This slide deck, in Italian, was presented at VMUGIT (5-Apr-2018) in Lecce (italy). It describes the need for remote compute resources next to where data is created and the relationship between edge and cloud, showing that cutting edge technology now present in the datacenter is also perfect for edge computing. The presentation is based on the content of this blog: https://blog.openio.io/data-management-in-industrial-iot-and-edge-computing
2. OpenIO
Un paio di info su di me
Head of Product Strategy - OpenIO
Io sono quello a sinistra Non ancora un campione
Per il resto c’è Google
2.5 anni di “duro” lavoro
6. Ma quanti dati produciamo in realtà?
Phone 4K video
350MB/minute
1080p Surveillance camera
2.1TB/day
Connected car
4TB/day
Small DC Logs
100GB/week
Human genome
715MB/each
Airbus A350
2.5TB/flight
My PC Backup
0.5TB
E questo è solo l’inizio!
11. Un esempio: l’auto autonoma
1. Dati istantanei
2. Dati a breve termine
3. Dati a lungo termine
12. I dati istantanei
• Servono per prendere decisioni
istantanee
• Latenza? Connessione? No, Grazie!
• Dopo che sono stati usati, molti di
questi si posso eliminare
• Usano molta potenza di calcolo e
relativamente poco storage
13. I dati a breve termine
• Durano per la lunghezza del viaggio o
poche settimane
• Diagnostica, DashCam.
• Usano poca CPU, ma relativamente più
spazio storage
• Vengono scaricati alla fine del viaggio
14. I dati a lungo termine
• Durano per “sempre”
• Storico, Analytics, R&D
• Sono tanti (soprattutto perchè si
accumulano!)
• Non usano CPU locale, scaricati sul
cloud appena possibile
15. Il Data Center che non si vede!
• 320TOPS
• CPU+GPU
• E molto altro
16. Perchè il Cloud da solo non basta
• Connettività e continuità del servizio
• Efficienza
• Costi
22. Ma quali sono i vantaggi?
• Astrazione HW-SW
• Facile da programmare
• Facile da gestire
• Resiliente
• Scalabile
23. OpenIO
Object Storage + Serverless Computing in practica
Everything happens into SDS (no external resources)
Add Category metadata tag to original object
Metadata enrichment
Queue manager
Category tag = Volcano
Processing
1- file is added, 2- event is intercepted and the function is triggered, 3- data is processed into the storage system 4- the result
(recognised content) is added to the original metadata, increasing its value and making it searchable!
OpenIO SDS
1
2
3
4
26. OpenIO
Builds cost effective storage
infrastructures for all kinds of
capacity-driven workloads, while
keeping them sustainable at any
scale.
OpenIO
SDS Grid for Apps
Processing triggered on data by events:
Grid for Apps
Hides complexity of infrastructure
management from operators and
developers, while helping to bring
data closer to the applications and
make them more scalable.
27. OpenIO
OpenIO SDS funziona anche su un computer da $5
OpenIO
Lightweight design and efficiency: 400MB of RAM and 1 ARM CPU core is all you need to run OpenIO SDS
28. OpenIO
Open source and proprietary code
OpenIO SDS core
(AGPLv3)
S3 and Swift
Gateway
File Gateway WebUI
Large scale
deployment and
reporting
Tools
Open source Proprietary
Grid For Apps
SDS Event Queue
(AGPLv3)