0% found this document useful (0 votes)
173 views8 pages

Data Center - Bicsi

Data Center Design

Uploaded by

mdlateef89
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
173 views8 pages

Data Center - Bicsi

Data Center Design

Uploaded by

mdlateef89
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 8
By David Kozischek As enterprises increasingly decide to outsource some or all of their infrastructure ‘to IT service providers, the result is not at all surprising: fewer data centers overall and hyper-sized facilities in their place. Structured cabling ls defined as bullding cor campus elecommunication cabling infrastructure that consists ofa number of standardized smaler elements (hence, *structured" galled subsystems Forit to be effedve, tuctured cabling is organize in Such away that individual fibers are easy to locate; moves, adds and changes (MACS) are easly managed; and there ts ample srfiow round cabling Perhaps no envionment requires effective structured cabling more than the datacenter With no tolerance for downtime rework alue, data center owners and ‘operators are among the main consumers of traning esouces devoted to stctured cabling. The reason is clear: Even asfewer ‘uadtional datacenters are being bull in favor of outsourcing to the coud (Le, some type of IT service provides there aes physical structures enabling the cloud, and these structures need tobe cabled Facnately what coorstunea effective structured cabling Is not open to Interpretation; athe it is clearly explained In ANSIT1A.942-, Telecmmmunicatins Inftasirace Standard for Data Cente. This article will xplore the standard and break TODS e) GE der Data Centers Consider the different types of data centers in operation today: In-house data centers: Also known as enterprise data centers, these facilities are privately owned by large ‘companies. The company designs, builds and operates its own facility, and can also provide a service for profit such as cloud services or music streaming. Wholesale data centers: (Owned by IT service providers, also known as cloud providers, these data centers are in the business of selling space. Instead of building thelr own facilities, enterprises buy space and deploy their data center infrastructure within the wholesale facility. Colocation data centers: These facilities are like wholesale data centers, but enterprises rent just a rack, cabinet or cage. The IT service provider runs the infrastructure, Dedicated and managed hosting data centers: IT service providers operate and rent server capacity in these facilites, but ‘each enterprise customer controls lts own dedicated server. Shared hosting data centers: In these facilities, enterpris customers buy space on an IT service provider’ servers. These servers are shared among enterprise customers. Today, a significant shift is underway in how these different types of data centers invest in their 8 IcTTODAY Telecom Cloud -eEnterprise $600 jo ge s200 Infrastructure Spending ($bn) 2010 2911 2012 2013 201 FPGURE 1: Growin cou Enterprise = Premises Sm —_ 2 dowd 008 2009 2010 201i 2012 2013 2014 2015 2036 2017 2018 2019 2020 FARE 2: Govt ca sence oie ser toes Infrastructure, LightCounting and their infrastructure to IT service Forbes report that cloud/IT service providers, the result i not at all provider spending is up while surprising: fewer data centers overall enterprise IT spending is down and hypersized facilities in thelr (Figure 1. place (Figure 3 on page 1 Further evidence ofthis shift i The structured cabling require- reflected in DellOro's report of server ments of the resulting hyperscale, investments, the lion's share of nultitenart data centers may which ate shipping for installation differ from what has been installed In cloud-type facilities (Figure 2). in the past in smaller single- As enterprises increasingly tenant, enterprise-owned facilities, decide to outsource some orell of but ANSI/IIA-942-B provides ay Enterprise ‘Contoren ie ee FFRURES: Sit fom enopise TT sovice rier gow guidance. It always recommends a star architecture, with different areas for cross-connecting and interconnecting cable. The standard defines five different cross-connect) interconnect areas, consisting of: ‘= Main distribution areas (MDA). ‘= Intermediate distribution areas. ‘= Horizontal distribution areas (HDA). = Zone distribution areas. Equipment distribution areas EDA). ‘These areas represent the full network from racks and cabinets to the main area where routers, switches and othe: components are located. TIA-922 also provides guidance on redundancy definitions by ranking them into four tiers, called ratings. Rated-1 is the lowest tier with the least redundancy. Rated-4 provides