CPU Problems with build 500

[ACM]

Scott
FYI Globex has issued an advisory that they were experiencing market data
freezes at 9:00 CST and that they were looking into this issue.
Alex

----- Original Message -----
From: "eSignal Forums Mailer" <[email protected]>
To: <[email protected]>
Sent: Thursday, October 31, 2002 11:02 AM
Subject: Reply to post 'CPU Problems with build 500'


> Hello Alexis C. Montenegro,
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>

[DionLoy]

Builtin studies on builtin studies

In v7.2, which is currently under pre-beta testing, you have the ability to use some builtin studies with the output from other builtin studies.

ie, using the builtin moving average (MAStudy) to calculate moving averages from momentum, RSI, etc ...

This should speed up those formulas that have hand-built moving averages once they are rewritten to use this new feature.

[ScottJ]

Robert,

We simply can not agree with your assertions. We recognize that you continue to experience these fast market freezes but most customers are not experiencing these problems. With the size of our customer base, if they were, we'd be hearing from thousands of customers, trust me.

Are you still running 3-4 instances of eSignal? If so, while technically possible, that is not the way the system is intended to run. We strongly feel that your multiple sessions are at least a partial reason why you are experiencing the problems you are.

If you want to test our data delivery, you should be able to simply parse down to 1 session of eSignal and a simple workspace. You'll see that our communication infrastructure works very well.

We will continue to work individually with customers and keep searching for more ways to optimize the performance of eSignal and each individual's situation. We will also continue to thoroughly test our internal systems so we leave no rock un-turned.

Thanks,
Scott J.
eSignal Support


From: "Robert Dee" <[email protected]>
Date: Thu Oct 31, 2002 11:20 am
Subject: Re: [eSignal] Esignal Freezing


Esignal definitely cannot handle fast market condition.

Esignal froze again the last surge around 1040 PST

You probably need to examine your communications infrastructure or your program codes.

[hotsauce]

CPU problems are very real!!

Hi All

I can assure all that the cpu problems are very real. I am running one instance of esignal and am now down to only 3 ACH windows, yet when the market gets real busy my cpu goes to 100% and stops responding to mouse and key clicks.

[ScottJ]

Hotsauce

We aren't trying to say the CPU problems aren't "real". We are working directly with customers and, in quite a few cases, can confirm the CPU spikes.

We can also confirm quite confidently that eSignal runs beatifully during fast market conditions so this is not a problem that impacts all users. For those that it does impact, we 100% understand how "real" it is.

We think we're making significant progress on this issue and should know quite a bit more next week. If you want to participate in the pre-beta, where new versions are tested, that would be great.

You need to register for File Share, then register for this group. You'll get an email confirming you have access.

http://share.esignal.com/groupcontents.jsp?folder=&groupid=52

Thanks,

[hotsauce]

Re: Hotsauce

Originally posted by ScottJ

We simply can not agree with your assertions.

We can also confirm quite confidently that eSignal runs beatifully during fast market conditions so this is not a problem that impacts all users.

I realize the above quotes are taken somewhat out of context, but, as a heads up, let me say that they come across as pure denial to those of us who have been laboring with this problem for how many weeks now. This problem DOES potentially affect all users. People who havent experienced it yet are using the same software version I am. Just because another user has not experienced the problem yet doesnt mean that they wont stumble onto the circumstances that trigger it sometime soon.

I am more than willing to work with the eSignal staff to help identify, isolate, and resolve this problem; but, as a customer paying almost $500.00 per month for all of the combined services I am receiving from eSignal, I shouldn't have to deal with "We simply can not agree with your assertions"! Enough Said!

Now, for more objective feedback

Since this problem has apparently been observed by a minority of users, it becomes fairly obvious to me that there must be a really unique set of circumstances that set the problem off.

1. I am not using any line tools!
2. I dont think I am using any EFS formulas. Can someone please describe in more detail what an EFS formula is?
3. I agree with a previous poster, that the issue seems to be more related to charting the ES or NQ issues during periods of high market activity. However they are, after all, the highest volume futures (I think). When I just have pure stock charts up I have never observed the problem, although it would be very seldom that I have stocks only charts up. When I have just a single ACH chart up on the ES, and another on the NQ, I see the problem frequently.
4. If, indeed, esignal is experiencing as much outbound network traffic as inbound I would be very concerned about that. I will watch for it on monday after the market open and report my findings. High Context switching rates in the network drivers would definately be an area of immediate concern.

I sincerely hope that we can keep the rest of the dialog on this thread limited to the problem at hand.

Cheers

[gspiker]

If you ever loaded a study via RightClick->Formulas, that would be an EFS formula (sometimes called an EFS Study).

FYI, I had no problems (other than the linetools problem), til build 500. With 500 I saw CPU spikes up to 100% for short periods during fast markets. With the prebeta 505 and 506 the CPU usage is back to where it used to be for me, maybe even better as I hardly ever spike up above 40% now. So for those that have only seen the problem with Build 500 - you might want to sign up for the prebeta and see if it helps.

