{"id":70,"date":"2025-07-02T12:19:00","date_gmt":"2025-07-02T09:19:00","guid":{"rendered":"https:\/\/www.ertanerbek.com\/?p=70"},"modified":"2025-07-02T17:07:59","modified_gmt":"2025-07-02T14:07:59","slug":"dogru-bir-sanallastirma-sistemi-nasil-tasarlanmalidir","status":"publish","type":"post","link":"https:\/\/www.ertanerbek.com\/?p=70","title":{"rendered":"Do\u011fru Bir Sanalla\u015ft\u0131rma Sistemi Nas\u0131l Tasarlanmal\u0131d\u0131r?"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma sistemleri, fiziksel donan\u0131m kaynaklar\u0131n\u0131n daha verimli kullan\u0131lmas\u0131n\u0131 sa\u011flamak amac\u0131yla geli\u015ftirilmi\u015f altyap\u0131lard\u0131r. Temel hedefleri, kaynak israf\u0131n\u0131 \u00f6nlemek, servislerin yal\u0131t\u0131m\u0131n\u0131 sa\u011flamak ve sistem s\u00fcreklili\u011fini garanti alt\u0131na almakt\u0131r. Bu ba\u011flamda, farkl\u0131 servislerin birbirini etkilemesini engellemek ve donan\u0131m s\u00fcr\u00fcc\u00fcleri gibi bile\u015fenlerden do\u011fabilecek sistemsel karars\u0131zl\u0131klar\u0131 azaltmak m\u00fcmk\u00fcnd\u00fcr. \u00d6zellikle servislerin kesintisiz \u00e7al\u0131\u015fmas\u0131 a\u00e7\u0131s\u0131ndan sanalla\u015ft\u0131rma altyap\u0131lar\u0131 \u00f6nemli avantajlar sa\u011flar.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma sistemlerinin yaln\u0131zca kaynak payla\u015f\u0131m\u0131n\u0131 sa\u011flamas\u0131 de\u011fil, ayn\u0131 zamanda servis s\u00fcreklili\u011fi perspektifiyle de kurgulanmas\u0131 gerekir. Bu noktada y\u00fcksek eri\u015filebilirlik kavram\u0131 \u00f6ne \u00e7\u0131kar. Cluster yap\u0131lar\u0131 ile entegre \u00e7al\u0131\u015fan sanalla\u015ft\u0131rma altyap\u0131lar\u0131nda, sistemin belirli bir par\u00e7as\u0131n\u0131n devre d\u0131\u015f\u0131 kalmas\u0131 durumunda geri kalan bile\u015fenlerin hizmetin kesintiye u\u011framadan devam etmesini sa\u011flayacak yeterlilikte olmas\u0131 beklenir. Bu nedenle kaynak planlamas\u0131 yap\u0131l\u0131rken, cluster yap\u0131s\u0131n\u0131n node say\u0131s\u0131 dikkate al\u0131narak rezerv kapasite hesaplanmal\u0131d\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u0130ki node&#8217;lu bir yap\u0131da, sistemin %50\u2019sinden fazlas\u0131n\u0131n tek bir node \u00fczerine ta\u015f\u0131nmas\u0131 durumunda kaynak darbo\u011faz\u0131 ya\u015fanmas\u0131 olas\u0131d\u0131r. Benzer \u015fekilde \u00fc\u00e7 node&#8217;lu bir yap\u0131da teorik olarak iki node\u2019un t\u00fcm sistemi \u00e7al\u0131\u015ft\u0131rmas\u0131 gerekir ve bu durumda kaynak planlamas\u0131 sistem y\u00fck\u00fcn\u00fcn maksimum %66\u2019s\u0131n\u0131 kapsayacak \u015fekilde yap\u0131lmal\u0131d\u0131r. Bu oran d\u00f6rt node i\u00e7in %75, be\u015f node i\u00e7in %80, alt\u0131 node i\u00e7in ise %83 olarak hesaplanabilir. Bu de\u011ferler, cluster yap\u0131lar\u0131nda \u201cN-1 tolerans\u0131\u201d prensibiyle belirlenir ve y\u00fcksek eri\u015filebilirlik hedeflerinin sa\u011flanmas\u0131nda temel kabul edilir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma ortamlar\u0131nda kullan\u0131lan baz\u0131 hiper y\u00f6neticiler (hypervisor) kaynak kullan\u0131m verimlili\u011fini art\u0131rmaya y\u00f6nelik baz\u0131 yaz\u0131l\u0131m d\u00fczeyi \u00f6zellikler sunar. Bunlar aras\u0131nda memory ballooning, memory deduplication, zRam, zSwap, FlasCache, FlashSwap, CPU affinity, CPU limit gibi mekanizmalar \u00f6ne \u00e7\u0131kar. Bu \u00f6zellikler sayesinde kaynaklar dinamik olarak da\u011f\u0131t\u0131labilir ve a\u015f\u0131r\u0131 y\u00fcklenme durumlar\u0131nda sistem performans\u0131 d\u00fc\u015fmeden servislerin ayakta kalmas\u0131 