{"id":54686,"date":"2020-01-01T00:00:00","date_gmt":"2019-12-31T21:00:00","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/5-8-mln-iops-zachem-tak-mnogo"},"modified":"2020-02-18T14:02:44","modified_gmt":"2020-02-18T11:02:44","slug":"5-8-mln-iops-zachem-tak-mnogo","status":"publish","type":"post","link":"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/5-8-mln-iops-zachem-tak-mnogo","title":{"rendered":"5,8 milionu IOPS: pro\u010d tolik?","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Ahoj Habr! Sady dat pro Big Data a strojov\u00e9 u\u010den\u00ed exponenci\u00e1ln\u011b rostou a mus\u00edme je st\u00edhat zpracov\u00e1vat. N\u00e1\u0161 p\u0159\u00edsp\u011bvek je o dal\u0161\u00ed inovativn\u00ed technologii v oblasti vysoce v\u00fdkonn\u00fdch v\u00fdpo\u010dt\u016f (HPC, High Performance Computing), kter\u00e1 byla p\u0159edstavena na st\u00e1nku Kingston na Supercomputing-2019<\/a><\/noindex>. Toto je aplikace hi-end syst\u00e9m\u016f \u00falo\u017ei\u0161\u0165 dat (S\u00daD) na serverech s grafick\u00fdmi procesory (GPU) a technologi\u00ed sb\u011brnice GPUDirect Storage. D\u00edky p\u0159\u00edm\u00e9mu p\u0159enosu dat mezi S\u00daD a GPU, obch\u00e1zej\u00edc\u00edmu CPU, se v\u00fdrazn\u011b zrychluje na\u010d\u00edt\u00e1n\u00ed dat do GPU akceler\u00e1tor\u016f, co\u017e umo\u017e\u0148uje, aby se aplikace Big Data vykon\u00e1valy na maxim\u00e1ln\u00ed v\u00fdkon, kter\u00fd GPU poskytuj\u00ed. Z\u00e1rove\u0148 zaj\u00edmaj\u00ed v\u00fdvoj\u00e1\u0159e HPC syst\u00e9m\u016f dosa\u017een\u00e9 \u00fasp\u011bchy v oblasti S\u00daD s nejvy\u0161\u0161\u00edmi rychlostmi vstupu\/v\u00fdstupu \u2014 takov\u00e9, jak\u00e9 vyr\u00e1b\u00ed Kingston.<\/p>\n

\"5,8
\n<\/a><\/noindex><\/p>\n

V\u00fdkon GPU p\u0159ed\u010d\u00ed na\u010d\u00edt\u00e1n\u00ed dat <\/h2>\n

\nOd doby, kdy byla v roce 2007 vytvo\u0159ena CUDA \u2014 softwarov\u00e1 a hardwarov\u00e1 architektura paraleln\u00edch v\u00fdpo\u010dt\u016f zalo\u017een\u00e1 na GPU pro v\u00fdvoj aplikac\u00ed obecn\u00e9ho pou\u017eit\u00ed, se hardwarov\u00e9 mo\u017enosti samotn\u00fdch GPU mimo\u0159\u00e1dn\u011b zv\u00fd\u0161ily. Dnes se GPU st\u00e1le v\u00edce vyu\u017e\u00edvaj\u00ed v oblasti HPC aplikac\u00ed, jako jsou velk\u00e1 data (Big Data), strojov\u00e9 u\u010den\u00ed (ML, machine learning) a hlubok\u00e9 u\u010den\u00ed (DL, deep learning). <\/p>\n

