File and Object Storage

File and Object Storage

Software-defined storage for building a global AI, HPC and analytics data platform 

 View Only

IBM Systems Technical University - Day 5 morning and Final Thoughts

By Tony Pearson posted Fri May 26, 2017 05:43 PM

  

Originally posted by: TonyPearson


Banner-IBM-TechU-Orlando-2017

This week, I am presenting at the IBM Systems Technical University in Orlando, Florida, May 22-26, 2017. Here is my recap of the sessions on the morning of Day 5, the last day of the conference.

Integrating IBM Storage in Container Environments

Dr. Robert Haas, IBM CTO Storage for Europe, presented IBM Storage for Docker containers. These are different from containers in IBM Cloud Object Storage, and different from the Container Pools used in Spectrum Protect.

Robert gave an overview of IBM Spectrum Conductor, part of the IBM Software Defined Infrastructure (SDI) Spectrum Compute family of software products. The goal is to analyze large amounts of data, access these data efficiently, and protect the data, results and insights as intellectual property.

IBM Spectrum Compute comes in several offerings. IBM Spectrum LSF (Load Sharing Facility) manages long-running batch jobs for modeling, design and simulations. IBM Spectrum Symphony provides low-latency for risk analytics in the financial services sector. IBM Spectrum Conductor comes in two flavors. Conductor for Spark (CFS) manages Spark analytics. Conductor for Containers (CFC) handles Docker and Kubernetes containers.

Docker is the run-time platform. While there are other container run-time platforms like RKT and LXD, Docker is clearly the marketshare leader, growing 40 percent per year.

Statistics from the latest DockerCon2016 conference showed the most popular use cases and workloads for Docker. What can run in Docker: Lots of applications can be "containerized", including Redis, MongoDB, PostgreSQL, OracleDB, Java, to name a few. Docker is well established in enterprises, including service providers, healthcare, insurance and financial services, public sector, and technology firms.

Kubernetes, Mesos and Docker/Swarm are a layer above, as orchestrators. Spectrum Conductor for Containers uses Kubernetes and other open source tools to coordinate activity. Orchestrators restart failed applications, and can scale up or scale down the number of instances as needed. Orchestrators can manage groups of applications, across clusters on-premises and off-premises Cloud.

From a storage perspective, containers access storage like bare-metal operating systems, bypassing all of the layers normally associated with bloated Virtual Machine hypervisors. It also eliminates single root I/O virtualization (SR-IOV) that VMs use to compensate.

Persistent storage can be isolated, so that containers cannot see the files of other containers. This provides multi-tenancy.

There are four ways to attach storage.

  1. Writeable layer, ephemeral (non-persistent) storage.
  2. Internal persistent storage (directory on host file system). However, if you move a container from one host to another, you may lose access to this internal storage.
  3. External volume, manually mounted.
  4. Volume driver plug-in REST API that automatically mounts it.

 

The fourth method is preferred. Plug-ins are available for IBM Spectrum Scale, GlusterFS, Portworx, Rancher Convoy, RexRay, and Contiv. The start-up Flocker have gone out of business last year.

The Docker hosts can attach to IBM Spectrum Scale in all of its supported offerings, including POSIX, NFS and SMB protocol. Containerized applications can move from one Docker host to another, and continue access the IBM Spectrum Scale namespace.

IBM has created the "Ubiquity Volume Service" that provides a consistent API for Docker and Kubernetes. This will use IBM Spectrum Control Base Edition to support IBM Spectrum Scale, Spectrum Accelerate, Spectrum Virtualize and DS8000 storage systems. For IBM Spectrum Scale, volumes are mapped to iSCSI volumes, filesets or directories. For other devices, volumes are mapped to block LUNs. Ubiquity is publicly available on GitHub.

Enterprise Applications for IBM Cloud Object Storage

Andy Kutner, IBM Cloud Architect, presented the various options available for NAS gateways that can front IBM Cloud Object Storage.

Ctera offers NAS gateways, and Endpoint agents for backup and Enterprise File Sync & Share (EFSS). This vendor targets Remote Office/Branch Office (ROBO) and small NAS consolidation that have less than 60 TB per office IBM is a reseller of Ctera, so you can get both Ctera and IBM COS from the same IBM sales rep.

Nasuni offers a global file system, accessible from any device, smartphone, tablet or desktop. They are focused on taking out EMC and NetApp NAS solutions. Performance at the edge, combined with capacity in the client's chosen Cloud (including IBM Cloud Object Storage or IBM Bluemix). Infinite snapshots replace backups, offering RPO of 1 minute for Disaster Recovery. Their global file system "UniFS" offers file locking.

Panzura focuses on Cloud Integrated NAS, File Distribution, and Collaboration. This can help eliminate "islands of storage". The File Distribution can be any type of file, but was originally designed for Media and Entertainment, such as videos. Collaboration employs EFSS features for workgroup shared file folders, such as CAD/CAM or engineering blueprints.

IBM Spectrum Scale can provide NFS and SMB access to files, and then move colder, less active data to IBM Cloud Object Storage, using Transparent Cloud Tiering feature. Spectrum Scale offers WAN caching across locations.

IBM COS now offers a native NFS v3 interface. This allows read/write NFS access, with S3 API read of the same content. Each file is mapped to a single object.

This is targeted for large scale archive, static-and-stable data, NFS-based backup software, and applications going through the transition from file-based to object-based. This is not intended for multi-site collaboration or primary NAS replacement. Regardless of the number of geographically dispersed IBM COS sites, the NAS can run on only one or two sites initially.

To provide NFS v3 support, IBM introduces new F5100 File Accessers, which talk to an IBM COS Accesser, which in turn acts on specific Vaults in the storage pools. The file-to-object mapping metadata is replicated on-premises across three File Accessers, and optionally replicated asynchronously to a second site for High Availability. S3 API can read access the file by file name, or by Object URI.

Initially, the "File Accesser" is only available as pre-built system, not as software-only.

There was not enough time to cover other solutions, including Avere, NetApp AltaVault, or Open Source S3FS.

This was a great event, just the right size, between 1,500 and 2,000 attendees. Similar IBM Technical University events coming up later this year:

  • August 8 - 10 -- Sao Paulo, Brazil
  • August 15 - 17 -- Melbourne, Australia
  • September 12 - 14 -- Johannesburg, South Africa
  • October 9 - 13 -- Munich, Germany
  • October 16 - 20 -- New Orleans, LA, USA
  • November 6 - 10 -- Prague, Czech Republic
  • November 13 - 17 -- Washington, D.C., USA

Check out the [IBM Systems Technical University 2017] page.

Save the date: Next year, we will be back in Orlando, April 30-May 4!

technorati tags: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

0 comments
8 views

Permalink