Page 1 of 2 12 LastLast
Results 1 to 13 of 26
  1. #1
    Join Date
    Sep 2007
    Location
    Kenilworth NJ
    Posts
    1,579
    Post Likes

    Anybody use Skyspark?

    This is being suggested by a 3rd party guy with trust from the customer, but I have reason to doubt his abilities and expertise. So before I get stuck trying to make some ill-advised pipe dream work, I wanted to ask how it goes implementing this.
    1)Things that it is good for
    2)info you cannot get from this that one might think they could get.
    3)Configuration issues on the site that might make it hard/impossible to implement (like it is not good for LON, or in systems with a Websup)
    4) common issues that tend to make this all harder than it needs to be.
    5)other analytics software sources that might be a better fit.

    Thanks guys, greatly appreciated

    https://skyfoundry.com/product
    Hmmmm....smells like numbatwo to me.

  2. #2
    Join Date
    Jan 2003
    Location
    USA
    Posts
    4,851
    Post Likes
    Yes
    Propagating the formula. http://www.noagendashow.com/

  3. #3
    Join Date
    Jan 2003
    Location
    USA
    Posts
    4,851
    Post Likes
    Answers to 1-5 could go on for pages... Without knowing how they are connection to your station, type and number of points they want, history vs live values, there isn't any way one can answer this.

    Its good at what it does. It can be setup a million different ways all impacting how its going to effect resources on a station.

    SP thinks they are above all others, yet I have in multiple cases killed their process because it was taking over a server in complete resource hog doing something absolutely retarded.

    I'm sure there are other platforms that can do what they do. It works, I'm I generally impressed...not really.
    Propagating the formula. http://www.noagendashow.com/

  4. #4
    Join Date
    Sep 2007
    Location
    Kenilworth NJ
    Posts
    1,579
    Post Likes
    Thread Starter
    Quote Originally Posted by orion242 View Post
    Its good at what it does.
    SP thinks they are above all others...
    It works, I'm I generally impressed...not really.
    I like unambiguous nature of the post here. Hehe
    Who is SP?
    Sounds like anything Niagara, especially complicated stuff. There is plenty of rope to hang one's self.
    Hmmmm....smells like numbatwo to me.

  5. #5
    Join Date
    Jan 2003
    Location
    USA
    Posts
    4,851
    Post Likes
    Quote Originally Posted by numbawunfela View Post
    Who is SP?
    My slang for Skyspark. SS might be seen as problematic for the triggered these days lol.

    Quote Originally Posted by numbawunfela View Post
    Sounds like anything Niagara, especially complicated stuff. There is plenty of rope to hang one's self.
    Bingo.
    Propagating the formula. http://www.noagendashow.com/

  6. #6
    Join Date
    Nov 2018
    Location
    Los Angeles, CA
    Posts
    44
    Post Likes
    1)It is a good analytics platform.

    2)The hardest part of any analytics system is going to be explaining why the customer would need an analytics program when they can just have alarms. The best way i have found to do this is explaining that analytics are trying to analyze data from over a long period of time, alarms are for things occurring real time. Also, the platform is used to continuously provide verification that your controls sequences are being implemented properly.

    3)The software is pretty flexible in how it gets data, so almost certainly it can be implemented. I have always used it in tandem with Niagara and Alerton.

    4)The Skyspark is not for the feint of heart. They have a proprietary functional programming language that will easily stymie any technician or engineer without any lower level programming experience. Chances are pretty high that the person that is going to implement this is either going to be lacking in the computer based knowledge or the HVAC and energy knowledge, both of which are very important to implementing this properly. If you doubt this guys abilities in any way as you said, then i would strongly recommend not moving forward.

    5)I have heard good things about Niagara's analytics, but nothing first hand.

    Analytics itself has it's own knowledge base and vocabulary that most are not familiar with, especially the people buying it. I find a lot of times it is like handing over a tool to someone that they have never seen or heard of before, and telling them that it will be one of the most useful tools they own. When they can't figure it out they just stop trying to use it and resent you for selling it to them. If the facilities team isn't using their building automation system to it's full capability, then they are almost certainly wasting their money on analytics.

