CPU Problems with build 500

[ScottJ]

If you are using the latest build of 7.1 ( 500 ) and are seeing your charts "freezing" or your CPU continually maxing out to 100%, we want to use this thread to sort out the problem. Please post as much as you can about your set-up and if willing, please post your layout to "Andy's Group" under File Sharing and reference it here.

Thank you.

Scott J.
eSignal Support

[Lancer]

I'm glad you mentioned this. On different PCs (identical hardware/software), I run eSignal and RealTick (TAL data). eSignal is very CPU intensive compared to RealTick. With only 10 active charts and one quote window, the eSignal PC has regular instances of 100% CPU, and slow responsiveness to actions (e.g. moving a line) even when less than 100% CPU. The problem is apparent when charting high trade-rate stocks or e-minis. The problem includes program freeze (for up to a couple minutes), and that's typical at high trade-rate events (e.g. the open, news, strong reversal points).

On the RealTick PC, using a RealTick layout with more than twice eSignal's number of active windows (all charted symbols the same as on the eSignal layout), and using more incoming bandwidth, CPU% is half that of the eSignal PC. RealTick's outgoing data is less than 10 kbps average (almost nothing), whereas eSignal's outgoing kbps is at all times nearly equal to its incoming kbps. With outgoing data almost equal to incoming, it looks like eSignal has double the data-handling load of RealTick.

As to setup specifics, because eSignal has no numeric quantifier in Data Mangager expressing the load demand of a layout (a simple way to compare demands of various layouts (RealTick has this feature)), I'd probably have to email you the layout and all child window files. As to the PC eSignal runs on: CPU: Athlon XP 1800+; RAM: 512 MB; Monitors: 4; OS: Win2000; Internet: 768 kbps SDSL.

[Lancer]

Rally day. Lots of 100% CPU.

[ScottJ]

Can you post your layout so we can run it in-house to see if we get the same results. You can use the File Share Group "Andy's Group" and save it as a page so it's just one step.

Thanks,

Scott J.
eSignal Support

[greg]

I have had this problem since i installed eSignal. I have upgraded to build 500. I still had the problem.

I rebuilt everyone of my charts in the layout and that helped immensely. I would say that it cleared up 95% of the cpu 100%, application not responding incidents.

I still get cpu freezes in fast mkts. but it is now a rare event. Any further suggestions, as i would dearly like to be rid of this for good.

greg

[Lancer]

Today, with three high trade-rate symbols charted (MSFT, ESZ2, and NQZ2), 100% CPU was a regular event. Often, eSignal was frozen for minutes, and sometimes more than 10 min. at a time. In RealTick, with the same symbols charted (and more), and twice the number of active windows, CPU% was the usual half that of eSignal. eSignal was really demanding the data today; eSignal's incoming kbps was extraordinarily high, and was always at least twice that of RealTick. As usual, eSignal's outgoing kbps was almost equal to the incoming kbps. RealTick's outgoing kbps was nearly nothing (under 10 kbps).

In after-hours (4:15 ET), with no e-minis charted, eSignal is again frozen with 100% CPU (on MSFT earnings release), and an average incoming 325 kbps. Ten minutes later, the kbps has dropped down now to 182 kbps avg. but CPU% is still 100%. eSignal is just frozen and unresponsive. Looks like Task Manager will be required to end the process. Twenty minutes later, on its own, eSignal just cleared from a 20 min. freeze (from 4:15 to 4:35 ET). CPU % is now averaging 65%. Nice MSFT moonshot to 54 by the way.

So, is eSignal's software design, data feed, or data handling just inherently inefficient (e.g. compared to RealTick and TAL data)? Why does eSignal have such a high outgoing kbps, almost equal to incoming kbps; isn't that doubling the workload and adding to the high CPU problem? Is there a software or data bug causing high CPU and program freeze? Do layouts and child windows created in a prior version not run well in the current version? Could the problem be attributed to eSignal's incompatibility with other running programs (e.g. ZoneAlarm firewall)?

[AndyS]

Lancer,

A quick note to let you know I sent you a private mail to discuss this further.

Take care,
Andy

[jcramped]

I'm having a similar problem. I use eSignal only as a feed for my Wave59 program, so don't have a lot of stuff running in the actual program. I only have 4 windows showing in the eSignal main window: 1. 15 equities, 2. 5 indices, 3. a chart, 4. another chart.

When I first load eSignal (if I look at my "Processes" under "Task Manager" in Win2000), it shows winros.exe using upwards of 17,000K memory usage and winsig.exe shows upwards of 40,500K. When I maximize and then minimize the TurboFeed Data Manager and the eSignal program, the memory usage for each goes down to a respectable 1700-ishK.

I originally was running Wave59 using eSignal 7.02 build 451, and everything ran fine. Once I upgraded to 7.1, I started experiencing all kinds of problems with Wave59 - it was freezing, it was slow, it was shutting down, data was disappearing right off the charts, etc. I uninstalled 7.1, and am now back to using 7.02, but Wave59 is still not running as smoothly as it used to. I suspect I might have a few "remnants" of 7.1 hiding somewhere in my system.

Anyway, just thought I'd add my observations about the extreme memory usage, and how to at least alleviate it temporarily (for the day).

