Design
The RSS storage subsystem provides that rare combination of security, performance, reliability and optimum acquisition, installation and operational cost profiles. Today’s computer performance/cost marketplace, along with so many heavily tested software subsystems, makes it possible for RSS to configure and manage and integrate them all to ‘check all the boxes’. From the simplicity and maintainability needs of a small office to the administration requirements of multi-location, multi-domain organization.
Security? Any bit of company data that doesn’t by its own inner nature need to cross the internet– doesn’t cross the internet: there is no better security possible than denying a target. Compare that against any cloud based solution with data crossing the net every read and write, hosted by services with ‘lots of eggs from all clients in one basket’. Fully supporting cloud functions– but only when useful and necessary, not ‘cloud for cloud’s sake’.
Clients without extra need to comprehend the inner nature of redundant storage setups don’t need to learn anything about them, it all ‘just works’. RSS automatically provisions and monitors and maintains storage operations and integrates them into the fabric of other RSS subsystems such as websites and databases that rely upon them.
Highlights of the storage subsystem’s capabilities and structural components, including links to background and details, are in the two columns on the right.
This is ‘highly performant, highly available storage’ starting in box that could grow to a room full, in groups of up to 16 locations.
- What does RSS mean by ‘Highly available data’ ?: Your data is stored not just on three different storage devices, but on not less than three separate tightly linked affordable ‘ganged together’ servers, servers each that could be plugged in to different power supplies or power circuits, or housed in up to 16 different locations. Competitive file-server solutions locate a few drives in one high-end file-server box (often costing more than the several RSS servers combined), but what happens when the power supply or the CPU board in that box fails? In the cost-effective RSS design: your location can withstand the loss of two smaller scale servers– yet your data remains available. Now that is highly available.
- It’s reasonable to consider a portion of RSS’s cost as a form of insurance– consider the benefit of avoiding cost and loss due to business disruption. And, what a nice change to be able to schedule repairs during working hours without disrupting organization operations!
- In the case of records stored in RSS’s multi-location capable database: for performance reasons three copies of each record are kept per server as well as among servers.
- Note: A major value RSS’s Monitor packages provides is the automatic setup, configuration and management of the many integrated subsystems, including storage. The extensive documentation linked below is useful for those who wish to extend an installation, or otherwise become familiar with internal details. All the value these systems offer is ‘configured, up and running’ when the system powers up– it’s not the client’s burden to ‘learn it all’ then ‘set it up’. Every facility exists to support those wanting a more ‘hands-on’ set-up and/or administration/monitoring experience– but a major value of RSS is: while ‘the ability to know everything about everything’ is fully possible at every stage– it is optional. An RSS system is a great way to start with a base of coordinated working and integrated subsystems, then grow in the direction of your choice.
- High availability File, Block (iscsi) and Object access. Via Ceph: ‘the highly available future of storage’. See our ‘Structure’ tab above to read about the thousands of Ceph deployments and history. By use of Ceph’s administrative GUI (video), RSS delivers access to ‘highly available’ storage in three general forms:
- As ‘Block Devices’, or ‘virtual disk drives’ available to connected devices either in the popular ISCSI format (broadly compatible across known operating systems, such as Windows), or via the ceph ‘rbd’ or ‘Rados Block Device’ format (Linux). Often useful as virtual machine images, Windows secondary data drives or even boot drives.
- As a ‘Network File System’, allowing users to browse files and directories in a traditional way. Simultaneously available system formats include the native CephFS, or broadly compatible NFS protocol or as a ‘Windows share’ via widely popular RSS’s Samba 4 servers. RSS Admin GUI allows for very fine-grained access control among users, groups, machines and even individual subsystem services running among others on a device (See RSS’s Admin section).
- As ‘Long Term Storage Objects’ via Cephs ‘Object Gateway’ for direct use by programming languages and client subsystems using the RESTful API. Many website systems deploy compatible interfaces, including Amazon S3, and some databases and enterprise management systems. The RSS speed and security difference is: your client information never has to cross the internet to take advantage of the latest in storage subsystem capabilities.
- Monitoring Device Health and “Hot Swapping” failed storage devices. The Ceph ‘Dashboard GUI’ provides alerts and useful-life-remaining predictions about all its active storage elements. When a system loses a drive, it automatically creates replicas of its data among the remaining drives — a feature which generates time for repair staff to engage, then swap out the failed unit. That’s peace of mind in action. Want more time to replace failed drives? Just add more storage devices per server than you need to maintain three copies, and the system will automatically move content to working drives when one has a problem. Whether in-house staff manages swapping out drives, or local IT support services– it’s meant to ‘all happen’ during scheduled visits during normal hours without disrupting client operations.
- Notice all of the above is in keeping with the RSS security promise– the only time your data crosses the internet is when you decide it absolutely has to, not because ‘cloud for cloud’s sake’. It’s not often performance, security and reliability go together, yet that’s what RSS systems offer.