sa\u011flanabilir. \u00d6rne\u011fin iki node\u2019lu bir yap\u0131, teorik olarak tam y\u00fck devri i\u00e7in %50 kapasite kullan\u0131m\u0131yla s\u0131n\u0131rland\u0131r\u0131lmal\u0131 olsa da, yukar\u0131da belirtilen \u00f6zelliklerin etkin kullan\u0131m\u0131 ile bu s\u0131n\u0131r %75 seviyesine kadar esnetilebilir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>N Nodelu Cluster Mimarilerinde Y\u00fck Planlama Tablosu (N-1 Kural\u0131)<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><td><strong>Toplam Node Say\u0131s\u0131<\/strong><\/td><td><strong>Maksimum Y\u00fckleme Oran\u0131 (%)<\/strong><\/td><td><strong>A\u00e7\u0131klama<\/strong><\/td><\/tr><\/thead><tbody><tr><td>2 node<\/td><td>%50<\/td><td>1 node ar\u0131zas\u0131nda di\u011fer node t\u00fcm y\u00fck\u00fc ta\u015f\u0131mal\u0131<\/td><\/tr><tr><td>3 node<\/td><td>%66<\/td><td>2 node ile sistem devam etmeli<\/td><\/tr><tr><td>4 node<\/td><td>%75<\/td><td>3 node ile t\u00fcm y\u00fck ta\u015f\u0131nabilir olmal\u0131<\/td><\/tr><tr><td>5 node<\/td><td>%80<\/td><td>4 node aktif kald\u0131\u011f\u0131nda sistem kesintisiz olmal\u0131<\/td><\/tr><tr><td>6 node<\/td><td>%83<\/td><td>5 node yeterli olmal\u0131, kaynak kullan\u0131m\u0131 dengelenmi\u015f<\/td><\/tr><tr><td>7 node<\/td><td>%85<\/td><td>Kaynak verimlili\u011fi artar, esneklik y\u00fckselir<\/td><\/tr><tr><td>8 node<\/td><td>%87<\/td><td>HA kapsam\u0131 geni\u015fler<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Bu oranlar, \u201cen k\u00f6t\u00fc durumda bir node kayb\u0131\u201d (N-1) senaryosuna g\u00f6re planlanm\u0131\u015ft\u0131r. Daha fazla node kayb\u0131 \u00f6ng\u00f6r\u00fcl\u00fcyorsa (\u00f6rne\u011fin N-2), oranlar d\u00fc\u015f\u00fcr\u00fclmelidir.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Ger\u00e7ek\u00e7i \u00d6rnek: 4 Node\u2019lu Bir Sanalla\u015ft\u0131rma K\u00fcmesinde Kaynak Planlamas\u0131<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Toplam Node:<\/strong> 4<\/li>\n\n\n\n<li><strong>Her Node:<\/strong> 2 x Xeon E5-2660 v4 (14 core) = 28 fiziksel \u00e7ekirdek<\/li>\n\n\n\n<li><strong>Toplam Fiziksel \u00c7ekirdek (4 node):<\/strong> 112 core<\/li>\n\n\n\n<li><strong>Hyper-threading ile Toplam vCPU:<\/strong> 224<\/li>\n\n\n\n<li><strong>Toplam RAM (\u00f6rne\u011fin):<\/strong> 4 x 256 GB = 1024 GB<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;N-1 Kural\u0131 Gere\u011fi Kullan\u0131labilir Maksimum Y\u00fck:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>vCPU kapasitesi: 224 \u00d7 0.75 = <strong>168 vCPU<\/strong><\/li>\n\n\n\n<li>RAM kapasitesi: 1024 GB \u00d7 0.75 = <strong>768 GB<\/strong><\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Sistemin en fazla bu kaynak y\u00fck\u00fcn\u00fc a\u015fmayacak \u015fekilde VM\u2019lerle yap\u0131land\u0131r\u0131lmas\u0131, olas\u0131 node kayb\u0131nda sistemin ayakta kalmas\u0131n\u0131 garanti eder.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\ud83d\udd01<\/strong><strong> Dinamik Kaynak Y\u00f6netimi Aktifse (Memory Ballooning, CPU Limit vb.)<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Baz\u0131 hiper y\u00f6neticilerle birlikte bu oranlar daha esnek olabilir. \u00d6rne\u011fin, yukar\u0131daki sistemde teorik %75 s\u0131n\u0131r\u0131, g\u00f6zlemlenmi\u015f y\u00fckler ve dinamik bellek y\u00f6netimi sayesinde <strong>%85\u201390\u2019a kadar<\/strong> \u00e7\u0131kar\u0131labilir. Fakat bu durumda:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Y\u00fck profili s\u00fcrekli izlenmeli<\/li>\n\n\n\n<li>Overcommit oran\u0131 kontroll\u00fc tutulmal\u0131<\/li>\n\n\n\n<li>VM&#8217;ler kritik \u00f6nceliklerine g\u00f6re