Traditional SFTP on Cloud
A square peg in a round hole
In today’s world of technology, cloud computing is everywhere. Vendors such as Google and Microsoft are releasing products that have features & capabilities to enable businesses to become more lean, efficient & flexible. With these benefits on offer, almost every business out there is aiming to move to the cloud, to be more competitive.
However, migration from traditional tech stacks to the cloud comes with its own baggage, which we need to let go of. This white paper scrutinizes one such traditional technology — the File Transfer Protocol (FTP) and the Secured File Transfer Protocol (SFTP).
FTP & SFTP
The FTP, with its roots from the early ’70s, is a standard network protocol used to transfer files from client machines to the server and vice-versa. This was later secured by SSL thereby creating SFTP. For the remainder of this paper, reference shall be made only to SFTP as FTP fails to be an option considering security aspects.
Architecturally, SFTP requires that a server be always available, storage space be pre-allocated & a stateful connection be maintained during a session. It also has limited capabilities for DR situations and back-ups. When it comes to cloud computing, these fundamentals themselves are a misfit.
THE CLOUD
Resilience
It is known that systems are bound to fail due to technical glitches, physical interruptions or natural calamities. Cloud computing provides us various levels of fault tolerance — regions, availability zones, availability sets etc depending on the provider.
Having a traditional SFTP on the cloud machine forces one to be on a single virtual machine only, thus, extricating the possibilities for disaster recovery. With data being valued even more than money, a data loss could be catastrophic for a business.
Serverless Computing
Cloud computing provides us with the facility of ‘pay per execution’. This refers to a payment model, where a customer is charged only when a piece of code (or function) executes. This is done, by provisioning the necessary computes and resources at run time, post which the resources are returned back to the pool.
An analysis of the resource usage graph of a traditional SFTP server will show that resources are used sporadically (during file transfers), lying idle for the rest of the duration, thus theoretically fitting the serverless concept, but not practically. This is because, an SFTP server (as mentioned earlier), by design, needs an always-up state. Thus, raising the operational costs.
Elasticity
Elasticity is a key feature when it comes to handling increased loads on a cloud environment especially for specific and (usually) short periods of time. This allows applications to perform with reasonable health even when there is a spike in the load.
With traditional SFTP servers on the cloud, such an elastic expansion is not possible. Though not a common scenario, when there is a peak in load, the performance of the server would purely depend on available resources only.
Cost
Cost is one of the biggest drivers for businesses to adopt cloud computing and storage space is a direct contributor to cost. For a traditional SFTP, dedicated storage space is needed. The approximate cost of minimal storage space is 150$/month (approx) as compared to a blob storage cost of around 22$/month (approx), thus, approximately 10 times more. [Costs from https://azure.microsoft.com/en-in/pricing/calculator/]
In addition, a back-up logic will need to be incorporated by the application team along with back-up storage. This will further shoot cost much higher compared to various back-up options provided by Azure storage at a nominal cost.
Thus it is evident that usage of a storage account is much more economical as compared to traditional disk storage when it comes to the cloud for file storage and access. This is in spite, charges being applied every time file storage is accessed.
Maintenance
One of the key benefits that cloud computing gives businesses, is that their application teams can focus on the core logic and not on the infrastructure, runtime etc. These ‘managed’ offerings improve not only TTM but also abstracts risks.
A traditional SFTP on cloud machine would force us to dive right back into the maintenance of the SFTP server, SSH keys, connections etc. It will also be among the application teams' responsibilities to maintain access controls, troubleshoot issues etc.
Conclusion
The sections so far highlight that a traditional SFTP is not an ideal fit for the cloud. In fact, most of the benefits and value add provided by the cloud computing environment are nullified with a traditional SFTP.
As there still exist requirements where files are needed to be transferred in a client-server environment, cloud vendors have launched managed offerings such as Azure SFTP Gateway. These offerings capture the essence of FTP while taking advantage of the benefits cloud computing offers.
As highlighted earlier, file transfer only requires sporadic use of resources. It thus becomes a prime use case for container-based SFTP servers (ACI based solution) that can be commissioned and destroyed on demand.
As of the date of publishing of this paper, Strati does not support any SFTP options on cloud and recommends to redesign applications to be cloud-native. In the long run, the economic and operational benefits and mitigated risks will outweigh the costs involved for revamping legacy solutions.
