Couple of questions

I watched the Whiteboard video and have a few questions:

1.  Lets say I have the data spread across 8 nodes.  How many nodes can I lose before I lose data?  Is there some sort of calculation that can be used to determine this based on the number of nodes used?

2.  Lets say I have nodes spread across 2-3 locations.  Is there a way to configure the software to force it that if I lose a site I don't lose any data?

3. Which applications is this best suited for?  I imagine that any write-intensive applications would suffer (or is this more of an asynchronous chunkification/dispersal of the data?).

Previously Jim wrote:

I watched the Whiteboard video and have a few questions:

1.  Lets say I have the data spread across 8 nodes.  How many nodes can I lose before I lose data?  Is there some sort of calculation that can be used to determine this based on the number of nodes used?

2.  Lets say I have nodes spread across 2-3 locations.  Is there a way to configure the software to force it that if I lose a site I don't lose any data?

3. Which applications is this best suited for?  I imagine that any write-intensive applications would suffer (or is this more of an asynchronous chunkification/dispersal of the data?).

Hi Jim, thanks for your questions.

To answer your first question, one has very wide flexibility when setting up a dispersed storage network.  You could have 8 nodes and configure it such that you need only 5 servers available to restore data.  This setup would have an overhead of 8/5 (about 60% increase in data size).  You could also create an extremely reliable configuration, requiring any 2 nodes to be available, however this would have a data expansion of 8/2 (400%).  Grids are typically configured to have around a 30% data expansion, which corresponds to 3/4ths of the grid nodes need to be available.

To answer the second question, if you have a grid with N number of nodes, the threshold (number of nodes required to recover the data) can be any value between 1 and (N-1), 1 corresponds to replication, while N is similar to a raid array with a single parity bit.  IDA's offer flexibility because the threshold can be anything you want.

Since dispersal operates over a network, the bandwidth limitations are what affect the types of uses it can be used for.  Dispersed grids can also be setup with all the nodes at a single data center / LAN and then the bandwidth limitations wouldn't be an issue but you lose the benefit of geographical dispersion.  I am not quite sure what you mean by the asynchronous chunkification, but essentially writing involves using the IDA to create slices, which on a modern machine can surpass hundreds of Mbps.

Jason

Hi,

can I get step-by-step installation guide and also I need configure steps  to configure multiple volts on multiple accessers?

Thanks

Hi,

You can download the install and usage guide by clicking on the following link:

https://www.cleversafe.org/documentation/User-guide-open-source.pdf

Please let us know if you need any clarification. 

You can create many Vaults on a single Accesser. The procedure is the same as your first Vault creation as detailed in the guide.

Thanks,

Sarah.

Hi,

Lets say I have created volt  “test”, can have same volt on another accesser?

If  your answer is Yes what is the procedure?

Thanks,

Hi,

Lets say I have created volt  “test”, can have same volt on another accesser?

If  your answer is Yes what is the procedure?

Thanks,

Previously ushakiranj wrote:

Lets say I have created volt  “test”, can have same volt on another accesser?

If  your answer is Yes what is the procedure?

It is possible to access the same vault on a different accesser. Unfortunately in the current release this require that you copy several files over from the first access to the second (which I describe below). However, you should note that it is unsafe to access the same block device vault from two accessers at the same time.

In your installation, there is a directory called "output". Just copy that from one accesser installation to another.