Data Center Simulation
We are in the process of evaluating 3 data center simulation tools:
We have 4 DCs with 65K sq ft. of space. Curious to get feedback from any users of the tools (speed, ease of use, validity of results, customer support, etc.).
evaluation of CFD data center tools
I have evaluated all 3 tools in the past in much detail and 6sigma is the only one of these tools that provides the details neccessary for problem resolution.
All the large engineering firms such as Syska and EYP use this application exclusivley because it is an engineering tool not a marketing tool.
You really can't expect to gain any credible or reliable information when your only input is CFM in and CRAC temperature.
The input parameters within 6sigma are extensive and neccessary as well as the ability to to define an object as its true structural demensions. The other prodcuts define items as a black box. with 6sigma an object can be defined just as it would appear in a CAD drawing. This is essential for understanding how air moves through the object and determing leakage areas.
Very simple definable attrubtes can make a tremndous difference in the result, such as the including the propper set of fans for your UPS or the gap beween the rail sets in a cabinet. none of these would be captured when an item is defined as a black box.
The ease of use should never be a question since its the final result that matters. The fundemental piece is the software will be used to make descions on what solutions you will need moving forward to help you achieve greater capacities eficiency reliabilty etc.. and you need to make sure your models will provide you with the correct information.
otherwise its a waste of a license.
Hope that helps.
Data Center CFD Experience
My organization derives 100% of its revenue from data center design services/PM. As part of these service offerings we do CFD analysis using 6sigma. In the last 4 months we have done 5 CFD analysis for various clients.
I can tell you that selecting a CFD analysis tool was a very difficult process. It is absolutely critical that your organization chooses a software that best meets your requirements. I had to weigh a number of items in order to determine which CFD tool licence I should purchase.
Items I evaluated where(there are more, but for simplicity these where the ones I examined, feel free to email directly if you want me to go into greater detail):
1) Licence cost - what financial risk was associated with my licence purchase
2) Resource cost - how much time would it take my staff to complete the work with one licence vs. another?
3) Accuracy - what type of accuracy could I expect to achieve using the various licences?
4) User friendliness - this goes to resource cost...but more importantly how easily could I train my staff to use the program? If one of my team members left, how fast could I get a replacement trained?
5) Customer Support - when I have issues...and believe me you will have many many technical issues...how fast does the liscence manufacturer provide support? How often do they update their software? (CFD softwares are not like microsoft word where they have 1 million users to help them resolve all the glitches, you will have limitations and work arounds that you will need to live with, and rely on the licence providers support to guide you through these issues. All CFD licences have them and the ones that tell you they don't are flat out lying to you. The good providers will fix the problems in their next software update)
I can tell you that I am a mechanical engineer by education. In school I focused on fluid dynamics, and HVAC systems. My first experience with CFD was scholastic. I used a non data center specific platform by ANSYS for some school projects. So other then getting a masters in fluid dynamics, I was well equipped technically to understand the theory behind CFD analysis(Computers and I also get along very well). Because of my scholastic experience, while working at my previous employer, I spearhead a CFD analysis initiative using ANSYS airpak.
2008-2009 - ANSYS Airpak
I selected this software after some minor evaluations, but primarily because it was the name I knew from school. Overall the software was functional. However, it was definitely not for the technically inept. I spent hours, upon hours, trying to correct meshing errors. As an example it would take me a day to build the model, and 3 days to fix the meshing errors. Extremely frustrating. Technical support was very helpful, they typically responded to my emails within a day or 2. Trying to get accurate results was quite challenging. The HVAC systems could not be modeled properly. If the room was symmetrical(both shape and load), it was fine, however, if it wasn't AC units would operate at 200% capacity. Technical support provided me with a work around for the problem, but it did not allow me to model the AC units to the accuracy that I desired. VFD was not an option. They may have resolved these since I used the program, but I know when I was working with them they didn't really see it as a big problem which really frustrated me. In terms of accuracy I would say the best ...absolute best, I could hope to achieve was +/- 10 F on 90% of points. Not really helpful, considering ASHRAES thermal guidelines for recommended IT operating ranges are 64.4-80.6F. In terms of cost they where middle of the road at roughly 30k/annum for a single core licence. On paper they looked great, and their CFD tetrahedral processing engine is definitely superior to any other CFD products on the market but they just aren't set up for data centers yet (I am sure if I was building a jet plane they would be the guys to use). So looking at my grading ruberic, they met criteria 5, and where halfway there on number 1, but didn't pass on the other items.