Uv\u011bdomme si, \u017ee i p\u0159es podobnost term\u00edn\u016f, posledn\u00ed dva p\u0159edstavuj\u00ed algoritmicky odli\u0161n\u00e9 \u00fakoly. ML \u0161kol\u00ed po\u010d\u00edta\u010d na z\u00e1klad\u011b strukturovan\u00fdch dat, zat\u00edmco DL na z\u00e1klad\u011b odezvy od neuronov\u00e9 s\u00edt\u011b. P\u0159\u00edklad, kter\u00fd pom\u00e1h\u00e1 pochopit rozd\u00edly, je pom\u011brn\u011b jednoduch\u00fd. P\u0159edstavme si, \u017ee po\u010d\u00edta\u010d m\u00e1 rozli\u0161it fotografie ko\u010dek a ps\u016f, kter\u00e9 se na\u010d\u00edtaj\u00ed z \u00falo\u017ei\u0161t\u011b. Pro ML byste m\u011bl poskytnout sadu obr\u00e1zk\u016f s mnoha tagy, z nich\u017e ka\u017ed\u00fd definuje n\u011bjak\u00fd jeden rys zv\u00ed\u0159ete. U DL sta\u010d\u00ed nahr\u00e1t mnohem v\u011bt\u0161\u00ed po\u010det obr\u00e1zk\u016f, ale pouze s jedn\u00edm tagem \u201eto je ko\u010dka\u201c nebo \u201eto je pes\u201c. DL velmi p\u0159ipom\u00edn\u00e1, jak se u\u010d\u00ed mal\u00e9 d\u011bti \u2014 prost\u011b jim ukazuj\u00ed obr\u00e1zky ps\u016f a ko\u010dek v knih\u00e1ch a v \u017eivot\u011b (\u010dasto dokonce ani\u017e by detailn\u011b vysv\u011btlovali rozd\u00edly), a mozek d\u00edt\u011bte si posl\u00e9ze za\u010dne ur\u010dovat typ zv\u00ed\u0159ete po ur\u010dit\u00e9m kritick\u00e9m po\u010dtu obr\u00e1zk\u016f k porovn\u00e1n\u00ed (odhaduje se, \u017ee jde o zhruba stovku nebo dv\u011b uk\u00e1zek za cel\u00e9 ran\u00e9 d\u011btstv\u00ed). Algoritmy DL zat\u00edm nejsou tak dokonal\u00e9: aby neuronov\u00e1 s\u00ed\u0165 mohla \u00fasp\u011b\u0161n\u011b pracovat na identifikaci obraz\u016f, je nutn\u00e9 poskytnout a zpracovat v GPU miliony obr\u00e1zk\u016f.<\/p>\n

Z\u00e1v\u011br \u00favodu: na b\u00e1zi GPU lze budovat HPC aplikace v oblasti Big Data, ML a DL, ale existuje probl\u00e9m \u2014 datov\u00e9 sady jsou tak velk\u00e9, \u017ee \u010das str\u00e1ven\u00fd na\u010d\u00edt\u00e1n\u00edm dat ze syst\u00e9mu \u00falo\u017ei\u0161t\u011b do GPU za\u010d\u00edn\u00e1 sni\u017eovat celkov\u00fd v\u00fdkon aplikace. Jin\u00fdmi slovy, rychl\u00e9 grafick\u00e9 procesory z\u016fst\u00e1vaj\u00ed nedostate\u010dn\u011b vyu\u017eiteln\u00e9 kv\u016fli pomal\u00e9mu vstupu\/v\u00fdstupu dat ze strany jin\u00fdch subsyst\u00e9m\u016f. Rozd\u00edl v rychlosti vstupu\/v\u00fdstupu GPU a sb\u011brnice k CPU\/\u00falo\u017ei\u0161ti m\u016f\u017ee b\u00fdt v \u0159\u00e1du.<\/p>\n

Jak funguje technologie GPUDirect Storage?<\/h2>\n

\nProces vstupu a v\u00fdstupu je \u0159\u00edzen CPU, stejn\u011b jako proces na\u010d\u00edt\u00e1n\u00ed dat ze \u00falo\u017ei\u0161t\u011b do grafick\u00fdch procesor\u016f pro n\u00e1sledn\u00e9 zpracov\u00e1n\u00ed. Odtud se objevila pot\u0159eba technologie, kter\u00e1 by zajistila p\u0159\u00edm\u00fd p\u0159\u00edstup mezi GPU a NVMe disky pro rychlou interakci mezi sebou. Prvn\u00ed takovou technologii navrhla spole\u010dnost NVIDIA a nazvala ji GPUDirect Storage. V podstat\u011b se jedn\u00e1 o variaci d\u0159\u00edve vyvinut\u00e9 technologie GPUDirect RDMA (Remote Direct Memory Address).<\/p>\n

