tfgrid/info_tfgrid
33
3
Fork 0
Code Issues Pull Requests 2 Packages Projects Releases Activity

s

Browse Source
This commit is contained in:
kristof de spiegeleer
2024-05-23 11:09:37 +03:00
parent 2d41a9a8a2
commit 00f1b87efa
115 changed files with 333 additions and 715 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
collections/tfgrid4/boxes/block_intro.md
View File

@@ -1,7 +0,0 @@
A Cloud, Storage or AI Block is a part of a server/computer (TF or INCA Node) which delivers a service which has well defined properties in relation to capacity, pricing, serviceabity, capabilities.
These Cloud, Storage or AI Blocks can be bought by the INCA/TF Community through the INCA marketplace.

14
collections/tfgrid4/boxes/aiblock.md → collections/tfgrid4/slices/aislice.md
View File

@@ -1,6 +1,6 @@
# AI Block (AB)
# AI Slice
!!wiki.include page:'block_intro.md'
!!wiki.include page:'slice_intro.md'
An AI box is a unit of AI capacity (GPU or future TPU driven).
@@ -10,13 +10,13 @@ An AI box can be launched in our Zero-OS and can enable any possible AI workload
## AI Hour (AH)
AI hour (AH) is like a kwatth unit for electricty: it represents a AI Block being used for 1h and billed as such.
AI hour (AH) is like a kwatth unit for electricty: it represents a AISlice being used for 1h and billed as such.
INCA Hosts (our cloud providers) can price the AI Hour themselves in a chosen currency.
## AI Block Properties
## AISlice Properties
The service provider (hoster) defines the following properties per AI Block:
The service provider (hoster) defines the following properties per AISlice:
- Type of GPU
- Price per hour
@@ -24,10 +24,10 @@ The service provider (hoster) defines the following properties per AI Block:
- Max bandwidth
- Min bandwidth (min 1 mbit/sec)
- Cost per GB bandwidth if any
- Max aggregation size (how many of the AI Blocks can be combined in one node if relevant)
- Max aggregation size (how many of the AISlices can be combined in one node if relevant)
- Link to support site if any (find info about hoster and service capabilities)
- If linked to Hosting Pool (company giving support on the machines)
- Location & type of location
Each AI Block has unique ID and can be looked at through a portal (or found).
Each AISlice has unique ID and can be looked at through a portal (or found).

32
collections/tfgrid4/boxes/cloudblock.md → collections/tfgrid4/slices/cloudslice.md
View File