the ‘most redundancy :n a data center's structured cabling and is typically deployed in large M/service provider LOMA Architecture Recommends asta topology architecture Cross-Connect vs. inerconnect | MA, IDA, HDA, ZDA, EDA Redundancy Definitions atod 1-4 Zone Architecture Reduced topologies and consolidated points Enery Eficiency Examples of routing cables and alow contention TABLE 1: Tops Cavered by ANSUIIA 942-8 Tobcommunatons nastuce Seat Dita Cnt 101 IcTTODAY data centers. The other basics covered by this standard include zone architectures and guidelines for energy efficiency (Table 1). ‘When it comes to structured cabling, the standard addresses backbone and horizontal cabling as shown in Figure 4, Each of the distribution areas, or squares, is an area where there isa patch panel and. fiber coming in. How much fiber is needed in each of those areas isa function of network speeds, network architectures, oversubsc:iption and switch configuration. Following are afew examples that illustrate how those considerations affect a data center's fiber count. Figure 5 on page 12shows how network speed influences fiber count asa data center moves from 10 to 100 Gigabit Ethernet (GbE). On the left is the physical architecture with four racks or cabinets each with a switch on top and a switch at the end of the row. In the center Is the logical architecture in TIA-942's recommended star configuration for cabling, and on the right is the network speed. 10 GbE can be supported by only two fbers; 40 GbE can operate over two or eight fibers; and 100 GDE requires two, eight or even 20 fibers, depending on the transceiver. The conclusion, is clear: Network speeds affect fiber count. Check road maps (IEEE for Ethernet and, on the storage side, ANSI for Fibre Channel) for detailed information on per-port fiber counts. Work Areas in tices, Operations center Support Horizontal cating (Offce and Operations (enterLAN sites) Backbone cabin Access Providers Primary Entrance Room (cariereauipment Sand Demaraton) (Routers Backbone ANS Ste, HDA (uayjsanfevms ‘Switches Horizontal cabling | Horizontal cabling Horizontaleabling | Hor ‘Backbone cabing Backbone bing ‘Access Providers ‘Secondary Entrance Room (Carer taupment Sd Demareon) acktone abling IDA (AN/SAN Switches) EDA (Rak/Cabinet DA | [: EDA acta | | aac Each architecture's speed will be constant at 40 GbE, with eight fibers connecting each switch (Figure 6 on page 12). Point-to- point architecture is the simplest —both logically, because itis a star, and physically cabled as a star with eight fibers to each cabinet. A full mesh architecture connects each switch to every other switch, totaling 32 fibers for the same sce antral eairg deuson seas five switches. That logical mesh is “cabled” physically at the cross- ‘connect, and it takes 32 fibers to do that. The final architecture in this ‘example is the spine and leaf, in ‘which every spine switch (Switches and 2) has to connect to every leaf switch (Switches 3 - 5). In the same physical configuration with the same five switches, the spine- aand-leaf logical architecture requires 16 fibers. Depending on the data ‘center's architecture, therefore, an ‘operator may need eight, 16 or 32 fibers for every cabinet. Conclusion: Architecture redundancy increases fiber count. ‘Next, consider how oversubscription impacts fiber count, Oversubscription is the ratio of circuits coming in compared to those going out of a switch. In March/April 2018. | 11 1006 FIOURE 5: Nek spe nonce oe zu a Sone + Loa le Laat oF PURE 6: Net ache ects cout, 12. | ICTTODAY ‘Network speeds affect fiber counts Check IEEE roadmaps {or Ethernet ‘Check ANS! roadmaps for Fibre Channel Gene 2asewes ; eeee p-Z. URE 7: Nobo ooeuberpon not ow count ese IB #h---- fic oe FIGURE @: Never cc contguaton sins ber cou. Gorn ‘Check Swich Morch/Apil 2018) 13, ServersiCompute (cauipment ditbuten area) Data Center ‘Campus Nain Oisttouton ara the example shown in Figure 7 on page 13, the star architecture is used physically and logically with a constant network speed of 10 GbE. The variable shown is the oversubscription rate. The example shows a 4:1 oversubscription with 24 10 GbE circuits coming in and six going out; in the middle, 24 10 GbE circuits come in and 12 go ‘out for a 2:1 rate; at the bottom, the rate is 1:1, with 24 10 GbE circuits both entering and exiting each switch. Depending on the 141 IetTODaY FGURE¢: Large lisence proce ct ore oversubscription rate, with all other variables remaining constant, the required per-switch fiber count can be 12, 24 oF 48 fibers. Conclusion: The lower the oversubscription ration, the higher the fiber count. Ultimately, the oversubscription rate isa function of network