Muna ci gaba da gabatar muku da tsarin haɗin gwiwar HyperFlex na Cisco.
A cikin Afrilu 2019, Cisco yana sake gudanar da jerin zanga-zangar sabuwar hanyar haɗin gwiwa ta Cisco HyperFlex a cikin yankuna na Rasha da Kazakhstan. Kuna iya yin rajista don zanga-zanga ta amfani da fam ɗin martani ta bin hanyar haɗin gwiwa. Shiga mu!
A baya mun buga labarin game da gwajin nauyi da ESG Lab mai zaman kansa yayi a cikin 2017. A cikin 2018, aikin Cisco HyperFlex bayani (version HX 3.0) ya inganta sosai. Bugu da kari, gasa mafita kuma na ci gaba da inganta. Shi ya sa muke buga sabon, sabon sigar ma'auni na damuwa na ESG.
A lokacin rani na 2018, dakin gwaje-gwaje na ESG ya sake kwatanta Cisco HyperFlex tare da masu fafatawa. Yin la'akari da yanayin halin yanzu na amfani da ƙayyadaddun mafita na software, masana'antun dandamali iri ɗaya kuma an ƙara su zuwa nazarin kwatancen.
Gwajin daidaitawa
A matsayin wani ɓangare na gwajin, an kwatanta HyperFlex tare da cikakken tsarin haɗin gwiwar software guda biyu waɗanda aka shigar akan daidaitattun sabar x86, da kuma tare da software guda ɗaya da mafita na hardware. An gudanar da gwaji ta amfani da daidaitattun software don tsarin hyperconverged - HCIBench, wanda ke amfani da kayan aikin Oracle Vdbench kuma yana sarrafa tsarin gwaji. Musamman, HCIBench yana ƙirƙirar injunan kama-da-wane ta atomatik, yana daidaita nauyin da ke tsakanin su kuma yana samar da rahotanni masu dacewa da fahimta.
An ƙirƙiri injunan kama-da-wane 140 kowace tari (35 kowace kullin tari). Kowane injin kama-da-wane ya yi amfani da 4 vCPUs, 4 GB RAM. Fayilolin VM na gida shine 16 GB kuma ƙarin faifan 40 GB.
Ƙididdigar gungu masu zuwa sun shiga cikin gwaji:
- gungu na nodes na Cisco HyperFlex 220C guda huɗu 1 x 400 GB SSD don cache da 6 x 1.2 TB SAS HDD don bayanai;
- mai fafatawa mai siyarwa Tari na nodes huɗu 2 x 400 GB SSD don cache da 4 x 1 TB SATA HDD don bayanai;
- gungun mai siyarwa mai siyarwa B na nodes huɗu 2 x 400 GB SSD don cache da 12 x 1.2 TB SAS HDD don bayanai;
- mai fafatawa C tari na nodes huɗu 4 x 480 GB SSD don cache da 12 x 900 GB SAS HDD don bayanai.
Masu sarrafawa da RAM na duk mafita sun kasance iri ɗaya.
Gwaji don adadin injunan kama-da-wane
An fara gwaji tare da nauyin aiki da aka tsara don yin koyi da daidaitaccen gwajin OLTP: karanta/rubutu (RW) 70%/30%, 100% FullRandom tare da manufa na 800 IOPS a kowane injin kama-da-wane (VM). An gudanar da gwajin akan VM 140 a kowace gungu na tsawon sa'o'i uku zuwa hudu. Makasudin gwajin shine a ci gaba da rubuta latency akan VM da yawa gwargwadon yiwuwa zuwa miliyon 5 ko ƙasa da haka.
Sakamakon gwajin (duba jadawali a ƙasa), HyperFlex shine kawai dandamali wanda ya kammala wannan gwajin tare da farkon 140 VMs kuma tare da latencies ƙasa da 5 ms (4,95 ms). Ga kowane ɗayan gungu, an sake kunna gwajin don gwadawa a daidaita adadin VMs zuwa maƙasudin latti na 5 ms sama da maimaitawa da yawa.