  7. #7
    Join Date
    Sep 2007
    Location
    Kenilworth NJ
    Posts
    1,579
    Post Likes
    Thread Starter
    Thanks for the response!
    How will this affect CPU usage on the JACEs?
    There is 1 Sup and 20 or so JACEs. I am assuming this goes on the Sup and does its thing using historical data there. Histories can be forwarded up to the sup every 15 mins so there is not much of a lag if done that way, and there is **no?** load on the JACEs. Is that right?
    I am sure there are other ways to impliment this, looking for one that does not load subservient devices.
    Hmmmm....smells like numbatwo to me.

  8. #8
    Join Date
    Nov 2018
    Location
    Los Angeles, CA
    Posts
    44
    Post Likes
    Quote Originally Posted by numbawunfela View Post
    Thanks for the response!
    How will this affect CPU usage on the JACEs?
    There is 1 Sup and 20 or so JACEs. I am assuming this goes on the Sup and does its thing using historical data there. Histories can be forwarded up to the sup every 15 mins so there is not much of a lag if done that way, and there is **no?** load on the JACEs. Is that right?
    I am sure there are other ways to impliment this, looking for one that does not load subservient devices.
    It can all be done on the server. It doesn't even have to be done on the same server, just needs to have a connection to the supervisor. Skyspark is pretty flexible on importing histories. You are corect, doing it like this won't increase the load on your JACEs.

  9. #9
    Join Date
    Mar 2006
    Location
    Austin Texas
    Posts
    89
    Post Likes
    As world largest skyspark distributor, I would like to add couple comments on this topic.

    A little history on technology. It is written on top of Fantom lang which is written by Brian and Andy Frank. You could say fantom is bigger brother of the Sedona language that runs on Java. Since sedona was written to be highly efficient on small embedded devices fantom turned out to be highly efficient JVM based language.

    SkySpark has 7 major components. Folio database, Haystack Semantic datamodel, Axon Scripting, Connector Framework, View Framework, Spark and KPI engines and clustering with replication.

    Folio database is written from scratch by our industry genius Brian Frank which the database is highly optimized on time-series data storage. Data storage is for historical data takes a lot less space then SQL DB’s and take a lot less to make the historical data calls.

    All the data stored into folio database within an open source semantic data model called “Project Haystack” Project haystack is semantic data model for internet of things. Semantic data model like html enables us to write and read these posts. Think of project haystack for the internet of things.

    Axon Script is utilized for large datasets to do following.

    Extract transform and load (ETL, I will explain further)
    Making calls to historical database
    Writing Spark and KPI rules
    Creating views and charts

    Connector framework helps us with extracting data from 3rd party sources like BACnet IP, Modbus IP, Greenbutton data, OPC UA, etc. Since it is written with fantom and extensible we had written over 10 connectors for our customers.

    View framework is a wsiwig plus axon scripting generated customizable and cached dashboarding system that enables you to look at semantically modeled data from multiple angles. IE electrical Hvac time span historian or any calculated data.

    With Spark and KPI engines, it generates historical and now live pattern analytics results automatically displayed in organized way. Again haystack enables most of the automation within the views. Sparks check into slices and dices historical data to find anomalies, operational deficiencies which they could lead us to operational issue where it could ended up with equipment failure. With the power of the analytics, we can issues before it happens. Think of your cars dashboard signaling you over gas consumption based on historical data and go change your air filter and check tire pressures. That is what skyspark does. Looks into historical time span as you would like to look into your data within your facilities. Each building is unique so is spark rules. That is the most power part of skyspark.


    Clustering uses dual way secured websockets to communicate instances with each other. Replication takes the historical data from remote sites and creates read only copies. These read-only copies can utilized to speed up historical data call for any part of the application. Analogy of this would be content delivery network. When you go to amazon images are served to you to the closest proximity server. Within replication data results from historical data is returned from the closest replicated database instance to speed up the data processing times and not from remote instances which may be far away from you.

