This article is part of our "Tiger Bridge - Under the hood" series, where we uncover technical details about our product, dive deep into the solutions it offers, and explain which decisions we had to make while designing it.
Dilemma
While considering the architecture of the product, we had to find out how we could integrate our software between the existing infrastructure of our clients and the cloud.
Background
Many people face the problem of needing to store more data while having to meet constantly growing faster data access requirements.
The speed at which data is accessed is dictated by the latency and bandwidth, as shown here:
|
Latency\Bandwidth |
Low |
High |
|
High |
Cloud, slow networks |
Networks (compared to local storage) |
|
Low |
Spinning disks and fast networks (1+ Gb, Infiniband, iSCSI) |
SSDs (best and most expensive option) |
This diagram lies at the core of tiering solutions. Price naturally goes left-to-right, top-to-bottom.
The speed-to-volume diagram looks similar:
|
Speed\Volume |
Low |
High |
|
High |
Cloud, slow networks |
Cloud, NAS |
|
Low |
Local disks, SSDs |
Disk arrays (best and most expensive option) |
If you want to have the best, it will be expensive. In bigger volumes, it also becomes more problematic.
For a given workflow, the need for speed is relatively constant and it usually generates some buffer. The need for bigger volumes, on the other hand, grows without stop.
People using disk arrays pay a lot and often encounter issues, which makes them unhappy with the solution. Companies in the upper-right or bottom-left corner tend to be satisfied.
The tiering concept offers to replace the expensive and problematic disk arrays with the other two options (upper-right and bottom-left). As a result, it removes complexity and reduces cost, providing an alternative while keeping the same volume and speed.
This hybrid approach helps even if you have been using one of the corner solutions so far. For the ones previously in the bottom-left, the hybrid solution will add volume without sacrificing speed. For solutions in the upper-right, the reverse will take place (speed added, volume kept).
This is what a typical on-premises setup looks like – we have a couple of clients accessing a shared server with some storage:
We considered the following options during the process of making a hybrid solution:
- With a cloud gateway – a machine which can replace the local server.
- With a file or application server change. This approach can also offer:
- Making an internal solution.
- Making an external component which can talk to the cloud.
- Making an internal solution.
Pros/Cons Analysis
Cloud gateway could be the easiest option because it replaces the existing file server with a new machine that acts as a gateway to the cloud. However, this means recycling an old file server which may not be “old” at all. And if it is an application server (e.g. SAP), this operation might not be possible.
Making an internal solution could seem more complex from a developmental perspective. On the plus side, it is much easier to implement into existing workflows and infrastructure.
An external component attached to the server will work flawlessly for some applications. However, for others that require local drives due to performance or functionality requirements, it might be a show-stopper. Making external solutions may be easier and could give us full control, but it is disruptive and sometimes impossible to implement.
Decision
We chose option 2a – an internal solution to be integrated in the existing server of the client.
Arguments
With an on-premises first hybrid model, our goal is to preserve the workflow and infrastructure. Replacing the local storage with the network might be undoable for the application. Users working with local storage usually do so because of better performance. We don’t want them to sacrifice that.
We believe that introducing new SMB storage points (NAS devices) instead of local storage is not a good idea as it is disruptive.
With this approach, we enable our software to be used as a gateway, too. However, we put efforts into providing an alternative as well, which works really well for hybrid solutions.
This way, clients do not need to adjust their cloud transformation strategies to IT infrastructure refresh cycles. Using the internal solution approach with Tiger Bridge, they have the option to immediately take advantage of benefits like AI services without replacing their existing infrastructure and disruptively adjusting their workflows.