes.
Setup Backup count based on how critical the data is and how much time it takes to pull the data from the actual source(DB/Files). Distributed maps have 1 backup by default.
In the client side, setup a NearCache and attach it to the Distributed map. This NearCache will hold the Key/Value pair in the local/client side itself. So the get operations would end up in milliseconds.
Things to consider with NearCache solution:
The first get operation would be slower as it has to go through network to get the data from cluster.
Cache invalidation is not fully reliable as there will be a delay in synchronization with the cluster and may end reading stale data. Again, this is same case across all the cache solutions.
It is client's responsibility to setup timeout and invalidation of Nearcache entries. So that the future pulls would get fresh data from cluster. This depends on how often the data gets refreshed or value is replaced for a key.
I have a java web server and am currently using the Guava library to handle my in-memory caching, which I use heavily. I now need to expand to multiple servers (2+) for failover and load balancing. In the process, I switched from a in-process cache to Memcache (external service) instead. However, I'm not terribly impressed with the results, as now for nearly every call, I have to make an external call to another server, which is significantly slower than the in-memory cache.
I'm thinking instead of getting the data from Memcache, I could keep using a local cache on each server, and use RabbitMQ to notify the other servers when their caches need to be updated. So if one server makes a change to the underlying data, it would also broadcast a message to all other servers telling them their cache is now invalid. Every server is both broadcasting and listening for cache invalidation messages.
Does anyone know any potential pitfalls of this approach? I'm a little nervous because I can't find anyone else that is doing this in production. The only problems I see would be that each server needs more memory (in-memory cache), and it might take a little longer for any given server to get the updated data. Anything else?
I am a little bit confused about your problem here, so I am going to restate in a way that makes sense to me, then answer my version of your question. Please feel free to comment if I am not in line with what you are thinking.
You have a web application that uses a process-local memory cache for data. You want to expand to multiple nodes and keep this same structure for your program, rather than rely upon a 3rd party tool (memcached, Couchbase, Redis) with built-in cache replication. So, you are thinking about rolling your own using RabbitMQ to publish the changes out to the various nodes so they can update the local cache accordingly.
My initial reaction is that what you want to do is best done by rolling over to one of the above-mentioned tools. In addition to the obvious development and rigorous testing involved, Couchbase, Memcached, and Redis were all designed to solve the problem that you have.
Also, in theory you would run out of available memory in your application nodes as you scale horizontally, and then you will really have a mess. Once you get to the point when this limitation makes your app infeasible, you will end up using one of the tools anyway at which point all your hard work to design a custom solution will be for naught.
The only exceptions to this I can think of are if your app is heavily compute-intensive and does not use much memory. In this case, I think a RabbitMQ-based solution is easy, but you would need to have some sort of procedure in place to synchronize the cache between the servers on occasion, should messages be missed in RMQ. You would also need a way to handle node startup and shutdown.
Edit
In consideration of your statement in the comments that you are seeing access times in the hundreds of milliseconds, I'm going to advise that you first examine your setup. Typical read times for a single item in the cache from a Memcached (or Couchbase, or Redis, etc.) instance are sub-millisecond (somewhere around .1 milliseconds if I remember correctly), so your "problem child" of a cache server is several orders of magnitude from where it should be in terms of performance. Start there, then see if you still have the same problem.
We're using something similar for data which is read-only and doesn't require updated every time. I'm in doubt, that this is good plan for you. Just imagine you should have one more additional service on each instance, which will monitor queue, and process change to in-memory storage. This is very hard to test.
Are you sure that most of the time is spent on communication between your servers? Maybe you run multiple calls?
I’m thinking about building an offline-enabled web application.
The architecture I’m considering is as follows:
Web server (remote) <--> Web server/cache (local) <--> Browser/Prism
The advantages I envision for this model are:
Deployment is web-based, with all the advantages of this approach
Offline-enabled
UI (html/js) synchronization is a non-issue
Data synchronization can be mostly automated
as long as I stay within a RESTful paradigm
I can break this as required but manual synchronization would largely remain surgical
The local web server is started as a service; I can run arbitrary code, including behind-the-scene data synchronization
I have complete control of the data (location, no size limit, no possibility of user deleting unknowingly)
Prism with an extension could allow to keep the javascript closed source
Any thoughts on this architecture? Why should I / shouldn’t I use it? I'm particularly looking for success/horror stories.
The long version
Notes:
Users are not very computer-literate.
For instance, even superficially
explaining how Gears works is totally
out of the question.
I WILL be held liable if data is loss, even if it’s really the users fault (short of him deleting random directories on his machine)
I can require users to install something on their machine. It doesn’t have to be 100% web-based and/or run in a sandbox
The common solutions to this problem don’t feel adequate somehow. Here is a short analysis of each.
Gears/HTML5:
no control over data, can be deleted
by users without any warning
no
control over location of data (not
uniform across browsers and
platforms)
users need to open application in browser for synchronization to happen; no automatic, behind-the-scene synchronization
different browsers are treated differently, no uniform view of data on a single machine
limited disk space available
synchronization is completely manual, sql-based storage makes this a pain (would be less complicated if sql tables were completely replicated but it’s not so in my case). This is a very complex problem.
my code would be almost completely open sourced (html/js)
Adobe AIR:
some of the above
no server-side includes (!)
can run in the background, but not windowless
manual synchronization
web caching seems complicated
feels like a kludge somehow, I’ve had trouble installing on some machines
My requirements are:
Web-based (must). For a number of
reasons, sharing data between users
for instance.
Offline (must). The application must be fully usable offline (w/ some rare exceptions).
Quick development (must). I’m a single developer going against players with far more business resources.
Closed source (nice to have). Yes, I understand the open source model. However, at this point I don’t want competitors to copy me too easily. Again, they have more resources so they could take my hard work and make it better in less time than I could myself. Obviously, they can still copy me developing their own code -- that is fine.
Horror stories from a CRM product:
If your application is heavily used, storing a complete copy of its data on a user's machine is unfeasible.
If your application features data that can be updated by many users, replication is not simple. If three users with local changes synch, who wins?
In reality, this isn't really what users want. They want real-time access to the most current data from anywhere. We had better luck offering a mobile interface to a single source of truth.
The part about running the local Web server as a service appears unwise. Besides the fact that you are tied to certain operating environments that are available in the client, you are also imposing an additional burden of managing the server, on the end user. Additionally, the local Web server itself cannot be deployed in a Web-based model.
All in all, I am not too thrilled by the prospect of a real "local Web server". There is a certain bias to it, no doubt since I have proposed embedded Web servers that run inside a Web browser as part of my proposal for seamless off-line Web storage. See BITSY 0.5.0 (http://www.oracle.com/technology/tech/feeds/spec/bitsy.html)
I wonder how essential your requirement to prevent data loss at any cost is. What happens when you are offline and the disk crashes? Or there is a loss of device? In general, you want the local cache to be the least farther ahead of the server, but be prepared to tolerate loss of data to the extent that the server is behind the client. This may involve some amount of contractual negotiation or training. In practice this may not be a deal-breaker.
The only way to do this reliably is to offer some sort of "check out and lock" at the record level. When a user is going remote they must check out the records they want to work with. This check out copied the data to a local DB and prevents the record in the central DB from being modified while the record is checked out.
When the roaming user reconnects and check their locked records back in the data is updated on the central DB and unlocked.