\"5,8
\nJensen Huang, gener\u00e1ln\u00ed \u0159editel spole\u010dnosti NVIDIA, p\u0159edstavuje GPUDirect Storage jako variantu GPUDirect RDMA na v\u00fdstav\u011b SC-19. Zdroj: NVIDIA<\/i><\/p>\n

Rozd\u00edl mezi GPUDirect RDMA a GPUDirect Storage spo\u010d\u00edv\u00e1 v za\u0159\u00edzen\u00edch, mezi kter\u00fdmi prob\u00edh\u00e1 adresace. Technologie GPUDirect RDMA je ur\u010dena pro p\u0159enos dat p\u0159\u00edmo mezi rozhran\u00edm s\u00ed\u0165ov\u00e9 karty (NIC) a pam\u011bt\u00ed GPU, zat\u00edmco GPUDirect Storage poskytuje p\u0159\u00edmou cestu pro p\u0159enos dat mezi m\u00edstn\u00edm nebo vzd\u00e1len\u00fdm \u00falo\u017ei\u0161t\u011bm, jako je NVMe nebo NVMe p\u0159es Fabric (NVMe-oF), a pam\u011bt\u00ed GPU.<\/p>\n

Ob\u011b mo\u017enosti, GPUDirect RDMA a GPUDirect Storage, se vyh\u00fdbaj\u00ed zbyte\u010dn\u00e9mu p\u0159enosu dat p\u0159es vyrovn\u00e1vac\u00ed pam\u011b\u0165 v pam\u011bti CPU a umo\u017e\u0148uj\u00ed mechanismu p\u0159\u00edm\u00e9ho p\u0159\u00edstupu k pam\u011bti (DMA) p\u0159en\u00e1\u0161et data z s\u00ed\u0165ov\u00e9 karty nebo \u00falo\u017ei\u0161t\u011b p\u0159\u00edmo do pam\u011bti GPU nebo z n\u00ed \u2013 a to v\u0161e bez z\u00e1t\u011b\u017ee na centr\u00e1ln\u00ed procesor. U GPUDirect Storage nen\u00ed d\u016fle\u017eit\u00e9 um\u00edst\u011bn\u00ed \u00falo\u017ei\u0161t\u011b: m\u016f\u017ee to b\u00fdt NVMe disk uvnit\u0159 jednotky s GPU, uvnit\u0159 racku nebo p\u0159ipojen\u00fd p\u0159es s\u00ed\u0165 jako NVMe-oF. <\/p>\n

\"5,8
\nSch\u00e9ma fungov\u00e1n\u00ed GPUDirect Storage. Zdroj: NVIDIA<\/i><\/p>\n

Hi-End \u00falo\u017ei\u0161t\u011b na NVMe jsou vyhled\u00e1van\u00e1 na trhu HPC aplikac\u00ed.<\/h2>\n

\nS v\u011bdom\u00edm, \u017ee s p\u0159\u00edchodem GPUDirect Storage se z\u00e1jem velk\u00fdch klient\u016f zam\u011b\u0159\u00ed na nab\u00eddku \u00falo\u017en\u00fdch syst\u00e9m\u016f s rychlost\u00ed vstupu\/v\u00fdstupu odpov\u00eddaj\u00edc\u00ed propustnosti GPU, na v\u00fdstav\u011b SC-19 Kingston p\u0159edvedl demonstra\u010dn\u00ed syst\u00e9m slo\u017een\u00fd z \u00falo\u017ei\u0161t\u011b na b\u00e1zi NVMe disk\u016f a jednotky s GPU, v n\u00ed\u017e prob\u00edhaly anal\u00fdzy tis\u00edc\u016f satelitn\u00edch sn\u00edmk\u016f za sekundu. O takov\u00e9m \u00falo\u017ei\u0161ti na b\u00e1zi 10 disk\u016f DC1000M U.2 NVMe jsme ji\u017e psali ve zpr\u00e1v\u011b z v\u00fdstavy superpo\u010d\u00edta\u010d\u016f.<\/a><\/noindex>. <\/p>\n

