Avere Systems has an idea of how to boost performance for those storing data in the cloud. It’s called the Edge Filer.
The basic idea is to use the principles of cache to reduce latency for the end user. Their Edge Filer appliance stores active data near the users and application servers to reduce the latency associated with transferring data over the WAN. It offloads sending data to the cloud for long-term retention in batch fashion. The application is not affected when data moves from one place to another, as the Edge Filer hides the details of that.
Their model is appropriate for those who have a combination of in-house and off-site cloud storage as opposed to those who are using all-cloud storage.
The diagram below from Avere Systems shows how their system is laid out.
The Edge filers can front-end multiple cloud object storage systems at once as well as local and remote network area storage (NAS). It supports object storage and traditional hierarchical file systems.
Its caching algorithm classifies data into three categories: (1) hot, (2) warm, and (3) cold. It does that by predicting whether data that is warm or cold and which data might become hot or warm, taking into consideration what data is currently in use and calling that hot. Although it is not spelled out in their whitepapers, their algorithm must in part follow the widely adopted premise that 90% of a company’s data is never accessed after it was first created and most data is never accessed again 30 days after creation.
The Edge filer stores hot data in and reads hot data from RAM (memory) and NVRAM (non-volatile ram). Warm data is read from and stored in solid state storage (SSD) and SATA (hard drives connected with fiber cable) storage. Data that is cold is sent to the core filer, which is what they call the cloud or the existing NAS storage system. The word “filer” just means storage, and “core” means data not stored in the Edge.
The system creates a global name space (one mount point) so that applications do not have to refer to storage by using different names depending where the data is located. The Edge filer moves hot or warm data to cloud storage when, for example, the user has updated a record and now is ready to save the change. Such changes and additions are sent to the cloud in a batch schedule set by the administrator.
The system also has the ability to convert data from native Windows and UNIX SMB\CFS and NFS file formats to object storage format. That would be required if, for, example you wanted to use Amazon S3 to store Windows file and block structure data, meaning a hierarchical file system.
The Edge Filers can scale vertically. The company says it reduces the total cost of ownership (TCO) by letting you move data for long-term storage to less expensive storage devices while deploying the most expensive and faster storage (solid state and SATA) closer to the application servers and users.