News Precis#9 – 30 October 2020

News Precis#9 – 30 October 2020

Chris Evans News Leave a Comment

A roundup of news and views on technology that piqued at least a little bit of interest. Fit to print? I’m sure they are.

Vendors

  • Scality announced the availability (in preview) of SOFS for Microsoft Azure. The premise is a low-cost scale-out file system backed by Azure blob storage running on Azure virtual instances that’s cheaper than a traditional scale-out file storage solution. I like the idea but the reporting of performance data is a mess. The solution quotes 1 Terabit per second performance – who measures storage throughput in bits? A quick calculation shows this system would require 224 VMs to run. There’s no mention of networking costs or the cost of using CosmosDB as a metadata server or the backend storage. There’s no data on multiple workload profiles. Perhaps if the solution is popular, we might see a deeper analysis.
  • Intel sells NAND business to SK Hynix. Not a bad $9bn payday for Intel and more importantly this signifies a focus on processor technology. Optane is not included in this sale (no surprise) as that technology is likely to provide higher margins over the coming years.
  • AMD to acquire Xilinx. Another aspect to the ongoing lovefest over SmartNICs. Architecture are evolving and hardware is on the up as we look to drive even greater performance for applications. Exactly how this all plays out is going to be so interesting to watch and we can see now why Intel is offloading distractions like NAND storage.
  • IBM divides into two companies. Even the BBC reported this one. IBM will spin off the managed infrastructure services part of the GTS business unit, leaving behind hybrid cloud, AI, quantum computing, mainframes and consulting. If I were a cynical person, I’d say that IBM is simply offloading the low margin solutions and keeping the rest, a bit like offloading the PC and server business in the past few decades. However, there’s no point keeping a business unit if it isn’t either strategically valuable or returning good value on cash for the shareholder. Must still be some money in those “legacy” mainframes…

Personal Computing

Industry

  • OpenStack becomes Open Infrastructure Foundation. I’ll be honest, I never warmed to OpenStack or the culture behind it. Developing code in a weekend isn’t a badge of honour, it’s a recipe for future technical debt. That’s what OpenStack is, a minefield of future technical debt packaged in a free software solution. There will still be places where OpenStack can play, but those will be few and for companies prepared to build proprietary stacks.
  • Goodbye Hadoop. I see Hadoop as another proprietary solution that never made the grade. I did try it out many years ago, but wasn’t 100% convinced and like many technologies I struggle to validate, I dropped it from my radar. That radar usually proves right and here we are with Hadoop fading away. Will anyone mourn?

Startups

  • Kasten is acquired by Veeam. Great news for Niraj and the team but I worry about integration. Veeam loves creating new, separate solutions then packing them with features. In a future hybrid world, that’s not going to fly. Integration will be the key and that’s the area I’m watching. Incidentally, this was a “cash and stock” transaction, however Veeam is a private company, so exactly how does Kasten quantify the value of the transaction – and more importantly – crystallise their assets in the future?
  • Docker pauses potential Hub changes. Docker did well to innovate and bring containers to the masses, but the company is a shadow of it’s former self. Nothing exemplifies this than the rate limiting to be imposed on images download from Docker Hub. Whilst the idea of downloading from a centralised hub is a good one, there’s no incentive to “build your own” and life is easy when you can just re-pull from a free repository. Perhaps we need a rethink on how repositories are used in general.

Post #prec009. Copyright (c) 2020 Brookend Ltd. No reproduction in whole or part without permission.