Hyperscaler



• The term “hyperscale” refers to a computer archit1ecture’s ability to scale in order to respond to increasing demand.

Computers rely on resources within a given node or set of nodes. Scaling some part of our computer architecture typically means increasing computing ability, memory, networking infrastructure, or storage resources.
Hyperscale is not just the ability to scale, but the ability to scale hugely and quickly.
The goal of scaling is to continue building a robust system, whether that system revolves around the cloud, big data, or distributed storage, or, as is increasingly likely these days, a combination of all three
https://www.bmc.com/blogs/hyperscale-data-center/

• A hyperscale cloud data center looks different from an enterprise data center, or even a large hosting provider. 

Here are some of the significant differences to think about if you’re considering a hybrid cloud that involves you running Azure Stack or a hyperconverged infrastructure:
https://www.computerworld.com/article/3138432/inside-a-hyperscale-data-center-how-different-is-it.html

• Hyperscaler Storage

Enterprise hyperscaled storage is one of the more recent storage trends SNIA is focusing on to advance storage and enable the future of IT. SNIA’s research indicates that large enterprises, in addition to utilizing cloud storage, are adopting methods from Hyperscalers such as Amazon, Facebook, Google and Microsoft Azure to build their own storage systems using software defined storage (SDS) and best-in-class commodity components, assembled in racks.
https://www.snia.org/hyperscaler

• Software-defined: Infrastructure where the functionality is completely decoupled from the underlying hardware and is both extensible and programmatic.

Commodity-based: Infrastructure built atop commodity or industry-standard infrastructure, usually an x86 rack-mount or blade server.
Converged: A scale-out architecture where server, storage, network, and virtualization/containerization components are tied together as a pretested, pre-integrated solution.
Hyperconverged: A scale-out architecture that takes converged infrastructure one step further by combining software-defined components atop commodity hardware, packaged as a single solution -- often a single appliance
Hyperscale: A scale-out architecture that is also software-defined and commodity-based, but where the server, storage, network, and virtualization/containerization resources remain separate. Each component is distinct and can be independently scaled.
https://www.infoworld.com/article/3040038/what-hyperscale-storage-really-means.html

• Today’s hardware-defined infrastructure

Software-defined infrastructure: the first step toward hyperscale

Hyperscale data-center-driven use cases

Individual component replacement

Flexible host setup

Dynamically optimized data centers

https://www.ericsson.com/en/white-papers/hyperscale-cloud--reimagining-data-centers-from-hardware-to-applications

• Veritas HyperScale Architecture

Veritas HyperScale for containers is based on a unique architecture to provide resilient storage with predictable performance. The predictability is provided by internally separating the storage for secondary operations from that required for primary operations. HyperScale employs an architecture with services segregated into two horizontal planes; the top plane (Compute Plane) responsible for active/primary IO from application containers and the lower plane (Data Plane) responsible for version(snapshot) management of volumes and the usage of these snapshots for secondary operations like backup, analytics

https://vox.veritas.com/t5/Software-Defined-Storage/Predictable-performance-with-containerized-applications/ba-p/829920

• We are very excited to introduce DirectFlash™ Fabric which extends the DirectFlashTM family outside of the array and into the fabrics via RDMA over converged Ethernet (also known as RoCEv2).

DirectFlash Fabric is generally available as of January 16, 2019 with a non-disruptive firmware upgrade and installation of a RDMA capable NIC in the storage controllers.

At hyperscale, efficiency is a requirement.  In the beginning, there was “web-scale”.  This was the simple and easy way to scale.  Our hyper-converged friends are still in this early evolution.  As you saw with the Stanford Flash Storage Disaggregation study, this type of architecture was immature and equates to high levels of inefficiency at scale.  To enable independant scale of compute and storage many hyperscale environments went to “rack-scale disaggregation.”

https://blog.purestorage.com/directflash-fabric-continuation-of-pures-nvme-innovation/

Atlantis USX (Unified Software-defined Storage) is a software-defined storage solution (100% software) that delivers the performance of an all-flash storage array at half the cost of traditional SAN or NAS.

https://vinfrastructure.it/2015/05/atlantis-hyperscale-products/

•     Choose hyperscale when… your organization has 5,000 employees or more, more than 500 terabytes of data, more than 500 applications, or more than 1,000 VMs.

    Choose hyperconverged when… you’re below these watermark numbers, have five or fewer staff managing your virtual infrastructure, or you’re in a remote or branch office.



https://www.enterpriseai.news/2016/11/14/demystifying-hyperconverged-hyperscale-storage/

power analysis

• In cryptography, power analysis is a form of side channel attack in which the attacker studies the power consumption of a cryptographic hardware device (such as a smart card, tamper-resistant "black box", or integrated circuit). The attack can non-invasively extract cryptographic keys and other secret information from the device.

Simple power analysis (SPA) involves visually interpreting power traces, or graphs of electrical activity over time.
Differential power analysis (DPA) is a more advanced form of power analysis, which can allow an attacker to compute the intermediate values within cryptographic computations through statistical analysis of data collected from multiple cryptographic operations
https://en.wikipedia.org/wiki/Power_analysis

• Using this information, the DPA attack uses the following steps:

    Use the current/previous plaintexts and ciphertexts to calculate four different possible AES inputs (the inputs for ECB, CBC, CFB, and OFB modes)
    For each of these calculated inputs, use one bit of the input to split the traces into two groups
    Calculate an average trace for each group and subtract them to get four differential traces
    Look at the differences to decide which mode is most likely:
        If one of the differential traces shows a large spike, the target is probably using that mode
        If none of the differential traces has a large spike, the target is probably using CTR mode
https://wiki.newae.com/Investigating_Block_Cipher_Modes_with_DPA