grupland\u0131r\u0131lmal\u0131<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma sistemlerinde ba\u015far\u0131m kadar deterministik davran\u0131\u015f analizi de \u00f6nemlidir. \u00d6zellikle a\u015f\u0131r\u0131 y\u00fck, node kayb\u0131, donan\u0131m ar\u0131zas\u0131 gibi senaryolarda sistemin nas\u0131l tepki verdi\u011finin bilinmesi, ileriye d\u00f6n\u00fck kapasite planlamas\u0131 a\u00e7\u0131s\u0131ndan de\u011ferlidir. Donan\u0131m uyumlulu\u011fu, a\u011f segmentasyonu, depolama yap\u0131lar\u0131n\u0131n niteli\u011fi ve yedekleme stratejileri de bir sanalla\u015ft\u0131rma sisteminin do\u011fru \u015fekilde yap\u0131land\u0131r\u0131lmas\u0131nda g\u00f6z \u00f6n\u00fcnde bulundurulmas\u0131 gereken di\u011fer ba\u015fl\u0131klard\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sonu\u00e7 olarak do\u011fru bir sanalla\u015ft\u0131rma altyap\u0131s\u0131, sadece misafir i\u015fletim sistemlerini \u00e7al\u0131\u015ft\u0131rmakla kalmaz; i\u015f y\u00fcklerinin s\u00fcreklili\u011fini sa\u011flayacak \u015fekilde tasarlanmal\u0131, olas\u0131 ar\u0131za senaryolar\u0131nda sistemin ayakta kalmas\u0131n\u0131 garanti edecek kaynak rezervleri ve mekanizmalar bar\u0131nd\u0131rmal\u0131d\u0131r. Bu ba\u011flamda sistemin tasar\u0131m\u0131; donan\u0131m, yaz\u0131l\u0131m ve operasyonel y\u00f6netim perspektifinde b\u00fct\u00fcnsel olarak ele al\u0131nmal\u0131d\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma altyap\u0131lar\u0131nda kaynak planlamas\u0131 yap\u0131l\u0131rken yaln\u0131zca toplam CPU ve RAM miktar\u0131na odaklanmak yeterli de\u011fildir. \u00d6zellikle sistemin efektif \u00e7al\u0131\u015fabilmesi i\u00e7in CPU-RAM oran\u0131, depolama mimarisi ve kullan\u0131lan teknolojilerin karakteristikleri birlikte de\u011ferlendirilmelidir. Bu ba\u011flamda, \u00fcretici firmalar taraf\u0131ndan \u00f6nerilen baz\u0131 oranlar ve yap\u0131land\u0131rma prensipleri, sistemin kararl\u0131l\u0131\u011f\u0131 ve performans\u0131 a\u00e7\u0131s\u0131ndan kritik \u00f6neme sahiptir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u00d6ncelikle, sanalla\u015ft\u0131rma sistemine eklenen RAM miktar\u0131n\u0131n, sistem performans\u0131n\u0131 do\u011frudan art\u0131raca\u011f\u0131 varsay\u0131m\u0131 her zaman ge\u00e7erli de\u011fildir. RAM\u2019in efektif kullan\u0131m\u0131, CPU kapasitesiyle do\u011frudan ili\u015fkilidir. \u00d6rne\u011fin, VMware\u2019in teknik d\u00f6k\u00fcmantasyonlar\u0131nda da belirtildi\u011fi \u00fczere, fiziksel CPU ba\u015f\u0131na d\u00fc\u015fen sanal CPU (vCPU) ve RAM miktarlar\u0131 dengeli olmal\u0131d\u0131r. Aksi takdirde, CPU darbo\u011faz\u0131 olu\u015fabilir ve bellek kaynaklar\u0131 etkin kullan\u0131lamaz hale gelir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Depolama mimarisi de bu dengeyi do\u011frudan etkileyen bir di\u011fer fakt\u00f6rd\u00fcr. Geleneksel depolama sistemlerinde (\u00f6rne\u011fin DAS veya klasik RAID yap\u0131lar\u0131), CPU ve RAM kaynaklar\u0131n\u0131n disk I\/O performans\u0131 \u00fczerindeki etkisi s\u0131n\u0131rl\u0131d\u0131r. Ancak yaz\u0131l\u0131m tan\u0131ml\u0131 depolama (SDS) \u00e7\u00f6z\u00fcmlerinde durum farkl\u0131d\u0131r. Ceph, GlusterFS, vSAN gibi SDS sistemlerinde veri replikasyonu, hata tolerans\u0131, metadata y\u00f6netimi gibi i\u015flemler do\u011frudan CPU ve RAM kaynaklar\u0131yla ger\u00e7ekle\u015ftirilir. Bu nedenle, SDS mimarilerinde disk ba\u015f\u0131na CPU ve RAM gereksinimleri daha belirgindir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Genel saha tecr\u00fcbelerine ve \u00fcretici \u00f6nerilerine g\u00f6re:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Disk ba\u015f\u0131na 1 fiziksel CPU \u00e7ekirde\u011fi<\/strong> \u00f6nerilir (\u00f6zellikle Ceph gibi replikasyon yapan sistemlerde).