[AndyS]

Hi Jcramped,


When first loading eSignal, particularly if the market is open, you will see the cpu meter go to 100%. It's bringing in a lot of data. However, it doesn't take too long for that to come in and resources to begin to free up.

If you are experiencing sluggish data and are not using Advanced Charts, then please call into our Tech Support, as we do not have reports of this problem.

We do know of some issues when using Advanced Charts. They are spelled out by our Director of Desktop Software in a post to the Yahoo groups. Please check this link out; http://groups.yahoo.com/group/eSignal/message/4885.

Let me know if there's anything else I can do.

Regards,
Andy S.
eSignal Support

[Lancer]

Andy, can you copy that post here? Thanks.

[AndyS]

Lancer,

I don't want to bother the readers of this board with the contents of the private email. Basically, I was asking if you got my previous email detailing what files to email me.

I'd like to help with your freezing problems, but do to so would require I have a copy of you layout & data files.

Let me know if I can help.

Thanks,
Andy S.
eSignal Support

[Lancer]

Andy, this post:

"We do know of some issues when using Advanced Charts. They are spelled out by our Director of Desktop Software in a post to the Yahoo groups. Please check this link out; http://groups.yahoo.com/group/eSignal/message/4885."

[AndyS]

Hi Lancer,

O.K., I got it now. Here's the post in question.

--------------------------------------------------------
From: Matt Gundersen <[email protected]>
Date: Sat Oct 19, 2002 11:25 am
Subject: Freezing, High CPU, et all.....

Hi Everyone,

There have been some reports of freezing and high cpu with the latest build of eSignal (7.1, build #500). I'd like to give everyone a status update as to what we've (eSignal) found and what is being done.

There are two things we have seen that contribute to the high cpu.
#1 - A large amount of line tools on a chart
#2 - Certain EFS formulas.

Regarding #1 (line tools), you may have a large amount of line tools on your chart that can cause high cpu. How do you determine if you have a large amount of line tools on your chart? Right now, there is really no definitive way. We've learned that looking at the size of the ACH file is not enough. However, if you do have an ACH file > 20K, I'd be a little suspicious.

How can you remedy this situation related to too many line tools? Right click on your charts, goto Line Tools->Delete for (all symbols)->all. Then save the ach file. This may help reduce CPU.

What's causing this accumulation of line tools? We are not sure but are researching the problem to figure it out.

What are we doing to aid in helping identify line tool issues in the future? There will be a "Line List" dialog that will let you quickly identify if there are an accumulation of line tools. There will also be an auto purge lines feature (but it will ask you first before purging). We are also reviewing the code and looking for areas where performance tweaks can be made.


Issue #2 - High cpu due to certain EFS formulas. First, I'd like to say "Thank you" too those of you that provided Layouts, ACH files, and EFS files. This really helped us identify these EFS issues.

In reviewing various layouts, we discovered a few formulas that were causing high cpu. In specific /library/volumema.efs, /library/stochrsi.efs, and
/pricebars/mapricebars.efs. These formulas were doing a lot of math in the EFS. We made modifications to them so they would use the builtins instead.

We've heard back from a few people and learned that replacing these formulas with the updated EFS's helped reduce cpu.

To help identify EFS/CPU bottlenecks in the future, we are adding an EFS Performance Monitor. This will allow an EFS Developer or end user to turn on the performance monitor and see which EFS's are using the most cpu. This way, if you are having high cpu, you can quickly determine if it's due to an EFS and which which one specifically. This will aid the developer in tweaking their own formula. It will aid an end-user if they call eSignal Customer Service in helping report which specific EFS is causing high cpu.

For anyone interested (you must have 7.1, Build 500), the tweaked version(s) of these formulas can be found at:
http://share.esignal.com/groupconten...eaks&groupid=1

We are continuing to research the EFS/CPU issues and look for ways to improve EFS engine performance.

Once again, thanks to all that worked with us to isolate these problems.

We are working towards having another pre-beta available with the features discussed above as well as many other enhancements (ie: line tool improvements, etc...). For anyone intersted in receiving a copy of the pre-beta, please send an email to [email protected]

Matt Gundersen
eSignal
Director Desktop Software

[EdKoF]

Formulas do the trick

Just to give an example of how deleting some formulas solve most of the cpu load problem.

I downloaded a formula which show Momentum with a moving average. I've changed it so I can show 2 moving averages on the Momentum.

I had 4 charts with 2 times this formula per chart. At busy times, when eSignal has to calculate 8 x 2 = 16 ma's on MOM, it just locked up completely.

Just putting the original MOM back, solved it.

Although this solved the problem, it is of course not really a solution. I hope that the developers can take a good look at how this problem is related to the program architecture.

Regards,

Edo.

[ScottJ]

We are working on a comprehensive "white paper" to provide customers that experience freezing or high CPU usage. Some of the elements of that white paper are already described on this thread by Matt.

We continue to work with impacted customers individually to isolate the issues and to gather more information on how we can optimize efs usage and the overall performance of eSignal.

For this handful of customers having this problem, there does appear to be a pattern related to volatile issues, like the emini. We're looking very closely at that.

Should have the first revision of the white paper ready by Monday.

Thanks,
Scott J.
eSignal Support