File Fabric Sizing Guide
The Enterprise File Fabric™ solution can be deployed as a single virtual machine or node. We recommend a single node for smaller environments.
To support larger environments and high availability the solution can be both scaled-up and scaled-out. Additional nodes can be added (with one node just providing database services), and larger nodes can be deployed.
Recommended sizing when deploying as a single node.
|Type||Users||Files||Memory||vCPU||Disk OS||Disk DB||Notes|
|Small||500||1,000,000||6 GB||6||60 GB||60 GB||Minimum|
|Medium||2,500||10,000,000||8 GB||8||100 GB||100 GB||Recommended for Evaluation|
|Large||5,000||50,000,000||16 GB||16||200 GB||200 GB|
If you are using M-Stream add an additional 4 vCPUs.
We are often asked about recommended architecture when companies wish to try and design their Enterprise File Fabric architecture for production use for 'x' number of users.
Let’s first try and define some terms that we can use to dig into this.
‘Concurrent users’ is defined as the number of users actively using the system at the same time. These ‘Concurrent users’ will all have an active session and will be engaged in a range of activities on the system. In addition, we define ‘Throughput’ as the requests per second that the system has to facilitate.
When any number of users are logged in they are not all necessarily synchronously interacting with the File Fabric. At any particular point in time, some users may be searching for data, others may be browsing file, others uploading or downloading a file, and perhaps some will stop for a minute to check email and not be actually interacting with the File Fabric at all.
We could make the assertion, for example, that there are 1000 concurrent users, but they are not all making 1000 concurrent requests. The throughput that such concurrent users generate would only be based on their actual interactions at any given moment in time and even these interactions may be dealing with cached data layers that do not necessarily increase system throughput.
This is often forgotten when people start thinking about the number of users they have to server or cater for. We are often asked to help prospect and customers to calculate that ‘x’number of users = 'y' Throughput.
We can provide some guidance on requests like these, but only if we have detailed data sets as baselines. Every deployment is different, and is heavily influenced by use cases and other factors such as network latency, storage under management: ie. each deployment has multiple variables that change the equation outside of just the number of users.
As a minimum single machine configuration please consider the below but as outlined above this will be dependent on the use case, number of users, and modules deployed but as a minimum:
- 8 GB RAM for the Virtual Machine
- 4-6 vCPUs
- 36GB for OS Volume
- 40GB for DB Volume on SSD or performant storage
Remember that the File Fabric is horizontally and vertically scalable, so if the initial design is thought out and executed well, scaling out in real-time based on demand can easily be achieved.
Please download and review the comprehensive sizing document to assist in scoping out enterprise deployment sizing.