<\/li>\n\n\n\n<li><strong>Disk ba\u015f\u0131na 1\u20134 GB RAM<\/strong> aras\u0131 bellek tahsisi yap\u0131lmal\u0131d\u0131r. E\u011fer veri s\u0131k\u0131\u015ft\u0131rma ve tekille\u015ftirme gibi i\u015flemler aktif de\u011filse, bu oran genellikle 1:1 seviyesinde tutulabilir. Ancak bu t\u00fcr i\u015flemler aktifse, RAM ihtiyac\u0131 artar.<\/li>\n\n\n\n<li>\u00d6rne\u011fin, 12 diskli bir SDS nodunda, minimum 12 \u00e7ekirdek CPU ve 48 GB RAM yaln\u0131zca depolama servisleri i\u00e7in ayr\u0131lmal\u0131d\u0131r. Bu kaynaklar, sanal makineler i\u00e7in kullan\u0131lacak toplam kapasitenin d\u0131\u015f\u0131nda planlanmal\u0131d\u0131r.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Bu t\u00fcr planlamalar yap\u0131l\u0131rken, sistemin yaln\u0131zca anl\u0131k performans\u0131 de\u011fil, ayn\u0131 zamanda servis s\u00fcreklili\u011fi ve hata tolerans\u0131 da g\u00f6z \u00f6n\u00fcnde bulundurulmal\u0131d\u0131r. \u00d6zellikle y\u00fcksek eri\u015filebilirlik hedeflenen yap\u0131larda, SDS sisteminin kendi i\u00e7indeki replikasyon ve iyile\u015ftirme s\u00fcre\u00e7leri, CPU ve RAM kaynaklar\u0131n\u0131 yo\u011fun \u015fekilde kullanabilir. Bu nedenle, bu kaynaklar\u0131n sanal makinelerle payla\u015f\u0131lmas\u0131 \u00f6nerilmez.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sonu\u00e7 olarak, sanalla\u015ft\u0131rma altyap\u0131lar\u0131nda CPU, RAM ve disk kaynaklar\u0131 aras\u0131ndaki ili\u015fki, kullan\u0131lan depolama mimarisi ve sistemin i\u015flevsel gereksinimlerine g\u00f6re dikkatle planlanmal\u0131d\u0131r. \u00dcretici dok\u00fcmantasyonlar\u0131 ve saha testleri, bu planlaman\u0131n temel referans noktalar\u0131 olmal\u0131d\u0131r. Aksi takdirde, sistemin kararl\u0131l\u0131\u011f\u0131 ve hizmet kalitesi olumsuz etkilenebilir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A\u015fa\u011f\u0131da, \u00f6nde gelen sanalla\u015ft\u0131rma ve altyap\u0131 \u00fcreticilerinin teknik dok\u00fcmantasyonlar\u0131nda yer alan CPU\/RAM oranlar\u0131na ili\u015fkin \u00f6nerileri kar\u015f\u0131la\u015ft\u0131rmal\u0131 olarak sunuyorum. Bu oranlar, sistemin kararl\u0131 \u00e7al\u0131\u015fmas\u0131, kaynaklar\u0131n dengeli kullan\u0131m\u0131 ve performans optimizasyonu a\u00e7\u0131s\u0131ndan \u00fcretici tavsiyelerine dayanmaktad\u0131r.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Sanalla\u015ft\u0131rma Sistemlerinde \u00dcretici Bazl\u0131 CPU\/RAM Oranlar\u0131<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><td><strong>\u00dcretici \/ Platform<\/strong><\/td><td><strong>\u00d6nerilen CPU:RAM Oran\u0131<\/strong><\/td><td><strong>A\u00e7\u0131klama<\/strong><\/td><td><strong>Kaynak<\/strong><\/td><\/tr><\/thead><tbody><tr><td><strong>VMware vSphere 8.0<\/strong><\/td><td>1 vCPU : 4\u20136 GB RAM<\/td><td>Genel i\u015f y\u00fckleri i\u00e7in \u00f6nerilen ba\u015flang\u0131\u00e7 noktas\u0131d\u0131r. Hafif y\u00fcklerde 1:2 de m\u00fcmk\u00fcnd\u00fcr.<\/td><td><a href=\"https:\/\/www.vmware.com\/docs\/vsphere-esxi-vcenter-server-80-performance-best-practices\">VMware vSphere 8.0 Performance Best Practices<\/a><\/td><\/tr><tr><td><strong>OpenStack (Nova)<\/strong><\/td><td>1 fiziksel CPU : 16 vCPU<\/td><td>CPU overcommit oran\u0131 16:1 olarak tan\u0131mlanm\u0131\u015ft\u0131r. RAM i\u00e7in \u00f6nerilen overcommit oran\u0131 1.5:1\u2019dir.