Modular UPS: What’s All the Hype?

by Visitor Siggy62 on ‎06-13-2013 04:27 PM - last edited on ‎08-26-2013 12:25 PM by Administrator

Lou Signorelli Caterpillar Blog.jpg

 

You can find a lot of opinions out there about “modular UPS,” but who can you trust?  While some of the features sound cool enough, I just can’t wrap my head around the impracticality of it all. I don’t understand the notion that an arrangement with ten 20kVA UPS can somehow protect my mission critical applications better than a 2N system, or even an N+1 system with an A-B bus.

 

Looking at the information available, it seems marketers just want us to adopt the concept of “hot-swappable” UPS without question. I don’t know about you, but I want my UPS to work and be properly maintained without the need for swapping anything short of hot and cold air from the data center.

 

All kidding aside, let’s look at the reality of maintenance. Component wear and failures related to fuses can be the biggest maintenance problems for a UPS system. When it comes to component wear and fuses, hot swappable modules are not cost effective because the standby module of the same vintage is just sitting around waiting for something to go wrong. A blown fuse requires the replacement of the entire expensive module rather than the fuse itself.

 

While I’m confident everyone can list a hundred more failure modes, it still doesn’t make sense to invest in modules that do not provide proactive support.  If I can afford redundant modules, why wouldn’t I put them into load protection services?

 

Neither the math of reliability nor availability agrees with the modular mentality.  More modules mean more parts and more opportunities for failure.  Some might argue that the opportunity for failure is offset by the redundancy. However, if you think about this from a customer’s perspective, you’d find that a customer would likely only purchase an N+1 modular system; thus leaving the N as mandatory to support the load.  In this case, your mandatory UPS plant “N” has more parts and more opportunities to fail than a more traditional 2N system with its inherent internal redundancies.

 

I’ve heard the term “pay as you grow”.  However, it seems the modular mentality wants buyers to pay now and then again as they grow.

 

We’d like to hear more about your experience with UPS systems.

  • What kind of UPS system works for your power needs?
  • Have you used or do you currently use a modular UPS system? What has been your experience with it?
  • What can we do to improve our UPS systems and equipment?

Comments
by Regular Contributor
on ‎06-21-2013 09:47 AM

you just found out that the big industry of great managers born in the shadow of the great old that nothing new can do, if not more than the stupid mass production of objects of any function they have in the market, hiding in speudo analysis is done only on numbers and provided they do not reflect reality,
now it's a ups or a microwave oven attitude of production is always the same, the excuse the cost, and if you use more than the products themselves increases the likelihood that the customer does not have a meltdown for which you can calculate the risk, very semplicie,
now in the energy market clearly is that, all at once, low-cost, possibly controlled risk, a real stupidity but it is what you are seeing if you want to see,
I only use it in my small ups for the PC, and high voltage in the cabin, I saw large stations ups in data centers, but do not go beyond that,
my thought is that you must also do the product that serves the customer,
greetings gordini

by New member Ian_K
on ‎06-21-2013 12:07 PM

The weakest link in most UPS systems is the energy storage system, not the power conversion electronics. Lead-acid batteries are subject to failure without warning. Regardless what type of UPS used, battery monitoring is a "must" is lead-acid batteries are employed. Nickel cadmium batteries are much more reliable and long-lived. Mechanical energy storage is also available.

 

The fundamental difference between the two approaches discussed by Lou is that one is based on standard building blocks produced in high volume, while the other is built of high power, semi-custom products built in relatively low volume. The high volume standard building blocks approach allows relatively more design dollars per unit to be invested in design validation such as worst-case analysis than is possible with a low volume product. More design investment and validation time delivers higher reliability. Standardized production processes and more highly automated assembly and test are also possible when producing high volumes. These are a couple reasons that when designed for reliability, high volume and automation frequently deliver superior quality over low volume hand-built equipment.

 

As a proof point, the DC power systems at all modern critical communications sites (central offices, large cellular base stations, small cell sites) employ modular, hot swap architecture. All the older monolithic 2N type power systems are being replaced with the modular systems. Some of these systems deliver thousands of amps. In addition to an inherently robust modular architecture, the large systems employ redundant AC feeds, A and B distribution buses, and redundant distribution circuits.

 

The question of modular redundant architecture (N+1) versus monolithic (N or 2N) has been studied in depth, and considerable quantitative data is available to compare the two schemes. A leading authority on probabilistic risk analysis of various architectures is Steven Fairfax at MTechnology. http://www.mtechnology.net/about_us.htm.

 

by New member waiteave
on ‎06-22-2013 12:27 PM

Lou,

 

Since modular, scalable is not the correct solution, what is?  Your arguement regarding wear components is nullified across the board because all UPS units have wear compnents. Sure modular has more, but they also have one thing that "legacy" UPS do not REDUNDANCY. 

 

I see and hear it everyday in the construction of critical facilities where contractors, engineers and designers state " I have been doing it this way for 30-years and I see no reason to change now."   If using legacy equipment is such a good idea then why are the sale of modular, scalable UPS on the rise? 

If using a hand crank to roll down the windows of our cars in the 70-80's was such a good idea, then why does almost every car today utilize power windows?  In other words, technology changes and technology improves and the failure to truly understand the advantages does a disservice to your clients. 