    With all of this knowledge, check open source project Sandstar as well in which it creates hardware agnostic DDC for the first time. Yes we utilized haystack and sedona framework and bundled all of that with skyspark.

    https://vimeo.com/ankabassg

  10. #10
    Join Date
    Mar 2006
    Location
    Austin Texas
    Posts
    89
    Post Likes
    Clustering does not require port forwarding and only northbound communication is good enough to create dual-way data transfer.

    Quote Originally Posted by alper View Post
    As world largest skyspark distributor, I would like to add couple comments on this topic.

    A little history on technology. It is written on top of Fantom lang which is written by Brian and Andy Frank. You could say fantom is bigger brother of the Sedona language that runs on Java. Since sedona was written to be highly efficient on small embedded devices fantom turned out to be highly efficient JVM based language.

    SkySpark has 7 major components. Folio database, Haystack Semantic datamodel, Axon Scripting, Connector Framework, View Framework, Spark and KPI engines and clustering with replication.

    Folio database is written from scratch by our industry genius Brian Frank which the database is highly optimized on time-series data storage. Data storage is for historical data takes a lot less space then SQL DB’s and take a lot less to make the historical data calls.

    All the data stored into folio database within an open source semantic data model called “Project Haystack” Project haystack is semantic data model for internet of things. Semantic data model like html enables us to write and read these posts. Think of project haystack for the internet of things.

    Axon Script is utilized for large datasets to do following.

    Extract transform and load (ETL, I will explain further)
    Making calls to historical database
    Writing Spark and KPI rules
    Creating views and charts

    Connector framework helps us with extracting data from 3rd party sources like BACnet IP, Modbus IP, Greenbutton data, OPC UA, etc. Since it is written with fantom and extensible we had written over 10 connectors for our customers.

    View framework is a wsiwig plus axon scripting generated customizable and cached dashboarding system that enables you to look at semantically modeled data from multiple angles. IE electrical Hvac time span historian or any calculated data.

    With Spark and KPI engines, it generates historical and now live pattern analytics results automatically displayed in organized way. Again haystack enables most of the automation within the views. Sparks check into slices and dices historical data to find anomalies, operational deficiencies which they could lead us to operational issue where it could ended up with equipment failure. With the power of the analytics, we can issues before it happens. Think of your cars dashboard signaling you over gas consumption based on historical data and go change your air filter and check tire pressures. That is what skyspark does. Looks into historical time span as you would like to look into your data within your facilities. Each building is unique so is spark rules. That is the most power part of skyspark.


    Clustering uses dual way secured websockets to communicate instances with each other. Replication takes the historical data from remote sites and creates read only copies. These read-only copies can utilized to speed up historical data call for any part of the application. Analogy of this would be content delivery network. When you go to amazon images are served to you to the closest proximity server. Within replication data results from historical data is returned from the closest replicated database instance to speed up the data processing times and not from remote instances which may be far away from you.

    With all of this knowledge, check open source project Sandstar as well in which it creates hardware agnostic DDC for the first time. Yes we utilized haystack and sedona framework and bundled all of that with skyspark.

    https://vimeo.com/ankabassg

  11. #11
    Join Date
    Sep 2007
    Location
    Kenilworth NJ
    Posts
    1,579
    Post Likes
    Thread Starter
    Quote Originally Posted by alper View Post
    As world largest skyspark distributor, I would like to add couple comments on this topic.

    A little history on technology. It is written on top of Fantom lang which is written by Brian and Andy Frank. You could say fantom is bigger brother of the Sedona language that runs on Java. Since sedona was written to be highly efficient on small embedded devices fantom turned out to be highly efficient JVM based language.

    SkySpark has 7 major components. Folio database, Haystack Semantic datamodel, Axon Scripting, Connector Framework, View Framework, Spark and KPI engines and clustering with replication.
    I actually REALLY appreciate the background - super helpful and makes me look like I know what I am talking about in a room full of people.