<\/td><td><a href=\"https:\/\/docs.openstack.org\/arch-design\/design-compute\/design-compute-overcommit.html\">OpenStack Architecture Design Guide<\/a><\/td><\/tr><tr><td><strong>Microsoft Hyper-V<\/strong><\/td><td>1 vCPU : 4 GB RAM<\/td><td>Dynamic Memory ile esneklik sa\u011flansa da ba\u015flang\u0131\u00e7 yap\u0131land\u0131rmas\u0131 i\u00e7in bu oran \u00f6nerilir.<\/td><td><a href=\"https:\/\/learn.microsoft.com\/en-us\/windows-server\/virtualization\/hyper-v\/hyper-v-technology-overview\">Microsoft Hyper-V Planning Docs<\/a><\/td><\/tr><tr><td><strong>Red Hat Virtualization \/ KVM<\/strong><\/td><td>1 vCPU : 2\u20136 GB RAM<\/td><td>NUMA uyumu ve hugepages kullan\u0131m\u0131yla birlikte bu aral\u0131k \u00f6nerilmektedir.<\/td><td><a href=\"https:\/\/access.redhat.com\/documentation\/en-us\/red_hat_virtualization\/4.4\/html\/planning_and_prerequisites_guide\/index\">Red Hat Virtualization Planning Guide<\/a><\/td><\/tr><tr><td><strong>Nokia NSP (VM ortam\u0131)<\/strong><\/td><td>vCPU ba\u015f\u0131na 2\u20134 GB RAM<\/td><td>Sanalla\u015ft\u0131r\u0131lm\u0131\u015f NSP bile\u015fenleri i\u00e7in kaynaklar ayr\u0131lmal\u0131, oversubscription \u00f6nerilmez.<\/td><td><a href=\"https:\/\/documentation.nokia.com\/nsp\/24-8\/NSP_Planning_Guide\/ai8g83isoq.html\">Nokia NSP Planning Guide<\/a><\/td><\/tr><tr><td><strong>Citrix XenServer<\/strong><\/td><td>1 vCPU : 2\u20134 GB RAM<\/td><td>Hafif i\u015f y\u00fckleri i\u00e7in 1:2, yo\u011fun uygulamalar i\u00e7in 1:4 \u00f6nerilir.<\/td><td><a href=\"https:\/\/docs.citrix.com\/en-us\/citrix-virtual-apps-desktops\/install-configure\/sizing.html\">Citrix Virtual Apps and Desktops Best Practices<\/a><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Bu oranlar, sistemin i\u015f y\u00fck\u00fcne, hypervisor mimarisine ve donan\u0131m altyap\u0131s\u0131na g\u00f6re de\u011fi\u015fiklik g\u00f6sterebilir. \u00d6zellikle SDS, VDI veya HPC gibi \u00f6zel senaryolarda \u00fcretici belgeleri dikkatle incelenmeli ve kaynak planlamas\u0131 bu do\u011frultuda yap\u0131lmal\u0131d\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong><u>NOT : Bu de\u011ferler \u00fcreticilerde versiyondan versiyona geli\u015fen teknoloji ile de\u011fi\u015fiklik g\u00f6stermektedir. L\u00fctfen bu de\u011ferleri ana baz de\u011ferleriniz olarak kabul etmeyin her versiyon i\u00e7in ayr\u0131ca inceleyiniz. Tablodaki veriler olabildi\u011fince \u00fcreticilerden toplamam kar\u015f\u0131n versiyonlarda farkl\u0131l\u0131klar bulunmaktad\u0131r.<\/u><\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma altyap\u0131lar\u0131nda kar\u015f\u0131la\u015f\u0131lan bir\u00e7ok performans ve eri\u015filebilirlik sorununun temelinde, a\u011f (network) mimarisinin yanl\u0131\u015f tasarlanm\u0131\u015f olmas\u0131 yer almaktad\u0131r. Bu t\u00fcr sistemlerde yaz\u0131l\u0131msal \u00e7\u00f6z\u00fcmler her ne kadar geli\u015fmi\u015f olsa da, fiziksel altyap\u0131n\u0131n do\u011fru planlanmamas\u0131 performans d\u00fc\u015f\u00fckl\u00fc\u011f\u00fc, beklenmeyen kesintiler ve hizmet kalitesi bozulmalar\u0131na neden olabilmektedir.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong>1. Y\u00f6netim A\u011f\u0131 (Management Network)<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma platformlar\u0131n\u0131n kontrol ara y\u00fcz\u00fcd\u00fcr. Genellikle hiper y\u00f6netici (hypervisor) platformunun web veya komut sat\u0131r\u0131 eri\u015fim noktas\u0131d\u0131r. Bu a\u011f \u00fczerinden yaln\u0131zca sistem y\u00f6neticileri eri\u015fim sa\u011flar gibi d\u00fc\u015f\u00fcn\u00fclse de, bir\u00e7ok yedekleme \u00e7\u00f6z\u00fcm\u00fcn\u00fcn bu a\u011f\u0131 ayn\u0131 zamanda veri aktar\u0131m\u0131 i\u00e7in kulland\u0131\u011f\u0131 g\u00f6zden ka\u00e7maktad\u0131r. E\u011fer ayr\u0131 bir yedekleme a\u011f\u0131 tan\u0131mlanmam\u0131\u015fsa, <strong>yedekleme s\u0131ras\u0131nda olu\u015fan trafik, y\u00f6netim trafi\u011fini etkiler<\/strong> ve bu durum hem yedekleme h\u0131z\u0131n\u0131 d\u00fc\u015f\u00fcr\u00fcr hem de sistem eri\u015fimini geciktirebilir. Yada yetersiz uplink h\u0131zlar\u0131 yedekleme senaryolar\u0131n\u0131 ba\u015far\u0131s\u0131z k\u0131labilir.