Educational Episodes: Hyperconverged Infrastructure (HCI)

The post describes the history of RAID 5 and how it became obsolete at some point in time, just because HDD capacity grew at an enormous rate. It happened due to the chance of failure that grew to literal imminence when spinning disks reached TB scale, because the reading speed still had the same physical limits. Basically, creating a RAID 5 even with 1 TB disks would mean certain failure of the whole array and quite soon. The array technology was “saved” by an unlikely ally – the SSD. Being faster than hard disk drives in everything, they almost nullify the chance of the abovementioned failures. The post is written for everyday reader, not just engineers, and is quite comprehensive even without special knowledge and skills.

Here is a comparison of two technologies with similar task but different methods of accomplishing it – Link Aggregation Control Protocol (LACP) and Multipath I/O (MPIO). Both are aimed at providing higher throughput when one connection can’t handle the task. To achieve that, LACP bundles several physical ports into a single logical channel. MPIO, on the other hand, utilizes more than one physical path, even if the working application does not support more than one connection. Both technologies seem to be equally effective at first glance, but further study confirms that one of them is better at achieving its goal. The post is practical, so expect detailed research with screenshots and complete analysis of the technologies in a test case.

This research is, basically, an answer to some statements about NFS shares and Hyper-V Virtual Machines that StarWind engineers considered false. Hyper-V is a native hypervisor from Microsoft and one of the most popular ones. It is capable of creating virtual machines on x86 and x64 Windows systems. Microsoft suggests that VMs should be stored in the Cluster Shared Volumes, in case live migration and failover are required. Storing them on an NFS share was never present in the design, so StarWind team was a bit skeptical about the idea at the first place. This research contains the full experiment, aimed at proving or disproving the initial statement, complete with a few different checks. The result is quite an expected one, but nevertheless, curiosity is strong in StarWind engineers.

There is a new feature in Windows Server from Microsoft, called Soft Restart. The idea itself is perfect, because its goal is to minimize downtime during reboot process. While it is possible to reduce software starting sequence time, when starting the hardware there are physical time limits. These limits made it impossible to reduce reboot time, before Microsoft thought about the “soft restart”. It is claimed to perform standard reboot without restarting the devices, basically resetting only software and reducing the required time dramatically. StarWind engineers decided to check if Soft Restart works fine in two use cases – with a simple physical setup and with virtualized infrastructure. As it always goes with our practical approach, the fool experiment complete with screenshots is in the post.

This post is about Microsoft Storage Replica, a new solution introduced by Microsoft in Windows Server 2016. Basically, it enables replication between two servers, clusters or inside a cluster, also being capable of copying data between volumes on the same server. Storage Replica is often utilized for Disaster Recovery, allowing the user to replicate data to a remote site, thus being able to recover from complete physical failure on the main location. The post is dedicated to building a “shared nothing” cluster. It is an experimental part of a series and features a “Shared Nothing” Hyper-V guest VM cluster. As always, there is a detailed instruction as to how to create the subject setup and there are results for everyone to check.

This post is dedicated to the new solution from Microsoft – the Microsoft Storage Replica. It is a practical part of a series of posts about this technology and features a “Shared Nothing” Hyper-V HA VM Cluster in practice. Microsoft Storage Replica is designed to perform replication between various media: servers, clusters, volumes inside a server, etc. It’s typical usage scenario is Disaster Recovery, which is essential for data protection in case anything happens to the main location. Critical data is replicated to a remote site, often located hundreds and thousands of miles away for better data safety. The experiment is performed by StarWind engineers, so the post contains detailed instructions and a comprehensive conclusion.

Awesome! Got some new goodies from Mellanox – brand new 40 GbE & InfiniBand gear. We officially love these guys! Stay tuned for new performance tests.

This post is about Microsoft Storage Replica – a disaster recovery tool, introduced by Microsoft in Windows Server 2016. It is a part of series about this technology, which is all featured on this blog. Basically, the post is a practical guide for building a “Shared Nothing” Scale-Out File Server, prepared by StarWind engineers after they did it themselves. The “Shared Nothing” is an architecture, which implies that each node is self-sufficient and completely independent, which makes the system more reliable. It got its name because the nodes don’t have a shared storage at all, moving closer to eliminating the single point of failure. The architecture is almost infinitely scalable, becoming popular with use cases where unpredictable and explosive growth is typical.

This is the first test in the series, dedicated to Microsoft Storage Replica – a new solution introduced in Windows Server 2016. It shows a step-by-step process of “Shared Nothing” Failover File Server building, based on the Server Message Block protocol (SMB3). Microsoft Storage Replica is a versatile data replication solution. It performs replication between servers, between volumes in a server, between clusters, etc. The typical use case for Microsoft Storage Replica is Disaster Recovery, which is basically replication to a remote site. It allows easy recovery of data in case the main site operation is disrupted by some sort of a force majeure, like natural disasters. The test is performed by StarWind engineers with a full report on the results.