ingress/egress traffic needs, meaning the fiber count is duiven by this requirement as well Finally, consider how the network's switch configuration drives fiber count (Figure 8 on page 13). Using constant architectures and running 10 GbE to all of the servers on the racks, what happens ‘on the right side when the switch is reconfigured? At the tap, al of the circuits going dowa are 10 GbE; two of the 40 GbE ports are quad small- form-factor pluggable (QSFP) optical transceivers (Le., eght-fiber multi- fiber push on (MPO] connection); they break out into four 10 GbE to total 16 more ports, yielding two 40 GE ports yolug up, or 2x 8= 16. In the middle of tke figure, the same switch with all four of the 40 GbE ‘ports is seen going back up to the ‘core, equating to 8 x 4 = 32 fibers. “The final scenario shows an equal distribution of 10 GbE going down as going up. 40 GbE ports break out Into 10 GbE for 16 x 10 GbE ports. “Adding more 10 GbE to make It even totals 64 fibers, Conclusion: Just deciding how to configure the switch, affects the fiber count in these scenarios from 16, 32 or 64 fibers. Note that this switching configuration only addresses the Bthernet side of these servers. The fiber count would continue to climb if the servers also had a Fibre Channel network and/or ports for Infiniband higi-speed computing, After seeing how the four ‘variables can independently increase ‘the aumber of fibers needed in data centers, imagine the impact ‘that mixed variables can have in ‘driving fiber counts up even higher. ‘Changing the network's operating. ‘speed affects the fiber count, but ‘what happens when the speed and ‘the architecture are changed? Or ‘the speed and the oversubscription rate? Fiber counts that were already relatively high go up even more. What remains is the question ‘of how to cable this type of data center. Typically, today’s increasingly large data centers extend to separate locations much like an enterprise campus, as shown in Figure 9. Indoor cable is typically used ‘within each building, connected by indoor/outdoor cable and transitional optical splice enclosures (Table 2). Toa een Meet Me room ‘= Demarcation» Gross-conrect, Main cstrbution area 1 Racks/cabnets« Cross-cconect, Indoor cabling 1 Plenum rated ‘Indoot/outdoor cabling 1 Plenumrise « Armored cable Optical splice enclosure (OSE) | « Transition from indoor to autor cables TABLE 2 Fact cating amas ‘There are three deployment methods to consider: ‘+ Preterminated cable: Typically deployed for indoor ples rated cabling, these trunks are factory-terminated on both, ends with eight- or 12-fiber ‘MPO connectors. They are ideal for MDA to HDA or EDA installations involving raceway oor racefloor where the entire fiber count is being dep! FIRE 10: Perinat cable one run at a single location at each end of the link (Figure 10). Pigtailed cable: These semi- ‘num pre-connectorized assemblies are factory-terminated on one end ‘with MPO connectors for easy high-fiber-count deployment while remaining unterminated on the other end to fit through small conduit or allow for on-site length changes. Often ‘used in building-to-building toyed in March/apri2018 15 16 installations, pigtailed cable is ideal for situations when conduit Is too small for pulling grips or the cable pathway cannot be determined before ordering (Figure 11). Bulk cable: This deployment option requires field connect crization on both ends, typically ‘with MPO splice-on connectors (Figure 12). Bulk cable is best for deployments requiring center- pull installation or extremely high fiber counts (such as 1728, fibers and up), FIGURE 11: Pld cc . Fleld-Terminated MTP® Connectors, FIGURE 12:Buk cela, to crecte an A fabric and a B fabric. Figure 13 on page 18 shows how these recommendations translate to a logical architecture Asshown in Figure 14 on Table 3 on page 18 provides an overview of the three deployment methods and their corresponding fiber counts. Putting all of this information to practice, the following example page 19, Cisco's spine-and-leaf illustrates how a four-way spine is cabled and the resulting fiber count. Based on manufacturer recommendations, there are 48 architecture guidance provides for a four-way spine with 48 leaf switches. Starting at the rack and ‘working backward, 32 ports go out leaf switches with 32 ports down ‘to every leaf switch which translates to servers and 32 ports going back up into the fabric. In this example, two Cisco 3064 switches are used at the top of each of the 24 cabinets ‘to 64 fibers required per switch. ‘With two switches on each rack, 128 fibers are needed to support this architecture for every cabinet. IcTTODaAY Seliinonsiive ice Ke 5, LOW Voltage Lighting