I'm considering using bittorrent for a large data dissemination problem where the data source is petascale and users will want up to several terabytes. Some details
Number of torrents potentially in the millions
torrent sizes ranging from 100Mb to 100Gb
A stable set of clusters around the world capable of acting as seeders each holding a large subset of the total torrents (say 60% on average)
A relatively small number of simultaneous users (less than 100) wanting to download on average a few terabytes of data.
I expect the number of active torrents to be small compared to the total available but quality of service is important so there must be several seeders for each torrent or some mechanism for launching new seeders.
My question is can bittorrent clients handle seeding huge numbers of torrents, most of which are idle? Would I need to stripe torrents across the seeders in a cluster or could each node be seeding all torrents it has access to? Which client would do the best job? Are there any tools for managing clusters of seeders?
I am assuming that trackers can be made to scale to this level.
There are 2 main problems:
Each torrent (typically) needs to announce to a tracker periodically, this might end up using a significant amount of bandwidth.
The bittorrent client itself need to be written in a way to scale with a large number of torrents
As for the tracker traffic, let's assume you have 1 million torrents, the typical re-announce interval is 30 minutes, but some tracker has it set to 1 hour. Let's be conservative and assume your tracker uses 1 hour announce intervals. You will have to make 1 million GET requests per hour, let's say each request is 400 bytes up and 100 bytes down (assuming most responses will not contain any peers), that's about 111 kB/s up and 28 kB/s down constantly. That's not so bad, but keep in mind that TCP requires an extra round-trip for establishing connections, so that's another 40 bytes down and 40 bytes up.
This can be mitigated by only using UDP trackers. Then you would only need a single connect-message, and you can reuse the connection ID for each announce. Each announce message would then be 100 bytes, and the returned message would be a bit more compact as well, let's assume 60 bytes. That would get you 28 kB/s up and 16kB/s down, just to keep the torrents announced. For this you would need a client with decent udp tracker support (one that caches the connection ID for instance).
Not too bad, assuming that's insignificant compared to the actual data your seeds would send.
However, you don't necessarily need to stripe your torrents across separate data centers, you could also use an HTTP server to seed the torrents. All major bittorrent clients support http seeding, and you wouldn't have to worry about announcing to the tracker (the URL is burned into the .torrent itself).
As for a client that scales well with torrents, I don't know for sure, I haven't done any measurements. It should be fairly straightforward to just generate a million random torrents and try to load it up.
I have done some optimization work in libtorrent rasterbar to make it scale well with many torrents, I haven't tried millions though.
I've written a blog post on this topic, here.
You may be looking for Hekate
It's in, at best, pre-alpha right now, but it's quite nearly what you're describing.
To not collapse under the overhead of useless tracker announces and scrapes in the millions (and that in every announce interval), you have to restrict your seeding clusters to only load the current working set of items that are requested right now. Downloaders need to get (download) the .torrent file from a central place anyway, and that could trigger loading it into the seeding clusters. Alternatively, determine activity for a particcular info-hash by recognizing announces that do NOT originate from one of your seed clusters.
rTorrent has fast-resume (meaning no hashing happens when an appropriately prepared .torrent is loaded), and is controllable via xmlrpc so you can decommission idle items. That way, a .torrent download can trigger the actual data to be available for the next 24 hours, or as long as there's activity in the swarm.
The protocol allows for this, but I do not know which clients would scale to millions of torrents. In the worst case, you would have to write your own seed-only client.
The protocol feature most relevant to your use case is that, when a peer connects to another, the connecting peer is supposed to send the torrent's info-hash first. This means that a single listening TCP port could be used to seed an unlimited amount of torrents, with almost zero resources used when idle.
This can be found on The BitTorrent Protocol Specification:
If both sides don't send the same value, they sever the connection. The one possible exception is if a downloader wants to do multiple downloads over a single port, they may wait for incoming connections to give a download hash first, and respond with the same one if it's in their list.
I also found the same on this Bittorrent Protocol Specification v1.0:
The initiator of a connection is expected to transmit their handshake immediately. The recipient may wait for the initiator's handshake, if it is capable of serving multiple torrents simultaneously (torrents are uniquely identified by their info_hash).
However, there is one thing that would increase your load, and it is the tracker. With the normal tracker protocol, each client has to periodically announce to the tracker each torrent it has, together with information like how much it has uploaded. With millions of torrents, this would present a somewhat high load. If you were writing your own mass-seed-only client, a separate protocol to announce your seeders to the tracker would be a good idea.
Related
I'm wondering, how modern DNS servers dealing with millions queries per second, due to the fact that txnid field is uint16 type?
