Hi Tomasz, Thank you for this. I did a similar test, and I get similar results before the market opens in the morning. Right after the market closed it is much slower. I agree that bacfilling long history of one symbol is fast. But if you do the benchmark with a list of 100 symbols you will notice that it takes 3 times longer to fill the same number of bars (time to backfill 100 symbols , say 1000 bars each = 3 times the time it takes to backfill 1 symbol with 100 000 bars)
You can notice in the small "backfill" window that AB does it one symbol at a time. And what is important it does not backfill all the history for one symbol in one go. If it has 2000 bars to fill it will collect 800 bars, then there is a short pause, then it will fill 300 bars, pause, and then the remaining bars (numbers I gave in this example are random, and they actually vary each time). Why is it not filling all the bars in one go? The sequence is: "backfilling" => "history arrived". My understanding is that "backfilling" is the actual data transmission and then there is that long pause called "history arrived" when actually nothing happens (to my knowledge) but it is just as long as the "backfilling" period. And I can live with that after the market hours but not in live market. To me the problem is that when I run the scan on a large list of symbols AB is blocking the network. It practically does not generate any network traffic but it takes over the control over the network and other applications "time out" when trying to connect to the network. I dont get this problem when streaming large nr of symbols with Quotetracker. Second problem is the delay of the feed. Today my charts were moving for 1min 46 seconds after the market close. I dont think AB itself connects to the internet so it has to be the plugin, IQ connection manager or there is a problem at the IQFeed end. Another thing I noticed is that sometimes IQ connection manager takes as much CPU as broker.exe and another time IQ CM takes 0% of the CPU (all when streaming). Sometimes "reconnect" helps with this issue. --- In [email protected], "Tomasz Janeczko" <[EMAIL PROTECTED]> wrote: > > Hello, > > I did the following benchmark for you (that you can also duplicate on your > end to verify if you are getting similar results). > > Reproduction steps: > 1. File->New->Database > 2. Create new database > 3. Select Data Source from the combo (DTN IQFeed or eSignal or other) > 4. Enter 100000 into "Number of bars > 5. Select Base time interval: 1 minute and click OK button > 6. Wait until green light appears - so it is connected to the data source > 7. Add MSFT (enter it in ticker box to display the chart) > > TIME MEASUREMENTS as of Nov 5, 2008, 8:44PM ET > > Data source: IQFeed, symbol: MSFT, interval: 1 minute, 74934 bars arrived > TOTAL BACKFILL TIME: 15.2 sec > > Backfill rate: 4929 bars per second > > Data source: eSignal , symbol: MSFT, interval: 1 minute, 59258 bars arrived > TOTAL BACKFILL TIME: 5.5 sec > > Backfill rate: 10774 bars per second > > You should be getting similar results when you replicate conditions mentioned above. > > Please note that these measurements were done AFTER MARKET CLOSE. > This makes significant difference as compared to backfills during market hours > when they are noticeably slower. > I would need to do measurements tommorrow, but it seems that at least for IQFeed > the difference between pre/after hours and during market hours can be as large > as 5-10 times. > > > Best regards, > Tomasz Janeczko > amibroker.com
