omidqrose
Hello,
We have been using the LightStreamer Vivace for updating our clients. Recently we have noticed strange delays in peak periods. Using the Monitor Console, this information is acquired: At the peak time we had 1650 online users, 3000 connections, 1700 sessions and 6000 total subscribers.
Currently we have 2 web pages, the first page contains a stock watch, watch-list and total market details. The second is a market list with 400 instruments. Each instrument is updated using three message types. MaxItemFrequency of the above three messages is 1 second. In the second page, the messages are updated every one second and MaxBandWidth is 250.
Are these delays normal?
Server specification:
OS: windows web server 2008 r2
System type: 64-bit Operating System
DL380G7 HP Server
2*CPU Quad-Core Intel Xeon Processors X5670
(2.80GHz,8MB L3 Cache,95W)
32GB PC3-10600 RAM
6*146GB/15K DP SAS 6G HDD
Smart Array P410i/512MB+BBWC Controller
2*750W Hot Plug Power Supply
Alessandro Alinone
As discussed via email, in your scenario a single page seems to cause 400 * 3 = 1200 subscriptions and the bandwidth is 250 / 8 = about 31 kilobytes per second.
In this scenario the bandwidth limit may be hit in peak time.
This would cause intense filtering for the most active items and might cause delays for the less active ones.
In any case, we should analyze the whole history of the internal monitor in order to understand the Server state. Please activate the internal monitor logging, by configuring <text_monitor> to Y and by providing a suitable appender for the LightstreamerMonitorText log4j logger.