@@ -1,17 +1,17 @@
<h1> CloudBlock Concepts </h1>
# Cloud Slice Concept
!!wiki.include page:'block_intro.md'
!!wiki.include page:'slice_intro.md'
## CloudBlock (CB)
## Cloud Slice
A cloudblock is a unit of compute, fast storage and memory.
There are unlimited different configurations of cloudblock.
A cloudslice is a unit of compute, fast storage and memory.
There are unlimited different configurations of cloudslice.
A configuration of a machine defines the cloudblock which can be made.
A configuration of a machine defines the cloudslice which can be made.
A cloudblock can be aggregated to make a bigger cloudblock.
A cloudslice can be aggregated to make a bigger cloudslice.
The default cloudblock has
The default cloudslice has
- 4GB memory
- 2 virtual cores of CPU
@@ -21,29 +21,29 @@ The default cloudblock has
Terms
- The virtual cores are max oversubscribed 4x
- Min passmark per cloudblock is 500 passmark (CPU benchmark)
- Min passmark per cloudslice is 500 passmark (CPU benchmark)
- At least 50 GB storage available, more is better
**Example a node with 64 GB or mem and 2 TB of SSD and 24 virtual cores.**
- 15 Storage Blocks each:
- 15 StorageSlices each:
- 4 GB of memory (60 GB total)
- 120 GB of SSD capacity
- 6.4 logical CPU core (oversubscription of 4, which means user can max use 4x CPU capacity if system allows)
- when a user choses the full machine, then he/she will have reserved all Storage Blocks capacity which means the machine is now dedicated reserved for the user, the hoster specifies the discount for this typically 50%. On a dedicated machine the user has full access to the GPU.
- when a user choses the full machine, then he/she will have reserved all StorageSlices capacity which means the machine is now dedicated reserved for the user, the hoster specifies the discount for this typically 50%. On a dedicated machine the user has full access to the GPU.
- Min 2GB always needs to be left as buffer for memory and 10% of SSD capacity
How does it work:
- When a developer wants to deploy a virtual machine, they need to define the required capacity and will be able to make a choice based on these CloudBlocks, each CloudBlock has different base specs (mem, ssd, gpu, cpu) and reputation, the developer can now make a selection of how many of these Storage Blocks need to be given to the virtual machine. This defines the monthly price which will have to be paid as well as capacity available to the VM.
- When a developer wants to deploy a virtual machine, they need to define the required capacity and will be able to make a choice based on these CloudSlices, each CloudSlice has different base specs (mem, ssd, gpu, cpu) and reputation, the developer can now make a selection of how many of these StorageSlices need to be given to the virtual machine. This defines the monthly price which will have to be paid as well as capacity available to the VM.
## Cloud Hour (CH)
A cloudhour is like a kwatth unit for electricty: it represents a CloudBlock being used for 1h and billed as such.
A cloudhour is like a kwatth unit for electricty: it represents a CloudSlice being used for 1h and billed as such.
INCA Hosts (our cloud providers) can price the CloudHour themselves in a chosen currency.
## CloudBlock Properties
## CloudSlice Properties
The service provider (hoster) defines the following properties per cloud box
@@ -57,12 +57,12 @@ The service provider (hoster) defines the following properties per cloud box
- Cost per GB bandwidth
- Cost for additional storage HDD or SSD per GB/month
- Max additional storage
- Max aggregation size (how many of the cloudblock can be combined)
- Max aggregation size (how many of the cloudslice can be combined)
- Link to support site if any (find info about hoster and service capabilities)
- If linked to Hosting Pool (company giving support on the machines)
- Pub ip address possible or not (is option)
- Link to monitoring page (if any)
- Location & type of location
Each cloudblock has unique ID and can be looked at through a portal (or found).
Each cloudslice has unique ID and can be looked at through a portal (or found).

0
collections/tfgrid4/boxes/cloudblock_tech.md → collections/tfgrid4/slices/cloudslice_tech.md
View File

7
collections/tfgrid4/slices/slice_intro.md Normal file
View File

@@ -0,0 +1,7 @@
A Cloud, Storage or AISlice is a part of a server/computer (TF or INCA Node) which delivers a service which has well defined properties in relation to capacity, pricing, serviceabity, capabilities.
These Cloud, Storage or AISlices can be bought by the INCA/TF Community through the INCA marketplace.

22
collections/tfgrid4/boxes/storageblock.md → collections/tfgrid4/slices/storageslice.md
View File

@@ -1,34 +1,34 @@
# StorageBlock
# Storage Slice
!!wiki.include page:'block_intro.md'
!!wiki.include page:'slice_intro.md'
## StorageBlock (SB)
## Storage Slice
A storageblock is a unit of ZDB storage as can be used as backend for Zero-Stor (our quantum safe storage system).
A storageslice is a unit of ZDB storage as can be used as backend for Zero-Stor (our quantum safe storage system).
The default cloudblock has:
The default cloudslice has:
- 100GB of storage
## StorageHour (SH)
## Storage Hour (SH)
A storagehour is like a kwatth unit for electricty, it represents a storageblock being used for 1h and billed as such.
A storagehour is like a kwatth unit for electricty, it represents a storageslice being used for 1h and billed as such.
INCA Hosts (our cloud providers) can price the StorageHour themselves in a chosen currency.
## StorageBlock Properties
## Storage Slice Properties
The service provider (hoster) defines the following properties per StorageBlock:
The service provider (hoster) defines the following properties per StorageSlice:
- Min size in GB (100GB+), max size
- Discounts if used for longer periods & if renter has large amount of TFT in wallet
- Max bandwidth
- Min bandwidth (min 1 mbit/sec)
- Cost per GB bandwidth
- Max aggregation size (how many of the Storage Blocks can be combined)
- Max aggregation size (how many of the StorageSlices can be combined)
- Link to support site if any (find info about hoster and service capabilities)
- If linked to Hosting Pool (company giving support on the machines)
- Location & type of location
Each StorageBlock has unique ID and can be looked at through a portal (or found).
Each StorageSlice has unique ID and can be looked at through a portal (or found).