    Would you be able to comment directly on the questions raised though?

    Quote Originally Posted by numbawunfela View Post
    This is being suggested by a 3rd party guy with trust from the customer, but I have reason to doubt his abilities and expertise. So before I get stuck trying to make some ill-advised pipe dream work, I wanted to ask how it goes implementing this.
    1)Things that it is good for
    2)info you cannot get from this that one might think they could get.
    3)Configuration issues on the site that might make it hard/impossible to implement (like it is not good for LON, or in systems with a Websup)
    4) common issues that tend to make this all harder than it needs to be.
    5)other analytics software sources that might be a better fit.

    Thanks guys, greatly appreciated

    https://skyfoundry.com/product
    Hmmmm....smells like numbatwo to me.

  12. #12
    Join Date
    Mar 2006
    Location
    Austin Texas
    Posts
    89
    Post Likes
    Quote Originally Posted by numbawunfela View Post
    1)Things that it is good for
    Analyzing, visualizing any data live and historically. Any data being numbers and bools on historical bases.

    Quote Originally Posted by numbawunfela View Post
    2)info you cannot get from this that one might think they could get.
    It is time-span analytics. The more defined the rules that you write, the more specific anomalies you will find. So in this essence, the more thoughtful engineering you put in this, the better it will get. What it can't do is without engineering (Writing rules), it does not tell you outliers that is done with other technologies. When we write the rule, we write it for the specific equipment's parameters. One size fits all situations is good for controlled environments such as wind turbines, substations etc., unfortunately, since humans play a huge factor on a daily operation of the buildings, it does not fit into that category.

    Quote Originally Posted by numbawunfela View Post
    3)Configuration issues on the site that might make it hard/impossible to implement (like it is not good for LON, or in systems with a Websup)

    Niagara with open-source nhaystack is the easiest. Getting data out of any SQL is done with axon scripting, so that is easy as well. The only caveat is that the point has to be trended on the BMS side, whichever BMS it is. It does break the work-flow. Lately getting data BACnet IP is utilized more in the throttled method without hurting the speed of the existing BMS. Some BMS's have APIs, and there are connectors written on StackHub. In the past, we had written a called air hippo for Alerton where we read data on ms access DB and push the data to SkySpark via haystack. Since Alerton moved to SQL on all their instance sales, we stopped maintaining Air Hippo.

    Hardest part of SkySpark was conforming tagging on all points. That took a lot of time. That is going to get only better with haystack 4. If I get permission from SkyFoundry I will post the haystack 4.0 video. It is technical but very valuable. My analog is haystack 4.0 is such. Haystack 1~3 records were at free will. Think of the stars in galaxy but there was no order of gravity. So things could get chaotic if you were not paying attention to tags. There were tooling for us to keep you informed for some rules, but it was not as complete as haystack 4.0. Now haystack 4.0 is gravity added. There is some order to the chaos.


    Quote Originally Posted by numbawunfela View Post
    3)common issues that tend to make this all harder than it needs to be.
    You would write a spark that is too generic without parameters that can lead to nuisance sparks. This problem is also being solved with axon components. In the past, we would add a parameter to function then add a tag to equip for functions parameter. IE discharge CFM range. Each unit can have this different. Having way too many sparks can be another big data problem. So having time-span parameters for each equips lower the number sparks and also their duration, so you get analytics results that taste better.

    Quote Originally Posted by numbawunfela View Post
    5)other analytics software sources that might be a better fit.
    I have been so involved with haystack and skyspark along with project sandstar etc. I did not get a chance to look at any other solution. There are only 70~90 hours workweek work hours.

    Cheers. I hope this answers your questions a little more in-depth.

  13. #13
    Join Date
    Dec 2005
    Location
    California
    Posts
    617
    Post Likes
    I guess this data would be for the ddc impaired individual? Suit's and bean counters maybe...

Page 1 of 2 12 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •