Compaing Availability of Vaious Rack Powe Redundancy Configuations By Victo Avela White Pape #48
Executive Summay Tansfe switches and dual-path powe distibution to IT equipment ae used to enhance the availability of computing systems. Statistical availability analysis techniques suggest lage diffeences in availability ae expected between the vaious methods commonly employed. This pape examines vaious electical achitectues fo edundancy that ae implemented in today s mission-citical envionments. The availability analyses of these vaious scenaios ae then pefomed and the esults ae pesented. The analysis identifies which appoach povides the best oveall pefomance, and how altenatives compae in pefomance and value. 2
Intoduction Equipment with edundant powe supplies is also efeed to as dual-coded equipment, having edundant powe supplies, each with its own cod. The use of dual-coded equipment is a best pactice that helps maintain optimal powe availability fo the IT equipment and povides the necessay edundancy to pevent downtime fom a single failue within the powe distibution system. This added edundancy also facilitates powe system maintenance. Unfotunately, the majoity of today s mission citical envionments do not fully benefit fom this best pactice. This pape pesents vaious electical achitectue scenaios that may be implemented in today s data cente. The availability analyses of these vaious scenaios ae then pefomed and the esults ae pesented. Appoaches to Distibuting Powe to Racks The following illustations povide an oveview of vaious appoaches fo inceasing availability to ackmounted equipment but can also be applied to stand-alone equipment as well. The diffeent appoaches ae typically selected with the objective of achieving a desied level of availability, with the moe costly appoaches pesumably poviding a highe level of availability. Figues 1 and 2 show how powe is often distibuted within a data cente ack today. Figue 1 Typical ack-mount powe Figue 2 Typical centalized powe Monito Keyboad Seve Seve Stoage Rackmount UPS Powe fom lage 3 phase UPS Monito Keyboad Seve Seve Stoage P o w e S t i p 3
Figue 1 shows a typical ack powe distibution configuation used in small o medium size data centes and wiing closets. This configuation allows fo easily moved acks with intenal UPS battey backup and suge potection. In data centes whee dozens o hundeds of acks ae used, Figue 2 with a lage centalized UPS is a moe common configuation. Thee is no powe edundancy in the powe distibution to the ack in eithe case. Othe electical achitectues use devices to switch fom a pimay powe souce to a seconday powe souce. Two such devices ae a Static Tansfe Switch (STS) and an Automatic Tansfe Switch (ATS). Both of these units ange in size fom about 1kW to ove 1MW. These devices ae discussed in detail in APC White Pape #2: Poweing Single Coded Equipment in a Dual Path Envionment. Examples of both of these switches ae shown below. Rack-mount 3-Phase kva ATS 3-Phase 300kVA STS Figues 3 and 4 demonstate how powe is sometimes distibuted in lage, mission citical facilities. In both cases thee ae two edundant paths leading to an STS, howeve the utility souces feeding the UPS may o may not be edundant, depending on factos such as cost and substation availability fom the utility company. The only diffeence between the two scenaios is that Figue 3 uses a single tansfome downsteam of the static switch while Figue 4 uses edundant tansfomes upsteam of the static switch. Howeve, in both cases, the STS, downsteam subpanel, and associated wiing ae potential single points of failue. These methods povide some edundancy, but the emaining components that have no edundancy pesent failue isk hazads and potential maintenance difficulty. 4
Figue 3 Redundancy to the load with STS Pimay powe path UPS 1 UPS 2 Static Tansfe Switch PDU with STS Step Down P o w e S t i p Monito Keyboad Seve Seve Stoage Backup powe path Figue 4 Redundancy to the load with STS (edundant tansfomes) Pimay powe path UPS 1 UPS 2 1 2 PDU with STS Static Tansfe Switch P o w e S t i p Monito Keyboad Seve Seve Stoage Backup powe path Figues 3 and 4 ae an impovement ove the data cente configuations shown in Figues 1 and 2, but they still do not offe full edundancy to the ack. Although a edundant UPS and tansfome ae added, the static switch, subpanel and thei associated wiing ae single points of failue. 5
Figue 5 addesses the single points of failue limitation found in Figues 3 and 4 by pushing edundancy towads the load. This solution emoves the STS and adds an exta subpanel theeby pushing the edundancy benefits close to the load by means of a Rack Automatic Tansfe Switch (ATS). Any maintenance upsteam of the Rack ATS can now be completed without taking down the load. Although this scenaio exhibits fewe non-edundant components than that of Figues 3 and 4, the Rack ATS emains a single point of failue, as does the equipment s own powe supply. Figue 5 Redundancy to the load with Rack ATS Monito P o w e S t i p Keyboad Seve Seve Stoage Pimay powe path PDU Rackmount Tansfe Switch UPS 1 1 1 Backup powe path UPS 2 2 PDU 2 Figue shows how full edundancy to the load can be achieved using dual-coded equipment with edundant powe supplies. This scenaio has two impotant changes to Figue 5: the Rack ATS is taken out, and dual-coded equipment is used. Full edundancy is now bought staight though to the load. Notice also that an exta powe stip is used to maintain edundancy. This solution is highly available
compaed to those discussed thus fa; howeve, it is also the most expensive solution and can only be used with dual coded equipment expessly designed fo this use. Figue Redundancy to the load with dual-coded equipment P o w e S t i p 1 Monito Keyboad Seve Seve Stoage P o w e S t i p 2 Pimay powe path UPS 1 1 PDU 1 Backup powe path UPS 2 2 PDU 2 The achitectue in Figue 7 combines the achitectues of Figues 5 and, and shows an altenate solution that accommodates both single and dual-coded loads. This solution employs a hybid combination of peviously eviewed designs. Full powe edundancy is maintained fo the dual-coded compute equipment. Fo the single-coded equipment, edundancy is maintained up to the Rack ATS, howeve, the switch and equipment powe supplies ae now single points of failue. Figue 7 also shows added physical sepaation. This is often efeed to as "compatmentalization," whee vaious subsystems within the powe distibution and backup system ae physically sepaated. Physical 7
sepaation, if implemented popely, can pevent an event as seious as a mechanical collapse in one path fom affecting the second path (common cause failue). Figue 7 Redundant achitectue fo single and dual-coded loads Equipment with 2 powe cods P o w e S t i p Monito Keyboad Seve Seve Stoage P o w e S t i p 1 2 Souce 1 UPS 1 1 1 Physical Sepaation Souce 2 UPS 2 2 2 Equipment with 1 powe cod P o w e S t i p Monito Keyboad Seve Seve Stoage Rackmount Tansfe Switch The achitectues discussed in Figues 3, 4, 5 and 7 incopoate tansfe switches. With a lage tansfe switch, one failue can bing down an extemely lage potion of equipment, wheeas failue in a smalle switch will bing down only one ack. Fo some uses, a failue of any one ack has equivalent business consequences as the failue of 50 acks; while fo othes the isolation of a failue to a single ack is an 8
advantage. Fo uses of the latte type, the Rack ATS povides an added availability advantage of fault isolation. Anothe facto to conside is the time needed to epai these switches. A small tansfe switch will not be epaied but eplaced, and can be kept as a spae pat fo a vey quick swap out. In addition, it can be quickly bypassed if needed. A lage switch will need to be epaied and depending on location, will take a few hous to get a epai peson on site. Additional time will be needed to diagnose and epai the system, and if the technician does not have the equied pat, even moe time is lost. Thus, when evaluating some of these moe advanced designs, a vaiety of issues should be evaluated to make an optimal decision. Repai time is consideed in the statistical availability model descibed in the next section. In geneal, equipment with only one powe cod can be a significant liability when tying to develop a high availability mission citical envionment. This is tue not only fo ack-mounted equipment but also fo any mission-citical equipment. Even with the best possible constuction, any single point of failue will fail eventually and esult in downtime. If a tue high availability envionment is equied, single points of failue in the powe distibution must be minimized as much as possible, if not emoved completely. Availability Analysis Appoach An availability analysis is done in ode to quantify the impact of having single vs. dual-coded devices. Five availability analyses ae pefomed: Case 1 Single-coded Load fom Figue 2 Case 2 Single-coded Load with Static Tansfe Switch fom Figue 3 (single tansfome) Case 3 Single-coded Load with Static Tansfe Switch fom Figue 4 (edundant tansfomes) Case 4 Single-coded Load with Rack ATS fom Figue 5 Case 5 Dual-coded Load fom Figue Linea combinatoial analysis, also efeed to as Reliability Block Diagams (RBD), is used to illustate the powe availability at the outlet fo these five configuations. This method of system modeling is the most diect, and woks well fo systems whee thee ae few state tansitions. Linea combinatoial analysis woks by using defined eliability data and then developing a system model that epesents the configuation being analyzed. Because this analysis focuses only on the diffeences between the configuations, it is assumed that eveything upsteam of the UPS system is pefect including utility powe. Theefoe the availabilities pesented hee will be highe then what is expected in an actual installation. The details of the analysis ae povided in the Appendix. 9
Data Used in Analysis Most of the data used to model the components is fom thid paty souces. Data fo the Rack ATS is based on field data fo APC s Rack ATS poduct, which has been on the maket fo appoximately 5 yeas and has a significant installed base. In this analysis the following key components ae included: 1. 2. s 3. UPS systems 4. PDU 5. Static Tansfe Switch (STS). Rack ATS The PDU is boken down into thee basic subcomponents: s, Step-down and. The subpanel is evaluated based on one main beake, one banch cicuit beake and teminations all in seies. The Rack ATS component is used in the fouth case only. The appendix includes the values and souces of failue ate 1 MTTF and ecovey ate 1 MTTR data fo each subcomponent, whee MTTF is the Mean Time To Failue and MTTR is Mean Time To Recove. The failue ates and epai ates used fo the analysis ae povided in the Appendix. Assumptions used in the analysis As with any availability analysis, assumptions must be made to ceate a valid model. Table 1 lists the basic assumptions used in this analysis. 10
Table 1 Assumptions of analysis Assumption Failue Rates of Components Repai Teams System Components Remain Opeating Independence of Failues Failue Rate of Wiing Human Eo Powe Availability is the key measue No benefit of fault isolation Desciption All components in the analysis exhibit a constant failue ate. This is the best assumption, given that the equipment will be used only fo its designed useful life peiod. If poducts wee used beyond thei useful life, then non-lineaity would need to be built into the failue ate. Fo n components in seies it is assumed that n epaipesons ae available. All components within the system ae assumed to emain opeating while failed components ae epaied. These models assume constuction of the descibed achitectues in accodance with Industy Best Pactices. These esult in a vey low likelihood of common cause failues and popagation because of physical and electical isolation. Wiing between the components within the achitectues has not been included in the calculations because wiing has a failue ate too low to pedict with cetainty and statistical elevance. Also pevious wok has shown that such a low failue ate minimally affects the oveall availability. Majo teminations have still been accounted fo. Downtime due to human eo has not been accounted fo in this analysis. Although this is a significant cause of data cente downtime, the focus of these models is to compae powe infastuctue achitectues, and to identify physical weaknesses within those achitectues. In addition, thee exists a lack of data elating to how human eo affects the availability. This analysis povides infomation elated to powe availability. The availability of the business pocess will typically be lowe because the etun of powe does not immediately esult in the etun of business availability. The IT systems typically have a estat time which adds unavailability that is not counted in this analysis The failue of any ack is consideed a failue, and equivalent to the failue of all acks at once. This assumption undestates the advantage of the Cases 4 and 5. Fo some businesses, the failue of a single ack is of less business consequence than the failue of all acks. Results It is impotant to undestand that the objective of this analysis is to compae the theoetical availabilities between cases. Since all components in all five cases shae the same failue ate data the only diffeences between each case ae the quantity, MTTR, and placement of the components. This method povides a vey effective demonstation of availability effectiveness of one achitectue when compaed with anothe. 11
Availability is measued with espect to the outlet(s) supplying powe to the citical load. In evey case, the same component eliability data is used. In Case 1, the failue of any one component in that chain would cause the load to dop. This is a baseline case. In both Cases 2 and 3, any one component fom each edundant path would have to fail simultaneously fo a dopped load to occu. Howeve the failue of any single component downsteam of the STS, including the STS, would also dop the load. The emakable esult in this case is how little the installation of the STS inceases system availability. The eason is that the STS is not significantly moe eliable than the upsteam UPS, and the STS is still a single point of failue. Note futhe that in case 2 the tansfome MTTR minimizes any benefit of the STS. In Case 4, any one component fom each edundant path would have to fail simultaneously fo a dopped load to occu. Despite being a single point of failue, the MTTR of the Rack ATS is small due to the fact that it can quickly be eplaced if a spae is available. The key finding hee is that although the Rack ATS is not necessaily moe eliable than the lage STS, the much lowe MTTR gives it a vey lage availability advantage. In Case 5, any one component fom each edundant path would have to fail simultaneously fo a dopped load to occu. Table 2 is an oveview of the esults of the five availability calculations. Table 2 Summay of availability esults Case Configuation Availability Numbe of 9 s Case 1 Single-coded Load 99.985 % 3.8 Case 2 Single-coded Load with STS (single tansfome) 99.9859 % 3.85 Case 3 Single-coded Load with STS (edundant tansfomes) 99.99715 % 4.5 Case 4 Single-coded Load with Rack ATS 99.999931 %.2 Case 5 Dual-coded Load 99.9999977 % 7. This analysis illustates the significance of dual-coded equipment in attaining high availability within a double-ended electical achitectue. The benefits of such an elaboate design ae not fully ealized with single-coded equipment but come close by implementing a Rack ATS. Fom the esults pesented above it is quite clea that binging edundancy to the load impoves availability. Figue 8 demonstates that even if the eliability (MTTF) of a poduct is inceased 10 fold, it still does not povide the same availability as does using a edundant set at a lowe eliability level. The edundant system is poviding nea 100% availability, o a lage numbe of 9 s. 12
Figue 8 Availability vs. MTTF 0.999995 Oveall System Availability 0.999985 0.999975 0.99995 0.999955 0.999945 0.999935 Stand Alone System Redundant System 0.999925 0.999915 50,000 107,000 14,000 221,000 278,000 335,000 392,000 449,000 50,000 53,000 20,000 77,000 734,000 791,000 MTTF (hous) Conclusions When implementing a high availability achitectue, powe distibution to the ack needs to be caefully consideed. The typical types of powe distibution descibed in this pape vay by a facto of 10,000 in the magnitude of the downtime they ceate. This analysis demonstates vey clealy the impotance of using dual-coded equipment in a citical data cente. The analysis pesented hee suggests a full dual path achitectue can povide up to 10,000 times less down time than a single path design. The common pactice of using tansfe switches to incease the availability of single coded loads povides highly vaiable esults depending on how it is implemented. In some cases, the analysis suggests almost no advantage fom the use of a lage STS. By contast, when the tansfe switch is moved to the ack the system downtime caused by the powe distibution system deceases by a facto of 250. In addition, the Rack based tansfe switch povides additional fault localization since a failue takes out only a single ack. Futhemoe, the Rack based tansfe switch can be deployed as needed and whee needed in a dual path envionment. 13
This data suggests that the common pactice of using lage STS systems to powe single coded loads should be e-evaluated, and that Rack-based tansfe switches have significant advantages fo appoximately simila costs. In geneal, the analysis suggests a geneal pinciple of binging edundancy close to the loads to impove availability. Caeful analysis should always be a peequisite to investing in any high availability system. How much money a custome can justify spending to einfoce thei electical infastuctue detemines which solution to select. A custome must have a clea undestanding of thei business pocesses so that the cost of downtime can be calculated. This cost is what should ultimately dive any investments in availability. About the Autho: Victo Avela is an Availability Enginee fo APC. He is esponsible fo poviding availability consulting and analysis fo clients electical achitectues and data cente design. Victo eceived a Bachelo s degee in Mechanical Engineeing fom Rensselae Polytechnic Institute in 1995 and is a membe of ASHRAE and the Ameican Society fo Quality. 14
Appendix Components and values Component UPS 75kW / 750kVA Static Tansfe Switch (STS) Step-down Failue Rate Recovey Rate Souce of Data 4.0000E-0 0.125 Failue Rate is fom Powe Quality Magazine, Recovey Rate data is based on assumption of 4 hous fo sevice peson to aive, and 4 hous to epai system 4.100E-0 0.17 Godon Associates - Raleigh, NC 7.077E-07 0.0041 MTBF is fom IEEE Gold Book Std 493-1997, Page 40, MTTR is aveage given by Macus Data 3.9954E-07 0.45455 IEEE Gold Book Std 493-1997, Page 40 8.988E-008 0.231 x IEEE value Computed fom value by IEEE Gold Book Std 493-1997, Page 41 8 1.1598E-007 0.231 8 x IEEE value Computed fom value by IEEE Gold Book Std 493-1997, Page 41 Rack ATS 2.0E-0 3 APC Redundant Switch field data Comments Used to supply uninteupted 480 VAC powe to the PDU. Includes contols Used to step down the 480 VAC input to 208 VAC outputs, which is equied fo 120 VAC loads. Used to isolate components fom electical powe fo maintenance o fault containment. Upsteam of the tansfome, one temination exists pe conducto. Since thee ae 2 sets of teminations between components a total of six teminations ae used. Downsteam of the tansfome, one temination exists pe conducto plus the neutal. Since thee ae 2 sets of teminations between components a total of eight teminations ae used. The APC Rack ATS MTTF is calculated to be 1 million hous. A consevative value of 500,000 hous is used. 15
Availability of a Single-coded Load [Case 1] The availability fo a single-coded load, fom Figue 2, is calculated based on the following RBD. Figue 9 epesents the top laye of the RBD, which calculates the steady state availability based on the seies components. This RBD incopoates "expandable" blocks fo the Pats" and the " Pats". An expandable block means that thee s a lowe level RBD that defines its sub-components. Laying the RBD out in this manne facilitates availability calculations. The is used to distibute powe diectly to the citical equipment. The contents of these blocks ae shown in Figues 10 and 11. Figue 9 Single-coded load ë =8.988e-008 ë =8.988e-008 ë =4e-00 ì =0.125 75kW UPS ë =8.988e-008 ë =8.988e-008 Pats Pats Figue 10 pats ë =8.988e-008 ë =7.077e-007 ì =0.0041 Step down ë =1.1598e-007 8 Figue 11 pats ë =1.1598e-007 8 Based on the RBD pesented above, the availability of the single-coded system is shown below. 1
Single-coded load availability [Case 1] Model-Name Availability Unavailability MTTR (hous) MTTF (hous) Annual Downtime (hous) Single-coded Load 99.98498 % 1.5021E-04 19.3 128,5 1.3158 UPS system 99.9940 % 3.5958E-05.5 180291 0.31499 Pats 99.98879 % 1.1205E-04 85.5 73,201 0.98158 Pats 99.99978 % 2.1987E-0 2.4 1,092,825 0.0192 Because the analysis is caied out using data with five significant digits, unavailability is anothe way to expess the esults. Unavailability is simply calculated as (1 Availability). Availability of a Single-coded Load with Static Tansfe Switch (single tansfome) [Case 2] The distibution method fom Figue 3 uses an STS and adds edundancy to eveything upsteam of it except fo the tansfome, which is placed downsteam. The availability fo this scenaio is calculated based on 7 RBD stings that ae boken out fo claity. Figue 12 epesents the top laye of the RBD. The "UPS system" block is a 1 out of 2 block, meaning that all components within that block ae edundant. Figue 13 shows the contents of the "UPS system" block. Figue 12 Single-coded load with STS 1 out of 2 UPS system STS & Distibution Figue 13 UPS system ë =8.988e-008 ë =8.988e-008 ë =4e-00 ì =0.125 75kW UPS ë =8.988e-008 ë =8.988e-008 17
Eveything upsteam of the STS is edundant, howeve, eveything inside the STS & Distibution block, shown in Figue 12 is a single point of failue. The STS & Distibution block contains the STS system, tansfome pats and the subpanel pats as illustated in Figue 14. The STS system is what allows the use of the upsteam edundant components. This system incopoates cicuit beakes, teminations and most impotantly, the Static Tansfe Switch. The RBD fo the STS system is shown in Figue 15. Figue 14 STS & distibution STS system Pats Pats Figue 15 STS system ë =4.1e-00 ì =0.17 Static Switch ë =8.988e-008 The contents of the Pats block and Pats block of Figue 14 ae boken down futhe in Figues 1 and 17 Figue 1 pats ë =8.988e-008 ë =7.077e-007 ì =0.0041 Step down ë =1.1598e-007 8 Figue 17 pats ë =1.1598e-007 8 Based on the RBD diagams pesented above, the availability of the single-coded system with STS and single tansfome is shown in below. 18
Single-coded load with STS availability (single tansfome) [Case 2] Model-Name Availability Unavailability MTTR (hous) MTTF (hous) Annual Downtime (hous) Singled-coded Load with STS 99.9859% 1.4041E-04 20.4 145,513 1.23002 (1 ) UPS system 99.99999987% 1.2930E-09.5 5,025,125,28 0.00001 Single UPS 99.9940% 3.5958E-05.5 180,291 0.31499 STS & Distibution 99.9859% 1.4041E-04 20.4 145,518 1.23001 STS system 99.99738% 2.14E-05 5. 215,214 0.22920 Pats 99.98879% 1.1205E-04 85.53 73,201 0.98158 Pats 99.99978% 2.1987E-0 2.4 1,092,825 0.0192 Availability of a Single-coded Load with Static Tansfe Switch (edundant tansfomes) [Case 3] The distibution method fom Figue 4 uses an STS and adds edundancy to eveything upsteam of it including the tansfome. The availability fo this scenaio is calculated based on 7 RBD stings simila to the pevious analysis. Figue 18 epesents the top laye of the RBD. The "UPS system & " block is a 1 out of 2 block, meaning that all components within that block ae edundant. Figue 19 shows the contents of the "UPS system & " block. The Pats block is composed of the same pats as those of Figue 1. Up to this point evey component has been edundant, howeve, eveything inside the STS & Distibution block, shown in Figue 18, is a single point of failue. Figue 18 Single-coded load with STS 1 out of 2 UPS system & STS & Distibution Figue 19 UPS system & tansfome ë =8.988e-008 ë =8.988e-008 ë =4e-00 ì =0.125 75kW UPS ë =8.988e-008 ë =8.988e-008 Pats 19
