Bitsharesbot Quick Update on downtime due to server overload[Server back up!]

aperson (53)in #bitsharesbot • 6 years ago (edited)

Update: Site is back up. Turns out it was a malicious DDoS attack on our servers. We have taken steps to prevent another future DDoS attacks. We are sorry for any inconvenience that has been caused. Also, if your account balance is wrong please create a new account.

The site (https://bitsharesbot.com) is temporary down due to the server being overload. At the moment we are unsure if it was a DDoS attack or it just too many users are attempting to access the bot. We had managed to restore the bot servers back up online. We are currently re-syncing the node(5 hours needed (ETA 1:30pm EST)). Access will be available when the node is fully synced . Sorry for the inconvenience.

#bitshares #bots #notbtsbots #btsbot

6 years ago in #bitsharesbot by aperson (53)

Sort:

abit (70) 6 years ago (edited)

Interesting. Seems it's much slower than the old btsbots.com for me. Not sure why. I'm not visiting it from China.
//Edit: better today.

$0.25

2 votes

[-]

aperson (53) 6 years ago

The current lag is not caused by geographical location. Rather, it occurs due to three main reasons:

Heavy traffic in the bitshares network. As the endpoint node for the wallet, we need to handle the massive fan-out of all new incoming operations in the bitshares network (around 50x fan out). In this case, our bottleneck is the CPU, as that each active connection require CPU time to process each request.
Heavy DB load when computing statistics and distributing them. This is an inherited problem from btsbot, and it lies in the design choices the original developer choice in the database (mango). Mango is a bad db choice largely due to its read/write latch. As a write is blocking for all reads in the db and vice versa, which is different from SQL which has smaller lock granularity.
Scalability problem in the original btsbot algorithm. The original algorithm is an O(n^2) for many calculations, including the 28-day volume calculation, where n is the number of tx in the bitshares network.

We do have a plan to migrate the lag with our build #2, which will be open for public testing around February. One of the ways we are doing it is by reducing the traffic thru our servers by implementing a way for the wallet to work side by side with existing node infrastructure, allowing the client to connect to existing bitshares nodes to reduce the amount of data they will need to pull from our servers. The other ways we are addressing this problem is to migrate to SQL. Currently, we are still optimizing our SQL servers, but we are achieving 1k tx (3k tx per block) per second under our stress testing. (Compare to mongo db which tops around 200 tx per second --600 tx per block ) Lastly, we have written a linear time ( O(n)) algorithm for most of the O(n^2) calculation.

$0.98

2 votes

[-]

abit (70) 6 years ago

Thanks for the explanation.

It's still a bit strange that your site is much slower than the old btsbots.com site, perhaps because alt was using more powerful hardware, or maybe there are more data / users now.

Your business model (via fees that I can see) is interesting, which is similar to a model I've thought of. Looking forward to see where it will head to.

By the way, is there something that I can help? I think you know how to reach me.

$0.00

[-]

petherbaath (33) 6 years ago

Still down? Constantly saying "login failed".

$0.00

1 vote

[-]

aperson (53) 6 years ago

It is working fine at the moment. Make sure you use your active private key, when you register. A login failed means our server is working fine, it is something wrong with the key or bitshare blockchain as it is not accepting the key. Sometimes the blockchain is slow and you should try mutiple times, but if it is constantly saying "login failed" then it is the problem of the key. Sometimes keys can get corrupted if your browser clears the history, so please try re-importing if that is the case.

$0.88

2 votes

[-]