While I have never explicitly used tile flow I did analyse it while trying to determine which licence to use before and after Airpak. I was primarily drawn to tile flow because of the price point (roughly 13k) I have seen a couple of demos on it, been pitched it by 2 sales guys, and am aware of all the features. The user interface seems good. It has definitely been built for data centers so I imagine it would be easy to use...the problem I have with it is I know, by reading the specs, the level of accuracy that I could achieve wouldn't be any better then Airpak(probably worse). They don't allow you to control the meshing parameters, they don't model the airflow under and above the raised floor as one simulation(hopefully this has changed by now). My catch phrase for CFD is garbage in = garbage out. If you don't have the tools to enter accurate parameters, then you aren't going to get accurate results. Because I didn't use the program I don't know what customer support was like so I can't comment. So looking at my criteria, 1,2, and 4 seem good on paper, 5 no comment, but 3 really suffers. For some users that just want a pretty picture they can show to their boss, or for marketing purposes, this software probably could work for you. But my clients want accuracy. If I don't give them accurate results I have liability issues I need to deal with. Other than hiring a consultant this would be your most affordable solution.
To be honest I never really looked at cool sim in detail so I can't really comment explicitly. Just looking at their website briefly now I see that they use the ANSYS fluent solver which is good. The operating platform and images look identical to that of Airpak. It doesn't appear that you can model the airflow within the cabinets, or on a server by server basis so the accuracy will still suffer. Overall looks like a more user friendly version of Airpak so it appears they have addressed some of the issues I had with items 2 and 4.
For me the biggest irritation that I have with 6sigma is the price tag. 55k/annum is steep, but I understand why they charge what they charge. They are the best, and the only software that I have found which will give me the level of accuracy that I need. When performing a thorough assessment I can get measurements at 90% of points to +/-2F of the actual conditions. This is primarily achieved because they model the airflow within cabinets, they have library's of equipment for CRACs etc from various manufacturers, and...heres the big one, you can model servers on an individual basis. For some clients to reduce costs, I do uniform rack loading(like the other softwares above offer). I have found I can get +/-5F@90% of points when doing it this way. This to me is why they are worth the big $$'s. In terms of problems. They have some issues with creating ducting...there are work arounds but they are very time consuming. They have told me these will be addressed in version 6 which should be released any day(they are about 2 months late on this, but I had similar issues with ANSYS so I don't hold it against them, I'd rather they get it right then rush deployment). Customer support is excellent. They understand the concept of an accelerated deliverable schedule and if I need them to build me a CRAC unit or rack that isn't already in their library, they can do it in as little as one day if I need them too. Another point is because they are so user friendly, I can significantly reduce my resource costs. Comparing the number of hours I spent with 6sigma to Airpak, I find I my resource requirements were reduced by about 60%. This is fantastic, and more than offsets the significant licence costs. This means I can have 1 guy could do 1 CFD analysis/month(size dependent obviously), vs. 1 every 5 months. This translates to a significant increase in revenue and profits for me once I overcome the initial licence fee shock. The ease of use also allows me to train my staff very quickly, which lets me scale up for large projects very well. Overall, number 2 more than offsets number 1, and number 3,4 and 5 are best in industry. Well worth the investment.
Hopefully this helps. Sorry I didnít have time to properly edit it. If you have any more questions feel free to email at email@example.com. I don't check the board very often so this is the best way to reach me and I can post my responses for all readers on here if there is interest.
Just to set the record straight on this, CoolSim does support server level modeling within a rack. We introduced this at CoolSim 3.2 in early 2009. We also support Crac failure analysis, which allows the user to turn off 1-N Crac units per run for up to 10 concurrent runs. And the post is automated so all the result are contained in one viewable HTML doc. And, as correctly pointed out, CoolSim is powered by the ANSYS/Fluent hybrid meshing and advanced solver technology, which is best in class. Finally, CoolSim uses a SaaS model of delivery and is priced proportionate to anticipated annual usage. Why should advanced data center modeling be the exclusive domain of consultants and other "heavy" users? CoolSim is a solid step in the democratization of CFD for "periodic users" with the same needs for functionality as the consulting community.:)
|All times are GMT -4. The time now is 20:12.|