You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Is IB making our orders & quotes compete with back-fill/charting bandwidth?
- Thread starter Raystonn
- Start date
This indeed confirms what we have been suspecting all the time. That new crap grossly interferes with our trading-order traffic, especially at critical moments. Throw it off and set up a different system using seperate communication channels, paid for by the chart & backfill crowd.Quote from IBsoft:
... Ditching the charts allowed the number of reconnects to drop below a level where the firewalls could start servicing the connection requests.
I think, if you use SSL, your trade and account info go over one connection (to gw1.ibllc.com:4001) and your market data goes over a different connection (to gw1.ibllc.com:4000).
I imagine each of these hosts are really server farms with lots of bandwidth.
And the bandwidth needed to get data for a real-time chart is (after the backfill gets done) probably about as much as the bandwidth needed to update 1 line on your market data page.
I imagine each of these hosts are really server farms with lots of bandwidth.
And the bandwidth needed to get data for a real-time chart is (after the backfill gets done) probably about as much as the bandwidth needed to update 1 line on your market data page.
I think, if you use SSL, your trade and account info go over one connection (to gw1.ibllc.com:4001) and your market data goes over a different connection (to gw1.ibllc.com:4000).
These are different ports on the same machine. The machine (likely a router/firewall) still must accept all traffic and sort it out. Every millisecond spent routing useless back-fill and chart data is a millisecond not spent routing an order or fill confirmation. They need the non-critical data to be run through a completely separate router with a different IP address attached to a different internet pipe. The T3 (or whatever) that services orders and fill confirmations should be dedicated to doing *only* that.
And the bandwidth needed to get data for a real-time chart is (after the backfill gets done) probably about as much as the bandwidth needed to update 1 line on your market data page.
Right, and that's why they needed to turn off charting and back-fill to fix the Denial of Service attack, which is a flood of excess bandwidth. I really don't care if the non-critical data is a back-fill request, a chart request, a reconnect request, or something else caused by a bug. That data should not be going through the same pipe to IB's system, period!
-Raystonn
Quote from IBsoft:
Ditching the charts allowed the number of reconnects to drop below a level where the firewalls could start servicing the connection requests.
They are still not servicing my connection requests and I have been trying to connect for like hours by now. I am ending in some black hole, no info on the connection to TWS, no connection failure boxes, no TWS, but since Sierrachart responds with the message to the effect that TWS is not running so it's safe to assume that I am not connected...
Any idea when the things are going to be back to normal?
Quote from IBsoft:
Two unrelated things:
- We had a faulty controller in our HighAvailability data server, which houses the transaction related persistent stores. That got corrected, shortening the order-acknowledgement times.
- We turned off the charts because we had a network event w/ our service provider, resulting in all the TWS on American continent reestablishing 1-3 connections each, virtually all at the same time, resulting in a condition that was similar to a DOS attack. Ditching the charts allowed the number of reconnects to drop below a level where the firewalls could start servicing the connection requests.
Thanks IB for all efforts wrt the faulty controller, i'm happy to return to rapid executions going forward. Samaritan's proactive attention and direct collaboration with me on the issue have been outstanding, and a fantastic support example.
Thanks very much
Quote from Avid_Consumer:
i can say this, since those features went down this afternoon, my chronic order acknowledgement and position column delays are 100% gone this afternoon, and things are consistently running as fast as I can ever recall. unmistakable difference.
Followup on my earlier post:
As I explained in my earlier post the improved order acknowledgement times are caused by the repair of the HA data server. When the charts are restored, please observe the acknowledgement times and let me know if you thing they deteriorated. Thank you.
will do, thanks for clarifying the fixes. i got nailed on some missed fills while connectivity was down the second time and confounded issues. in general as ranges contract, executions play an increasing role here and i appreciate the attention/info on both fronts.
thanks
thanks