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong>2. Sanal Makine A\u011f\u0131 (Front-End Guest Network)<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Bu a\u011f, sanal makinelerin (guest operating systems) d\u0131\u015f d\u00fcnya ile haberle\u015fti\u011fi temel ta\u015f\u0131y\u0131c\u0131d\u0131r. Genellikle y\u00fcksek bant geni\u015fli\u011fi (10GbE, 25GbE) \u00fczerinden yap\u0131land\u0131r\u0131l\u0131r. Uygulamalarda yap\u0131lan hata ise, bu portlar\u0131n yede\u011finin de benzer bant geni\u015fli\u011finde planlanmas\u0131d\u0131r. Hata ihtimali d\u00fc\u015f\u00fck olan bu portlar i\u00e7in ayn\u0131 bantta yedekleme yapmak, kaynak israf\u0131d\u0131r. Daha ekonomik ve i\u015flevsel \u00e7\u00f6z\u00fcm, \u00f6rne\u011fin bir 10GbE ba\u011flant\u0131y\u0131 1GbE ba\u011flant\u0131 ile yedeklemektir. Ar\u0131za durumunda ilgili host \u00fczerindeki y\u00fcksek trafi\u011fe sahip sanal makineler, canl\u0131 g\u00f6\u00e7 (live migration) ile ba\u015fka hostlara ta\u015f\u0131nabilir.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong>3. Cluster A\u011f\u0131 (Cluster Interconnect \/ Backend Network)<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">K\u00fcmeleme yap\u0131lar\u0131nda, d\u00fc\u011f\u00fcmler (nodes) aras\u0131nda ileti\u015fim kurmak i\u00e7in kullan\u0131l\u0131r. Bu a\u011f, yaln\u0131zca durum bilgisi payla\u015f\u0131m\u0131 (heartbeat), veri \u00e7o\u011faltma (replication), sanal makine ta\u015f\u0131ma (migration) gibi i\u015flevsel g\u00f6revler i\u00e7in yap\u0131land\u0131r\u0131l\u0131r. Bu a\u011fda ya\u015fanacak yo\u011funluk, do\u011frudan servis kalitesini etkiler. Bu nedenle, m\u00fcmk\u00fcnse bu trafik di\u011fer a\u011flardan (management ve front-end) kesinlikle ayr\u0131lmal\u0131d\u0131r.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong>4. Depolama A\u011f\u0131 (Storage Network)<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Depolama sistemine eri\u015fimin IP tabanl\u0131 protokollerle sa\u011fland\u0131\u011f\u0131 yap\u0131larda (\u00f6rne\u011fin iSCSI, NFS), bu a\u011f kritik \u00f6neme sahiptir. \u00d6zellikle yaz\u0131l\u0131m tan\u0131ml\u0131 depolama (SDS) veya IP SAN mimarilerinde <strong>y\u00fcksek bant geni\u015fli\u011fine<\/strong> ve <strong>d\u00fc\u015f\u00fck gecikmeye<\/strong> ihtiya\u00e7 vard\u0131r. Bu a\u011f, hem cluster trafi\u011finden hem de sanal makine trafi\u011finden ayr\u0131 olmal\u0131d\u0131r. Aksi takdirde random\/sequential I\/O ak\u0131\u015flar\u0131, kullan\u0131c\u0131 trafi\u011fini negatif etkileyebilir.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong><u>\u00d6rnek Fiziksel A\u011f Tasar\u0131m\u0131: 2 Node\u2019lu K\u00fcme<\/u><\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">A\u015fa\u011f\u0131da belirtilen yap\u0131, iki node\u2019lu bir sanalla\u015ft\u0131rma k\u00fcmesinde minimum gereksinimleri kar\u015f\u0131layacak \u015fekilde planlanm\u0131\u015ft\u0131r:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><td><strong>A\u011f Tipi<\/strong><\/td><td><strong>Ana Ba\u011flant\u0131<\/strong><\/td><td><strong>Yedekleme \/ Failover<\/strong><\/td><\/tr><\/thead><tbody><tr><td>Management Network<\/td><td>1 \u00d7 10GbE<\/td><td>1 \u00d7 1GbE<\/td><\/tr><tr><td>Front-End Guest Network<\/td><td>1 \u00d7 10GbE<\/td><td>1 \u00d7 1GbE<\/td><\/tr><tr><td>Cluster (Backend) Network<\/td><td>2 \u00d7 10GbE (Active\/Passive)<\/td><td>Yok (Ayr\u0131 yol tan\u0131m\u0131 \u00f6nerilir)<\/td><\/tr><tr><td>Storage Network<\/td><td>2 \u00d7 10GbE (Active\/Passive)<\/td><td>Yok (Yedeklilik yap\u0131labilir)<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"wp-block-paragraph\">Bu mimari sayesinde;<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Yedekleme ve y\u00f6netim trafi\u011fi birbirini etkilemez<\/li>\n\n\n\n<li>Sanal makineler ile kullan\u0131c\u0131lar aras\u0131ndaki a\u011f, cluster hareketlerinden izole edilir<\/li>\n\n\n\n<li>SDS \/ IP SAN trafi\u011fi di\u011fer a\u011flara y\u00fck bindirmez<\/li>\n\n\n\n<li>Ar\u0131zalar esnas\u0131nda d\u00fc\u015f\u00fck h\u0131zl\u0131 ba\u011flant\u0131larla ge\u00e7ici \u00e7al\u0131\u015fma sa\u011flanabilir<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1536\" height=\"1024\" src=\"https:\/\/www.ertanerbek.com\/wp-content\/uploads\/2025\/07\/Dogru-Bir-Sanallastirma-Sistemi-Nasil-Tasarlanmalidir-1.png\" alt=\"\" class=\"wp-image-69\" srcset=\"https:\/\/www.ertanerbek.com\/wp-content\/uploads\/2025\/07\/Dogru-Bir-Sanallastirma-Sistemi-Nasil-Tasarlanmalidir-1.png 1536w, https:\/\/www.ertanerbek.com\/wp-content\/uploads\/2025\/07\/Dogru-Bir-Sanallastirma-Sistemi-Nasil-Tasarlanmalidir-1-300x200.png 300w, https:\/\/www.ertanerbek.com\/wp-content\/uploads\/2025\/07\/Dogru-Bir-Sanallastirma-Sistemi-Nasil-Tasarlanmalidir-1-1024x683.png 1024w, https:\/\/www.ertanerbek.com\/wp-content\/uploads\/2025\/07\/Dogru-Bir-Sanallastirma-Sistemi-Nasil-Tasarlanmalidir-1-768x512.png 768w\" sizes=\"auto, (max-width: 1536px) 100vw, 1536px\" \/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sanalla\u015ft\u0131rma sistemlerinde en s\u0131k kar\u015f\u0131la\u015f\u0131lan altyap\u0131 sorunlar\u0131ndan biri, depolama birimlerinin g\u00f6revlerinin yanl\u0131\u015f tan\u0131mlanmas\u0131ndan kaynaklanmaktad\u0131r. \u00d6zellikle Storage Area Network (SAN) cihazlar\u0131, sahip olduklar\u0131 donan\u0131m \u00f6zellikleri nedeniyle \u00e7o\u011fu zaman yanl\u0131\u015f kullan\u0131m senaryolar\u0131na maruz b\u0131rak\u0131lmakta; bu da performans darbo\u011fazlar\u0131na ve yat\u0131r\u0131m\u0131n erken a\u015famada de\u011fer kaybetmesine yol a\u00e7maktad\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Geleneksel olarak Fibre Channel (FC) protokol\u00fc ile ba\u011flanan SAN sistemleri, rastgele eri\u015fim (random I\/O) senaryolar\u0131 i\u00e7in optimize edilmi\u015ftir. Ancak bu yap\u0131lar, ard\u0131\u015f\u0131k veri yazma ve okuma i\u015flemlerinde (sequential I\/O) yeterli verim sa\u011flayamazlar. Y\u00fcksek maliyetli FC-SAN sistemlerinin dosya sunucular\u0131 (file server) veya yedekleme platformlar\u0131 gibi ard\u0131\u015f\u0131k veri i\u015fleme yo\u011fun servislerde kullan\u0131lmas\u0131, megabayt ba\u015f\u0131na performans oran\u0131n\u0131 ciddi anlamda d\u00fc\u015f\u00fcrmekte ve sistemlerin planlanandan \u00e7ok daha k\u0131sa s\u00fcrede yenilenmesine neden olmaktad\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Yayg\u0131n kan\u0131n\u0131n aksine, IP SAN ile FC SAN aras\u0131nda veri aktar\u0131m kapasitesi a\u00e7\u0131s\u0131ndan belirleyici fark, aktar\u0131m\u0131n bant geni\u015fli\u011finde de\u011fil, veri eri\u015fim s\u00fcresinde (latency) ortaya \u00e7\u0131kar. Bu nedenle milisaniye seviyesinde hassasiyet gerektirmeyen veri ak\u0131\u015flar\u0131nda (\u00f6rne\u011fin yedekleme s\u00fcre\u00e7leri, dosya