GPU Network Management Solution (G-NMS)


 

http://taejin.wiz-wig.com/smartEditorCustom/images/bullet/arrow/bullet03.png 시스템 개요




네트워크 관리 시스템 개요

인공지능(AI), 사물인터넷(IoT), 로봇기술, 드론, 자율주행차, 가상현실(VR) 등이 주도하는 4차 산업혁명으로 변화하는 새로운 패러다임에 따라 데이터 관리역량은 물론 관련 연산을 병목현상 없이 실시간으로 작동시킬 수 있는 빠른 연산 및 전송 속도를 갖춘 고확장성 서버 인프라 시스템을 제공드립니다.
이에 태진티엔에스는 고객 비즈니스 환경과 연구, 개발 등 목적에 최적화 된 혁신적인 인프라와 네트워크 관리 솔루션을 제공합니다.



구조도


 


 

 

 

 

 

 

http://taejin.wiz-wig.com/smartEditorCustom/images/bullet/arrow/bullet03.png 핵심기능


Private image library installation and deployment

· Import docker private repository image

· Linux Docker tag or push Commands can be simply executedon G-NMS

 

 

Deployment of deep learning environment

· Multiple deep learning frameworks: Caffe/TensorFlow/CNTK etc.
· Support various models: GoogleNet/VGG/ResNet etc.
·On e-Key deployment of the distributed computing environment
· Application arrangement, rapid start of APPs


Management of deep learning training tasks


· Support for application templates, rapid submission of training tasks
· workflow, data pre-processing, model training, visualization
· Management of training tasks, observation of training progress and precision, parameter-tuning 




Resource Monitoring

· GPU resource total / usage. 

· GPU core average utilization and cache utilization



Computing resources management and scheduling 


· Monitoring of CPU/GPU running state and performance state
· Scheduling of GPU resourceson demand, resource isolation
· Scheduling strategy: fair share, preemption, backfilling



Job Schedules 

· Option Deep Learning Settings
· Personal Options Deep Learning setting
· Support TensorFlow, caffe, Mxnet, etc
· Deep learning can be edited in the middle
· Deep learning plan script editing can be supported by coding