\"5,8
\n\u00dalo\u017ei\u0161t\u011b na b\u00e1zi 10 disk\u016f DC1000M U.2 NVMe skv\u011ble dopl\u0148uje server s grafick\u00fdmi akceler\u00e1tory. Zdroj: Kingston<\/i><\/p>\n

Takov\u00e9 \u00falo\u017ei\u0161t\u011b se realizuje ve form\u011b rackov\u00e9 jednotky 1U nebo v\u011bt\u0161\u00ed a m\u016f\u017ee b\u00fdt \u0161k\u00e1lov\u00e1no v z\u00e1vislosti na po\u010dtu disk\u016f DC1000M U.2 NVMe, kde ka\u017ed\u00fd m\u00e1 kapacitu 3,84-7,68 TB. DC1000M je prvn\u00ed model NVMe SSD ve form\u00e1tu U.2 v \u0159ad\u011b disk\u016f Kingston pro datov\u00e1 centra. M\u00e1 rating odolnosti (DWPD, Drive writes per day), kter\u00fd umo\u017e\u0148uje p\u0159episovat data na plnou kapacitu jednou denn\u011b po celou garantovanou dobu \u017eivotnosti disku. <\/p>\n

V testu fio v3.13 na opera\u010dn\u00edm syst\u00e9mu Ubuntu 18.04.3 LTS, Linux kernel 5.0.0-31-generic, uk\u00e1zal v\u00fdstavn\u00ed vzorek \u00falo\u017ei\u0161t\u011b rychlost \u010dten\u00ed (Sustained Read) 5,8 milionu IOPS p\u0159i trval\u00e9 propustnosti (Sustained Bandwidth) 23,8 Gbit\/s.<\/p>\n

Ari\u00ebl P\u00e9rez, business manager SSD at Kingston, described the new storage systems as follows: \"We are ready to equip the next generation of servers with U.2 NVMe SSD solutions to eliminate many of the data transfer bottlenecks traditionally associated with storage systems. The combination of NVMe SSD drives and our premium Server Premier DRAM makes Kingston one of the most comprehensive providers of data processing solutions in the industry.\"<\/p>\n

\"5,8
\nThe gfio v3.13 test showed a bandwidth of 23.8 Gbps for the demonstration storage system on DC1000M U.2 NVMe drives. Source: Kingston<\/i><\/p>\n

What will a typical system for HPC applications look like where GPUDirect Storage technology or similar ones are implemented? It is an architecture with physical separation of functional blocks within a rack: one or two units for memory, a few more for GPU and CPU compute nodes, and one or more units for storage systems.<\/p>\n

With the announcement of GPUDirect Storage and the potential emergence of similar technologies from other GPU vendors, demand for storage systems designed for high-performance computing is increasing for Kingston. A marker will be the speed of reading data from the storage system, comparable to the bandwidth of 40 or 100 Gbps network cards entering the compute unit with the GPU. Thus, ultra-fast storage systems, including external NVMe through Fabric, will transition from exotic to mainstream for HPC applications. In addition to science and financial calculations, they will find applications in many other practical areas, such as mega-city level security systems Safe City or transportation monitoring centers, where recognition and identification speeds at the level of millions of HD images per second are required,\u201d identified the market niche for top-tier storage systems.<\/p>\n

More information about Kingston products can be found at ofici\u00e1ln\u00edm webu<\/a><\/noindex> spole\u010dnost.
\n
Zdroj: habr.com<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