Mai siyarwa A yayi nasarar sarrafa 70 VMs tare da matsakaicin lokacin amsawa na 4,65 ms.
Mai siyarwa B ya sami larurar da ake buƙata na 5,37 ms. kawai tare da 36 VMs.
Mai siyarwa C ya sami damar sarrafa injunan kama-da-wane 48 tare da lokacin amsawa na 5,02 ms
SQL Server Load Emulation
Na gaba, ESG Lab ya kwaikwayi nauyin SQL Server. Gwajin ya yi amfani da girman toshe daban-daban da ƙimar karantawa/rubutu. An kuma gudanar da gwajin akan injuna 140.
Kamar yadda aka nuna a hoton da ke ƙasa, gungu na Cisco HyperFlex ya zarce dillalai A da B a cikin IOPS da kusan ninki biyu, kuma mai siyarwa C fiye da sau biyar. Matsakaicin lokacin amsawar Cisco HyperFlex ya kasance 8,2 ms. Don kwatanta, matsakaicin lokacin amsawa na mai siyarwa A shine 30,6 ms, na mai siyarwa B shine 12,8 ms, kuma na mai siyarwa C shine 10,33 ms.
An yi kallo mai ban sha'awa yayin duk gwaje-gwaje. Mai siyarwa B ya nuna babban bambanci a matsakaicin aiki a cikin IOPS akan VMs daban-daban. Wato, an rarraba nauyin da ba daidai ba, wasu VM sunyi aiki tare da matsakaicin darajar 1000 IOPS +, wasu kuma - tare da darajar IOPS 64. Cisco HyperFlex a cikin wannan yanayin ya fi dacewa da kwanciyar hankali, duk 140 VMs sun sami matsakaicin 600 IOPS daga tsarin tsarin ajiya, wato, nauyin da ke tsakanin injunan kama-da-wane an rarraba su daidai.
Yana da mahimmanci a lura cewa irin wannan rarrabawar IOPS mara daidaituwa a cikin injunan kama-da-wane a mai siyar da B an lura da shi a kowane juzu'in gwaji.
A cikin samarwa na gaske, wannan ɗabi'ar tsarin na iya zama babbar matsala ga masu gudanarwa; a zahiri, injunan kama-da-wane guda ɗaya sun fara daskarewa kuma kusan babu wata hanya ta sarrafa wannan tsari. Hanya ɗaya, ba hanya mai nasara sosai don ɗaukar ma'auni ba, lokacin amfani da mafita daga mai siyarwa B, shine amfani da ɗaya ko wani QoS ko aiwatar da daidaitawa.
ƙarshe
Bari mu yi tunani game da abin da Cisco Hyperflex ke da injunan kama-da-wane 140 a kowace kumburi ta jiki da 1 ko ƙasa da haka don wasu mafita? Don kasuwanci, wannan yana nufin cewa don tallafawa adadin aikace-aikace iri ɗaya akan Hyperflex, kuna buƙatar ƙarancin nodes sau 70 fiye da hanyoyin masu fafatawa, watau. tsarin karshe zai kasance mai rahusa sosai. Idan muka ƙara a nan matakin sarrafa kansa na duk ayyukan don kiyaye hanyar sadarwa, sabobin da dandamali na ajiya HX Data Platform, ya zama bayyananne dalilin da yasa Cisco Hyperflex mafita ke samun karbuwa cikin sauri a kasuwa.
Gabaɗaya, ESG Labs ya tabbatar da cewa Cisco HyperFlex Hybrid HX 3.0 yana ba da sauri da daidaiton aiki fiye da sauran hanyoyin daidaitawa.
A lokaci guda, ƙungiyoyin haɗin gwiwar HyperFlex suma sun kasance a gaban masu fafatawa dangane da IOPS da Latency. Hakanan mahimmanci, an sami aikin HyperFlex tare da kaya mai kyau da aka rarraba a cikin duka ajiya.
Bari mu tunatar da ku cewa za ku iya ganin Cisco Hyperflex bayani kuma tabbatar da iyawar sa a yanzu. Akwai tsarin don nunawa ga kowa:
source: www.habr.com