 

The arguement you make has serious flaws and I would consider a Gartner Report or 451 Group Report to better understand the way Critical Facilities are being constructed today. In fact, the movement to Datacenter 2.0 most likely will make all the current practices obsolete. 

 

by New member bucur
on ‎06-24-2013 02:51 PM

There are not only about technology, or risk of component failure, mass production or managers who want to launch a new item, here, is about functionality and advantages.

One of the advantages is repair time. In case of modularity you are just simply replacing the power module. Some UPS's uses hotswapable, some not, but even so, the time to repair is not more than 5 minutes after you are in the front of the UPS.

Second advantages is footprint. Imagine your load is 200KVA and you are using 2 x 200KVA UPS, each with its own backup battery system to secure, let's say 10 minutes. That is using lot of space. In a datacenter cost per sq meter is a key factor when deciding the equipment to be used.

Another advantages I see is not only pay as you grow, but, let me say "decrease the power as you need". This is when in actual datacenter cloud computing is showing up and the servers are reducing. By using same UPS your load is decreasinjg even to 25%, where the efficiency of the UPS is very low. By takking out one or many power modules, your UPS will be load in the area of good efficiency level and therefore you save on electrical bill.

Many other arguments can be provide. Also is very importat the optimum power modules. Another important topic would be the true modular versus regular modular, what is centralized parallel arhitecture, versus decentralized parallel arhitecture, but to stay on the curent topic, I sumarise that modular is the best choice.

Latest argument that comes to my mind is reducing cost for cabling, fuses, work. Imagine that at load of 200KVA you are using a modular UPS 250KVA set up as N+1 at 200KVA level, with 50KVA power modules, versus 2 units of 200KVA standalone UPS.

by Visitor Siggy62
on ‎06-25-2013 08:36 AM

Glad to see some great discussion.  Just to clear up a few points.

 

Energy storage is not the issue in this blog. There are several alternatives, all with their own positives and negatives.

 

Just to level set a bit, most of the data centers I’m involved with are 1MW and above.  I could see how the small rack mounted “modular” UPS might be a fine fit for small to mid-sized data centers with limited expertise and/or resources to fund proper maintenance and repair.  In those situations “hot swapping” might be the right solution.

 

Ian makes a good point about “mass production” however after touring many manufacturing facilities, I’m not bought into the argument that this makes anything more reliable … only cheaper.

 

Let’s not confuse “modular” with scalable.  10x10kVA UPS to make a N+1 90kVA UPS system is a modular approach. The key difference being that in a modular system you are required to have 9 of 10 functioning modules at all times.  Scalable systems by definition speak to the ability to take on more power (or less power) as needed. Most any UPS today is capable of paralleling for capacity or redundancy as required.  Thus, they are scalable.

 

Redundancy, for me,  is predominantly a system level discussion, not a UPS level view.  Most quality UPS today have redundant power supplies, fans etc… Redundancy becomes critical in the design of the overall power system.  Using the 10X10kVA UPS referenced above, one must recognize this 100kVA modular UPS is at one end a 50kVA solution (N+5) or at the capacity end of the scale a N+1 solution at 90kVA (assumes redundancy is a must).

  

A further look at this solution finds: 10 inverters, 10 rectifiers, 10 sets of inductors etc. Redundancy is defined as “superfluous”.  10x the parts = 10x the likelihood of failure, therefore a single module of redundancy (1 of 10) is not enough.    Where does it end?

 

One reader mentioned floor space as a benefit of the “modular” UPS citing the expense of data center floor space.  If raised floor or data center floor space is so expensive, why are owners using the space for power and not for processing? 

 

Is this a reflection of the size of the data center?  Do small to mid-sized data centers not have the ability to designate an “electrical” room for such equipment?  In the larger data centers I see, the goal is to maximize processing power by only placing servers in the raised floor area.  Occasionally we’ll see a rack mounted PDU, but rarely anything else utilizing the expensive “white” space of the data center.

You may have gotten the idea from my original post that I’m against redundancy.  This is far from the truth.  My methodology for design is simplistic and direct.  2N from Utility to the dual corded server loads.  Undoubtedly, and sadly, it’s the budget that forces most designs to a lesser level.

 

I’m very interested in more opinions and experiences?  Contrary to an earlier comment, market data indicates some growth, but not nearly a wholesale conversion of the market to these little, rack mounted, UPS.

 

1)    Are the customers purchasing these little UPS unique? 

2)    Are there applications where this is a better fit?  Why?

by Visitor essarups
on ‎06-26-2013 12:59 AM

Hi Lou Signorelli !

 

You are right, practical and 2-d-point. We agree with your approach that certainly ensures higher overall performance at ground level for majority of the real world applications. In datacentre design space, the engineers like the kick of multi-variable techno-commercial simulations, super imaginary possibilities while at times they seem to be missing the obvious.

 

The answers to your questions are :

1. Yes, majority of these are unique set of customers looking for unique propositions in terms of the '`gain for fame"' that `We did it differently"' and "gain for achievment" that justifies their multi-variable techno-commercial simulation.

2. Ideally Small Data-centres, Server Racks, Network Racks. Because the user wants the power to be closest to eliminate the possibility of far-ended-distribution issue (s) and have full command control (in all respects- physical, virtual and visual) over every network critical device for his application.

 

Everything happening around is just part of Eternal Evolution.. Cheers !

 

 

Announcements
Help us grow the Caterpillar Community: Invite a Friend

Meet Our Bloggers