\u041f\u0440\u0438\u0432\u0435\u0442 \u0425\u0430\u0431\u0440! \u041d\u0430\u0431\u043e\u0440\u044b \u0434\u0430\u043d\u043d\u044b\u0445 \u0434\u043b\u044f Big Data \u0438 \u043c\u0430\u0448\u0438\u043d\u043d\u043e\u0433\u043e \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u044f \u044d\u043a\u0441\u043f\u043e\u043d\u0435\u043d\u0446\u0438\u0430\u043b\u044c\u043d\u043e \u0440\u0430\u0441\u0442\u0443\u0442 \u0438 \u043d\u0430\u0434\u043e \u0443\u0441\u043f\u0435\u0432\u0430\u0442\u044c \u0438\u0445 \u043e\u0431\u0440\u0430\u0431\u0430\u0442\u044b\u0432\u0430\u0442\u044c. \u041d\u0430\u0448 \u043f\u043e\u0441\u0442 \u043e \u0435\u0449\u0435 \u043e\u0434\u043d\u043e\u0439 \u0438\u043d\u043d\u043e\u0432\u0430\u0446\u0438\u043e\u043d\u043d\u043e\u0439 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0438 \u0432 \u043e\u0431\u043b\u0430\u0441\u0442\u0438 \u0432\u044b\u0441\u043e\u043a\u043e\u043f\u0440\u043e\u0438\u0437\u0432\u043e\u0434\u0438\u0442\u0435\u043b\u044c\u043d\u044b\u0445 \u0432\u044b\u0447\u0438\u0441\u043b\u0435\u043d\u0438\u0439 (HPC, High Performance Computing), \u043f\u043e\u043a\u0430\u0437\u0430\u043d\u043d\u043e\u0439 \u043d\u0430 \u0441\u0442\u0435\u043d\u0434\u0435 Kingston \u043d\u0430 Supercomputing-2019. \u042d\u0442\u043e \u043f\u0440\u0438\u043c\u0435\u043d\u0435\u043d\u0438\u0435 Hi-End \u0441\u0438\u0441\u0442\u0435\u043c \u0445\u0440\u0430\u043d\u0435\u043d\u0438\u044f \u0434\u0430\u043d\u043d\u044b\u0445 (\u0421\u0425\u0414) \u0432 \u0441\u0435\u0440\u0432\u0435\u0440\u0430\u0445 \u0441 \u0433\u0440\u0430\u0444\u0438\u0447\u0435\u0441\u043a\u0438\u043c\u0438 \u043f\u0440\u043e\u0446\u0435\u0441\u0441\u043e\u0440\u0430\u043c\u0438 (GPU) \u0438 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0435\u0439 \u0448\u0438\u043d\u044b GPUDirect […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[688],"tags":[],"class_list":["post-54686","post","type-post","status-publish","format-standard","hentry","category-administrirovanie"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd475.8 million IOPS: why so much? | ProHoster","description":"Hello Habr! Datasets for Big Data and machine learning are growing exponentially, and we need to keep up with processing them.","canonical_url":"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/5-8-mln-iops-zachem-tak-mnogo","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"cs_CZ","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd475.8 \u043c\u043b\u043d IOPS: \u0437\u0430\u0447\u0435\u043c \u0442\u0430\u043a \u043c\u043d\u043e\u0433\u043e? | ProHoster","og:description":"\u041f\u0440\u0438\u0432\u0435\u0442 \u0425\u0430\u0431\u0440! \u041d\u0430\u0431\u043e\u0440\u044b \u0434\u0430\u043d\u043d\u044b\u0445 \u0434\u043b\u044f Big Data \u0438 \u043c\u0430\u0448\u0438\u043d\u043d\u043e\u0433\u043e \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u044f \u044d\u043a\u0441\u043f\u043e\u043d\u0435\u043d\u0446\u0438\u0430\u043b\u044c\u043d\u043e \u0440\u0430\u0441\u0442\u0443\u0442 \u0438 \u043d\u0430\u0434\u043e \u0443\u0441\u043f\u0435\u0432\u0430\u0442\u044c \u0438\u0445 \u043e\u0431\u0440\u0430\u0431\u0430\u0442\u044b\u0432\u0430\u0442\u044c.","og:url":"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/5-8-mln-iops-zachem-tak-mnogo","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2019-12-31T21:00:00+00:00","article:modified_time":"2020-02-18T11:02:44+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"54686","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":"2026-01-24 12:22:25","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-02-28 20:00:11","updated":"2026-01-24 12:22:25","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/54686","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/comments?post=54686"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/54686\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/media?parent=54686"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/categories?post=54686"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/tags?post=54686"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}