Let me explain. There is intermediate server, from one side clients sending to it DNS requests, and from other side server itself sending requests to upper DNS server (8.8.8.8 for example). So the thing is, that according to DNS protocol there is field txnid in the DNS header, which should be unchanged during request and response. Obviously, that intermediate DNS server with multiple clients replace this value with it's own txnid value (which is a counter), then sends request to external DNS server and after resolving replace this value back to client's one. And all of this will work fine for 65535 simultaneous requests due to uint16 field type. But what if we have hundreds of millions of them like Google DNS servers?
Going from your Google DNS server example:
In mid-2018 their servers were handling 1.2 trillion queries-per-day, extrapolating that growth says their service is currently handling ~20 million queries-per-second
They say that successful resolution of a cache-miss takes ~130ms, but taking timeouts into account pushes the average time up to ~400ms
I can't find any numbers on what their cache-hit rates are like, but I'd assume it's more than 90%. And presumably it increases with the popularity of their service
Putting the above together (2e7 * 0.4 * (1-0.9)) we get ~1M transactions active at any one time. So you have to find at least 20 bits of state somewhere. 16 bits comes for free because of the txnid field. As Steffen points out you can also use port numbers, which might give you another ~15 bits of state. Just these two sources give you more than enough state to run something orders of magnitude bigger than Google's DNS system.
That said, you could also just relegate transaction IDs to preventing any cache-poisoning attacks, i.e. reject any answers where the txnid doesn't match the inflight query for that question. If this check passes, then add the answer to the cache and resume any waiting clients.
The getstream.io documentation says that one should expect retrieving a feed in approximately 60ms. When I retrieve my feeds they contain a field named 'duration' which I take is the calculated server side processing time. This value is steadily around 10-40ms, with an average around 15ms.
The problem is, I seldomly get my feeds in less than 150ms and the average time is rather around 200-250ms and sometimes up to 300-400ms. This is the time for the getting the feed alone, no enrichment etc., and I have verified with tcpdump that the network roundtrip is low (around 25ms), and that the time is actually spent waiting for the server to respond.
I've tried to move around my application (eu-west and eu-central) but that doesn't seem to affect things much (again, network roundtrip is steadily around 25ms).
My question is - should I really expect 60ms and continue investigating, or is 200-400ms normal? On the getstream.io site it is explained that developer accounts receive "Low Priority Processing" - what does this mean in practise? How much difference could I expect with another plan?
I'm using the node js low level API.
Stream APIs use SSL to encrypt traffic. Unfortunately SSL introduces additional network I/O. Usually you need to pay for the increased latency only once because Stream HTTP APIs supports HTTP persistent connection (aka keep-alive).
Here's a Wireshark screenshot of the TCP traffic of 2 sequential API requests with keep alive disabled client side:
The 4 lines in red highlight that the TCP connection is getting closed each time. Another interesting thing is that the handshaking takes almost 100ms and it's done twice (the first bunch of lines).
After some investigation, it turns out that the library used to make API requests to Stream's APIs (request) does not have keep-alive enabled by default. Such change will be part of the library soon and is available on a development branch.
Here's a screenshot of the same two requests with keep-alive enabled (using the code from that branch):
This time there is not connection reset anymore and the second HTTP request does not do SSL handshaking.
I'm running into a lot of trouble when trying to seed a lot of torrents( > 10k) with libtorrent.
They include:
Choking my network connection
Tracker requests timing out(libtorrent tracker error)
When using auto-manage(they go from checking to seeding very slowly, even when my active_seeding is set to unlimited.
I used to let them be automanaged, but I'd find that it makes nearly all of them unavailable.
Here are my current settings:
sessionSettings.setActiveDownloads(5);
sessionSettings.setActiveLimit(-1);
sessionSettings.setActiveSeeds(-1);
sessionSettings.setActiveDHTLimit(5);
sessionSettings.setPeerConnectTimeout(25);
sessionSettings.announceDoubleNAT(true);
sessionSettings.setUploadRateLimit(0);
sessionSettings.setDownloadRateLimit(0);
sessionSettings.setHalgOpenLimit(5);
sessionSettings.useReadCache(false);
sessionSettings.setMaxPeerlistSize(500);
My current method is to loop over all my 10k+ torrents, and run torrent.resume(). When using automanage, this basically only starts ~ 50 of the torrents, and the others start about at a rate of 1 torrent per 10 minutes, which wouldn't work. When not using automanage, it chokes my connection.
BUT, when I do only 30 of them, they all seem to seed correctly, so my next plan is try to resume() them in groupings either with a time delay, or after they've received a tracker_reply.