- Integrated Subsystems:
- Ceph – “The future of storage”. Updated annually since 2012 and deployed by over 4,000 enterprises across the corporate size spectrum: Ceph strikes the ‘sweet spot’ providing highly reliable, scalable storage with a long term and watchful eye on capital and operational costs. By design Ceph delivers the full capability of the whole range of storage hardware performance and architecture– from commodity spindle drives, to solid-state devices and high-end special purpose solutions. General Information. Details.
- At each of up to 16 locations per department, RSS deploys up to seven multi-purpose servers which provide bare-metal storage functions, each location may add up to 55 further dedicated-purpose storage servers. There are two commonly expected installation variations:
- Company/department/location file/storage solution — most typical. A modest installation might consist of used 2U or 3U half-height cabinet rack mounted servers costing around $900 each, each populated with $500 worth of 5 7200 rpm 1TB spindle drives. Four of those and a fifth as a hot-spare with just two drives. Because the design expects and tolerates failure, and makes use of all resources in parallel, this system clocks in at much less than the cost of a single high end new server while providing service even should one, often two of the systems fail. The cost is roughly proportional to the speed and size of the spindle drives or solid-state memory chosen.
- To gauge the size required by your potential application: As noted below, RSS generally stores three ‘live’ copies of all important information and does basic automatic compression. Each storage server advises 2 1TB drives for internal functions. To optimize performance: RSS recommends each storage server have four additional drives each, roughly the same size and speed. So Drives Required DR = 3*(total compressed storage need) / (Drive size) such that DR is greater than or equal to 4 drives/server. RSS/Ceph automatically balances data across all available drives, delivering best-available performance.
- Notice this RSS facility discussion regards live, shared and fully available storage provided in a fault tolerant way — which is critically very different than ‘backup’. RSS recommends very high capacity additional removable drives to provide backup functions. (We advise more, smaller capacity, drives for operations as the speed of data access is much less than the maximum drive capacity. Many new storage admins who appreciate the low cost of very large drives are surprised later when the system ‘that can store so much’ seems ‘so slow’. Backup capacity is not live performance.)
- Communications/connectivity administrative solution — usual for satellite locations and implementations that require storage enough for RSS internet and communications related functions, but are not intended to provide long term file backup or short term highly-available file or data block access beyond RSS’s internal needs. As few as three low end ‘servers on a card’ with three small storage devices each would suffice.
- Company/department/location file/storage solution — most typical. A modest installation might consist of used 2U or 3U half-height cabinet rack mounted servers costing around $900 each, each populated with $500 worth of 5 7200 rpm 1TB spindle drives. Four of those and a fifth as a hot-spare with just two drives. Because the design expects and tolerates failure, and makes use of all resources in parallel, this system clocks in at much less than the cost of a single high end new server while providing service even should one, often two of the systems fail. The cost is roughly proportional to the speed and size of the spindle drives or solid-state memory chosen.
- RSS Monitor takes responsibility for configuring and deploying Ceph in an ‘RSS standard’ fashion: Any data trusted to RSS is stored not only on no fewer than three separate devices, but those devices are each on different servers. As the smallest RSS storage solution incorporates no fewer than four servers, administrators can still provide access should even any two servers fail at the same time. (Note: While access remains possible in such degraded conditions, RSS advises repair of any server or failed storage device before the end of the second business day the failure is detected.)
- Client Customizable – While RSS depends on a very few particular Ceph data storage capabilities, clients may at any time define further data pools, access capabilities and use patterns atop the RSS data storage under-structure.
- Security Technologies – To avoid giving those with ill intent information, RSS server installations at client locations have further security related technologies which are documented only for admin users. However, other than described publicly, nothing in RSS permits client data or metadata to leave the client’s installations.
- Foundational Subsystems:
- Host OS: Ubuntu LTS running on ‘bare metal’ and/or virtual machine/KVM – Each RSS location supports two separate database subsystems. One higher speed but site-local (used for DHCP and security needs not requiring data to be copied in realtime across the net to all company locations). A second, larger capacity, multi-location database subsystem. Each subsystem uses not fewer than four simultaneously running database servers (scales up to 59 at each of up to 16 client locations). Also ‘galera arbiter’ technology to maintain database state across extraordinary outages. Even should only two be running, database services will appear normal. Scales to thousands of users natively. For security, monitoring and performance, all database functions are isolated both in a private subnet and vlan. Larger installations can host database servers on ‘bare metal’ hosts, smaller installations host database engines in an isolated virtual machine both to protect the overall system from email breaches and to allow the installation of the latest database related patches without concern for dependencies and incompatibilities in unrelated subsystems. While ‘docker’ and related isolation technologies provide small speed advantages, virtual machine isolation remains the gold standard for security and is therefore the basis for RSS database hosts running as a subsystem on a server. The underlying Linux distribution for the database subsystem is Ubuntu. Ubuntu Linux General Information, Details. KVM General Information , Details.