Upcoming commission: storage for machine learning

Data storage is an often-overlooked part of machine learning and other AI deployments.

This article will appear in Computer Weekly. It will cover:

  • Definitions of machine learning/deep learning
  • Its storage requirements including
    • Sizing, capacity, performance (to match compute)
    • Scale
    • Media (SSD vs HDD, hybrids of the two)
    • Parallelism
    • Throughput vs IOPS
    • Locations – including use of the cloud

For this article we are open to comments from vendors, as well as analysts, consultants and other experts. Examples of ML use cases and how systems were designed to run it are most welcome.

Initial pitches and leads by Wednesday May 29th by email please.

Upcoming commission: virtual server storage

My next article for Computer Weekly will look at the best storage options for virtual servers, including SAN, NAS and hyper-converged infrastructure (HCI).

Specifically, the piece will ask:

  • What kind of storage requirements virtual servers and their data have?
  • What are the characteristics of a) SAN b) NAS and c) hyper-converged storage? 
  • What are the pros and cons of SAN vs NAS for virtual machine storage? What are the management and performance issues?
  • What about scale? Is a SAN, NAS or HCI better suited to large or smaller deployments?
  • What impact do workloads have on storage choices? Are all virtual machine workloads created equal in I/O terms?
  • What other factors affect storage choices, such as the applications being used, scale of the deployment and even skills on the IT team?

First and foremost I am looking for background information, analyst research/ technical papers and case studies which will help to answer the points above. If you or your client has expertise in this area, please contact me by email in the first instance. The deadline for input is Monday, 13 May.