In this case the contents of the STS & Distibution block, Figue 20, contains only the STS system and the subpanel pats because the tansfome is pushed upsteam as a edundant component. The STS system in this scenaio is identical to that of Figue 1 except that thee ae 8 teminations athe then as illustated in Figue 21. The components of the Pats block ae identical to those of Figue 17. Figue 20 STS & distibution STS system Pats ë =4.1e-00 ì =0.17 Static Switch Figue 21 STS system ë =1.1598e-007 8 Based on the RBD diagams pesented above, the availability of the single-coded system with STS and edundant tansfomes is shown below. Single-coded load with STS availability (edundant tansfomes) [Case 3] Model-Name Availability Unavailability MTTR (hous) Singled-coded Load with STS (2 s) UPS system & Pats MTTF (hous) 99.99715% 2.8495E-05 5.1 178,839 0.2491 99.9999978% 2.190E-08 21. 985,221,75 0.00019 UPS System 99.9940% 3.5958E-05.5 180,291 0.31499 99.98879% 1.1205E-04 85.5 73,201 0.98158 STS & Distibution 99.99715% 2.8473E-05 5.1 178,872 0.24942 STSsystem 99.99737% 2.274E-05 5. 213,880 0.2301 Pats 99.99978% 2.1987E-0 2.4 1,092,825 0.0192 Annual Downtime (hous) 20
Availability of a Single-coded Load with Rack ATS [Case 4] The analysis fo a single-coded load with a Rack ATS, fom Figue 5, is calculated based on the RBD in Figue 22, which epesents the top laye of the RBD. This model now povides edundancy to the ack howeve the Rack ATS becomes the single point of failue. Figue 23 shows the components of the UPS system & Distibution block. The contents of the Pats and Pats blocks ae identical to those of Figues 1 and 17 espectively. Figue 22 Single-coded load with Rack ATS 1 out of 2 UPS system & Distibution ë =2e-00 ì =3 POU Switch Figue 23 UPS system & distibution ë =8.988e-008 ë =8.988e-008 ë =4e-00 ì =0.125 75kW UPS ë =8.988e-008 ë =8.988e-008 Pats Pats Based on these RBD, the availability of the single-coded system with Rack ATS is shown below. Single-coded load with Rack ATS availability [Case 4] Model-Name Availability Unavailability MTTR (hous) MTTF (hous) Annual Downtime (hous) Single-coded Load with Rack ATS 99.999931 % 3.558950E-07 0.4 499,705 0.0004 UPS system & Distibution 99.999998 % 2.252E-08 19.3 85,898,029 0.00018 Pats 99.98879 % 1.1205E-04 85.5 73,201 0.98158 Pats 99.99978 % 2.1987E-0 2.4 1,092,825 0.0192 Rack ATS 99.999933% 3.3333E-07 0.3 500,000 0.00584 21
In this case, by simply adding anothe PDU, the availability has dastically impoved. Howeve, the Rack ATS is the single point of failue in this system, limiting the oveall availability to six "9 s". Because of this, a Rack ATS should always be chosen based on its eliability, and spaes should always be kept on site to minimize MTTR. Availability of a Dual-coded Load [Case 5] The analysis fo a dual-coded load, fom Figue, is calculated based on the RBD in Figue 24, which again epesents the top laye. Like the system with Rack ATS, this RBD calculates the steady state availability based on the oveall UPS and PDU failue and ecovey ates, howeve, it does not include the Rack ATS because the load is dual coded and can fully utilize the edundant paths. Only 1 of the 2 paths must be opeational to maintain the citical loads. Thee ae no single points of failue in this system. As a matte of fact, even the citical load powe supplies ae edundant. Figue 24 Dual-coded load 1 out of 2 UPS system & Distibution The lowe level RBD that the "UPS system & Distibution" block is composed of ae identical to those in Figues 9 11. Based on these blocks, the availability of the dual-coded system is shown below. Dual-coded load availability [Case 5] Model-Name Availability Unavailability MTTR (hous) MTTF (hous) Dual-coded Load 99.9999977 % 2.252E-08 19.3 85,898,029 0.000197 UPS system & Distibution Pats Pats 99.9999977 % 2.252E-08 19.3 85,898,029 0.000197 99.98879 % 1.1205E-04 85.5 73,201 0.98158 99.99978 % 2.1987E-0 2.4 1,092,825 0.0192 Annual Downtime (hous) In this final case, the UPS system & Distibution availability is identical to the pevious case yet oveall availability has inceased to seven 9 s. The majo diffeence is that the Rack ATS is no longe needed when using dual-coded equipment. As is shown in the last system, the Rack ATS is a single point of failue and limited the availability to six 9 s. 22