I tried to garner what I could from this, but don't know what my settings should be specifically:
http://blog.libtorrent.org/2012/01/seeding-a-million-torrents/
I'd really appreciate someone sharing their settings for seeding thousands of torrents,
When not using automanage, it chokes my connection.
Since you say it can run either on a hosted server or domestic internet connection then you will have not much of a choice but to throttle torrent startups. Domestic internet connections are generally behind consumer grade routers and possibly CGNAT, both of which have fairly small NAT tables that will eventually choke from concurrently established TCP connections (peer-peer connections, tracker announces) or UDP pseudo-connections (UDP trackers, µTP, DHT)
So to run many torrents at once you will have to limit all active maintenance traffic of that kind so that the torrents are only started to listen passively for incoming connections.
I'm implementing a websocket-secure (wss://) service for an online game where all users will be connected to the service as long they are playing the game, this will use a high number of simultaneous connections, although the traffic won't be a big problem, as the service is used for chat, storage and notifications... not for real-time data synchronization.
I wanted to use Alchemy-Websockets, but it doesn't support TLS (wss://), so I have to look for another service like Fleck (or other).
Alchemy has been tested with high number of simultaneous connections, but I didn't find similar tests for Fleck, so I need to get some real info from users of fleck.
I know that Fleck is non-blocking and uses Async calls, but I need some real info, cuz it might be abusing threads, garbage collector, or any other aspect that won't be visible to lower number of connections.
I will use c# for the client as well, so I don't need neither hybiXX compatibility, nor fallback, I just need scalability and TLS support.
I finally added Mono support to WebSocketListener.
Check here how to run WebSocketListener in Mono.
10K connections is not little thing. WebSocketListener is asynchronous and it scales well. I have done tests with 10K connections and it should be fine.
My tests shows that WebSocketListener is almost as fast and scalable as the Microsoft one, and performs better than Fleck, Alchemy and others.
I made a test on a Windows machine with Core2Duo e8400 processor and 4 GB of ram.
The results were not encouraging as it started delaying handshakes after it reached ~1000 connections, i.e. it would take about one minute to accept a new connection.
These results were improved when i used XSockets as it reached 8000 simultaneous connections before the same thing happened.
I tried to test on a Linux VPS with Mono, but i don't have enough experience with Linux administration, and a few system settings related to TCP, etc. needed to change in order to allow high number of concurrent connections, so i could only reach ~1000 on the default settings, after that he app crashed (both Fleck test and XSocket test).
On the other hand, I tested node.js, and it seemed simpler to manage very high number of connections, as node didn't crash when reached the limits of tcp.
All the tests where echo test, the servers send the same message back to the client who sent the message and one random other connected client, and each connected client sends a random ~30 chars text message to the server on a random interval between 0 and 30 seconds.
I know my tests are not generic enough and i encourage anyone to have their own tests instead, but i just wanted to share my experience.
When we decided to try Fleck, we have implemented a wrapper for Fleck server and implemented a JavaScript client API so that we can send back acknowledgment messages back to the server. We wanted to test the performance of the server - message delivery time, percentage of lost messages etc. The results were pretty impressive for us and currently we are using Fleck in our production environment.
We have 4000 - 5000 concurrent connections during peak hours. On average 40 messages are sent per second. Acknowledged message ratio (acknowledged messages / total sent messages) never drops below 0.994. Average round-trip for messages is around 150 miliseconds (duration between server sending the message and receiving its ack). Finally, we did not have any memory related problems due to Fleck server after its heavy usage.
I need to send identical information to 100's of clients over the Internet. I currently maintain a list of client connections and iterate over the list. Obviously the longer the list gets the more latency there is toward the end of the list.
I have looked at multicasting. However unless I am missing something it is only good for LAN-based communications at present. It requires routers that support multicasting and most routers do not. There is no mechanism that I can see where one requests an available multicast address to avoid broadcasting to an address already in use.
So my questions are:
1) Am I missing something and can I use multicasting to accomplish this? (have tried without success)
2) Other than multicasting, is there a short cut to sending identical packets to many recipients?
I solved the problem by multicasting between threads in the server. Every client connection results in the creation of an object. These objects are stored in a queue. Each object has its own thread and joins the multicast group. When the server multicasts a string to the client objects the delay that arose from the list iteration no longer occurs.
Every now and then there is huge latency (nearly a second). I suspect that this is a JVM thing.
If you need high performing low latency IO, you shoud try http://nodejs.org/
You may be also interested in some cache http://memcached.org/