[AndyS]

Hi,

Let me clarify Scott's post a bit for you, cause as you say, it's taken out of context. To fully understand what he's addressing you would had to have follow Robert Dee's post over time.

The reply deals mostly with the assertion (intended or not) from Robert that we are not able to deliver the S&P emini ticks in real-time when surges occur. We know the data is there to the client's machine when it happens. The freezing that you and others are experiencing is not a result of data delivery being delayed during market surges. Thus we cannot agree with the assertion that we are unable to deliver the data in real-time when market surges occur. We believe Robert's freezing experiences are due to multiple sessions of eSignal running simultaneously. Because we haven't designed or tested the system to handle multiple sessions of eSignal, it's difficult to use his layout & experiences to determine the problem.

As for what you are experiencing. We have your layout and I personally have been running it on and off since I received it without it freezing up on me. Gary, the Tech Support rep whose been working with you has not been able to reproduce your problem as well. Your layout consist of 4 or 5 Advanced Charts with mostly the Advanced Get indicators. Each chart is an interval and not a tick. We previously advised you on the chart sizes & deleting old lines and you have said that you don't use lines, so we know that this is not the problem. This is an example whereby we know you are experiencing freezing, but we are unable to reproduce it with your layout, and this makes it difficult to trouble shoot.

Others that have submitted their layouts, I have been able to reproduce the freezing. In most of these cases, they were running multiple charts along with multiple tick charts that also had custom formulas attached. With these, we see where we need to look to find the problem and correct. We also know that with the exploding volume on the s&p emini's there are limitations the system can handle as compared to stocks. I believe all vendors are experiencing this problem.

The message here is, we hear what our clients are saying and we are working at providing solutions that will work for all of our clients. We ask that those experiencing this problem to post their layouts to my group in file share (Andys Group) so that we may trouble shoot their layouts for gathering further information in fixing this problem. I will be working with our engineers with your layout as well and we will try to nail down what is causing your system to freeze.

Thank you for taking the time to express your concerns on this forum. I hope this helps in assuring you and others that we are actively working on this problem. We look forward to your continued feedback as we work through this situation.

Regards,
Andy S.
eSignal Support

[hotsauce]

Hey Garth

Thanks for the response and the suggestion. I am working at getting the 7.2 beta as we speak.

Andy - Thanks also for the response and the clarification. I would agree wholeheartedly that I dont think the data is being held up. Its seems to be something on the local machine IHMO.

Once again, I have only seen this problem when I have ES and NQ up on the screen. When the market gets real busy, the program seems to "get behind" in its processing, CPU goes to 100% and stays there, and the program stops responding to key and mouse clicks - even the clock stops updating; soemtimes for 10, 20, and even 30 minutes. Then, after the market volume slows down the program suddenly comes back and behaves normally again. I didnt see the problem at all on friday, but then again the volume was really low and I only played the NQ then and didnt have ES up at all.

The first hour is a good time to reproduce this problem, as volume usually spikes up during the first 15 minutes and then tapers off. During the middle of the trading session you wont usually see this problem very often unless there is unusual news. Another good opportunity to see the problem would be on days when a major report is due out; like at 10:00Am eastern - leading to a major trading burst just after the annoincement.

[Lancer]

Andy,

Regarding: "We also know that with the exploding volume on the s&p emini's there are limitations the system can handle as compared to stocks. I believe all vendors are experiencing this problem."

With respect to other vendors/software, my posts earlier in this thread described some comparisons with RealTick, running eSignal and RealTick on two different PCs side-by-side. It's notable that RealTick handles a much heavier load (twice the number of open charts, etc., and higher incoming kbps) while at a much lower CPU%. Further, RealTick's outgoing kbps is nil while eSignal's is almost equal to incoming kbps. There's obviously a difference in software design and data handling. eSignal is comparatively inefficient with both bandwidth and CPU resources. In addition to a user's particular layout (tick charts, lines, etc.), maybe changes to improve overall software/data efficiency are something to look at?

[EdKoF]

When closing the eSignal application during high volume/cpu periods, and just using Ensign (which runs on the eSignal datafeed), cpu levels are normal. This rules out the datafeed (winros.exe) as source of failure, in my opinion, but points to the eSignal application architecture.

The small hickups in presentation of exact prices for ES and NQ futures during a spike in volume are common also with other vendors and brokers, as far as I can judge.

[mattgundersen]

Re: Reply to post 'CPU Problems with build 500'

I believe we found the cpu problem on Friday AM and addressed it. We did some
testing with users before the close of market and received positive feedback.

If you are interested in a pre-beta copy to try out for Monday AM, please post
or email me directly before Monday AM ([email protected]). I'll get you
setup with the prebeta.

m.

--- eSignal Bulletin Board Mailer <[email protected]> wrote:
> Hello mattgundersen,
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>