payla\u015f\u0131m\u0131, medya d\u00fczenleme gibi senaryolar) IP SAN veya Network Attached Storage (NAS) cihazlar\u0131n\u0131n tercih edilmesi hem performans hem de maliyet a\u00e7\u0131s\u0131ndan daha uygun \u00e7\u00f6z\u00fcmler sunmaktad\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Nitekim IP tabanl\u0131 NAS sistemleri, ard\u0131\u015f\u0131k veri i\u015flemlerinde \u00f6zelle\u015ftirilmi\u015f \u00f6n bellekleme ve veri s\u0131ralama teknikleriyle FC-SAN \u00e7\u00f6z\u00fcmlerine k\u0131yasla daha y\u00fcksek aktar\u0131m performanslar\u0131 sergileyebilmektedir. \u00d6rne\u011fin video d\u00fczenleme (video editing) ya da m\u00fchendislik \u00e7izim yaz\u0131l\u0131mlar\u0131 (CAD) gibi b\u00fcy\u00fck dosyalar\u0131n ge\u00e7ici bellekte i\u015flenmesine dayal\u0131 uygulamalarda, IP SAN\/NAS \u00e7\u00f6z\u00fcmleri daha verimli sonu\u00e7lar do\u011furabilmektedir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Ayr\u0131ca raporlama gibi zamanlanm\u0131\u015f veri \u00e7\u0131kt\u0131lar\u0131 \u00fcreten sistemler (\u00f6rne\u011fin otomatik g\u00fcn sonu raporlar\u0131) i\u00e7in kullan\u0131lan veri taban\u0131 b\u00f6l\u00fcmleri de IP tabanl\u0131 sistemlere ta\u015f\u0131nabilir. Bu yakla\u015f\u0131m, y\u00fcksek fiyatl\u0131 FC-SAN cihazlar\u0131n\u0131n yaln\u0131zca milisaniye duyarl\u0131l\u0131\u011f\u0131 gerektiren i\u015flemler i\u00e7in ayr\u0131lmas\u0131n\u0131 sa\u011flayarak hem donan\u0131m \u00f6mr\u00fcn\u00fc uzat\u0131r hem de toplam sahip olma maliyetini azalt\u0131r.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Bu ba\u011flamda depolama mimarilerinin kullan\u0131m senaryolar\u0131na uygun olarak ayr\u0131\u015ft\u0131r\u0131lmas\u0131, sadece performans kazanc\u0131 de\u011fil, ayn\u0131 zamanda sistem s\u00fcrd\u00fcr\u00fclebilirli\u011fi a\u00e7\u0131s\u0131ndan da kritik \u00f6nem ta\u015f\u0131r. Etkin sistem m\u00fchendisli\u011fi, donan\u0131m al\u0131m g\u00fcc\u00fcn\u00fc de\u011fil, eldeki kaynaklarla en verimli, uzun \u00f6m\u00fcrl\u00fc ve i\u015fletme maliyeti d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcm\u00fcn \u00fcretilmesini esas almal\u0131d\u0131r. Sistem m\u00fchendisinin temel g\u00f6revi, sistemi en uygun maliyetli ve s\u00fcrd\u00fcr\u00fclebilir \u015fekilde yap\u0131land\u0131rmakt\u0131r.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Sanalla\u015ft\u0131rma sistemleri, fiziksel donan\u0131m kaynaklar\u0131n\u0131n daha verimli kullan\u0131lmas\u0131n\u0131 sa\u011flamak amac\u0131yla geli\u015ftirilmi\u015f altyap\u0131lard\u0131r. Temel hedefleri, kaynak&#8230;<\/p>\n","protected":false},"author":1,"featured_media":109,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1,9,27],"tags":[],"class_list":["post-70","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-genel","category-virtualization","category-sanallastirma-genel"],"_links":{"self":[{"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=\/wp\/v2\/posts\/70","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=70"}],"version-history":[{"count":2,"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=\/wp\/v2\/posts\/70\/revisions"}],"predecessor-version":[{"id":72,"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=\/wp\/v2\/posts\/70\/revisions\/72"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=\/wp\/v2\/media\/109"}],"wp:attachment":[{"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=70"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=70"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ertanerbek.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=70"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}