and SEMAN To Nec Iroc Oy Uneven eCaeoneCt Int eee Se oe ooo bec Po ee Ibol UU! nlermational Standards ra Ca PUT ania eeC ee CABLE METHOD ENVIRONMENT CONNECTOR COUNTS TRUNKTYPE FIBER TYPE | Premises |r weotenro 14,192.26, | tonamows | ana Pigtall rf 283, 432, 576 Singlemode | Pretermirated Indoor/Outdoor ‘MPO connector to fiber 144, 192, 216, ‘Armored Armored | Pigtail | \Won-armored Non-armored | } = [uk a rowtme | wore | Amace | Amores tana | Novem This design called for 10 GbE and a 4:1 oversubscription as previously alscussed, and we will proceed with this example using fiber counts that ae divisible by 12. There are several options for cabling this scenario. Some of them ate not good options like the one depicted in Figure 15 that uses jumpers — over 3,000 of them. A better option would be consolidating jumpers into 48 72-iber cables (Figure 16 on page 20). Better yet is the third option: using high-fber- count trunks, $76 fibers in each ‘one, reducing the number from 3,000 jumpers to six $76sfiber trunks (Figure 17 on page 20) To understand why some of these options are better than others, consider their relative ease of use (which translates to labor spent) both in the intial installation and during the MACs that are inevitable ina data center (Table §on page 21) Furthering the case for high- fiber-count trunks are thelr impact ‘on valuable data center real estate — the pathway for cabling, T1A-S69 provides calculations to understand ‘what percentage of tray/conduit/ raceway is taken up by cabling along 18 1 ICTTODAY TABLE 8: Decoynent ances zt ating cones ‘Cisco Nexus 7000 Cisco Nexus 3064 2 32 Servers {| eee wy) i Oh Ow FIGURE 13: Fou se with a recommendation that the individual maximum fl rationot exceed 25 percent. Though net intuitive, itis a fact that a $0 percent fill ratio actually uses up an entire pathway, because the spaces between cables are part of the equation. With this in mind (Figure 18 on page 21), the first option using more than 3,000 jumpers is no option at all. However, the second cabling option, (48 72-fber trunks) does workin a *=101.6 x 152 millimeters (mm [4 x re ese 6 inch (in)) tray but not quite as well in a =101.6 x 101.6 mm (4 x 4 {n) tray. Both tray sizes can easily accommodate the six 576.fiber trunks option. High-fiber-count trunks can be the best it (when it comes to fill ratios) in today's data centers. The days of trusty 12-and 24-fiber trunks to each rack are no more. Today, data centers ae increasingly growing {in scale and in the fiber counts required to support higher speeds, ‘oN 7000 ear Requied par Suiteh or ‘TRE Tank Cable per Switch FPGURE 14:Spneacel ecto, (oN 7000 GeoNews Sere FIGURE 18: Cling opion March/Apri 2018 119 OT iptaeisy Cra ro OPTION 2 corr) | ‘co Nes 09 ‘oxoNenus 7001 BETWEEN. Bexar rere | Test and dean 6,144 2-Fiber Duplex 576 12-Fber MTP® 576 12-Fer MIP | Lc connectors Connectors Connectors ; suctone Document and label 3,072 Jumpers and 48 Trunks, sicTrunks, 7 ZB : E] oe 6,146 Connectors 576 Connectors | $76 Connectors co ee oe l —= = | e Pull and instal 3,072 Jumpers ot ends) | ° 48 Trunks bth ends) ic Trunks (oth ends) Seon Fa a _ . Purchase 3.072 Jumpers 48 (72 Fiber Trunks) ‘Sx (576 Fiber Trunks) 4 xo Nens a ye ee wee Tratistot—_| 8072Unt,> 6000 Grmectrs | BL 576 Comers | _ Sli 576 Comoros = _ Move, add or change One Jumper ata tne, Create Cross-Conect, Creat Oross-Comect — Or oy OG Point-to-Point Configuration Use Short Jumpor — “ABLES: Delamont itso an urs geo is URE 1: aoa oon 2 a . (el [isal [al ‘Ge eu 7000 ‘so es 09 = ‘00 feted > Badbne os coo | ‘ae i» “ “a ‘ (Be L. | La] » * tenes eee me Ww Ww coeanen URE 18a rs ar anges ibe nt is “Be 2 * ‘greater oversubscription rates, methods and global manufacturers da coils and LAN Hes ches inthe Biever te ° redundant architectures, and creative __ with many years of experience Sock ate Telcormutisios Espns switch configurations. It is clear that solving data center challenges — and a Thelin! Soo x Op Enger large facilities are the new normal; _the assurance of TIA-942 continuing He abopapatsn vous courses icriog Tr ‘enterprise ITcustomers will continue to provide guidance. Inoccton osm Pot Coniguan 00 me +o shift away from small, single- Irieretcrk Deon, LANWAN Furdarens FGURE 17:2ing enon ‘tenant facility operators toward AUTHOR BIOGRAPHY: David Kazsoek Broad Communication Network Dson outsourcing all or part of their data Appian ad Maret Marae Coving ioducen o Newark Proc and asia ‘center infrastructure, Fortunately, Oped Commuricsion, whe eistezensble —_ruadn Des. He can be react there are proven structured cabling i ova mse poston nd protaly i dai shes Dosrng com. 20 1 ICTTODAY Morch/Apri2018 1 21

You might also like