Written By: Stephen Scribner, Systems Engineer, ScaleIO
I joined the ScaleIO team as a Systems Engineer in a very busy Q4 of 2015. Since joining the team there have been a TON of releases, both from a product standpoint as well as the ecosystem supporting and around ScaleIO. Even more recently we just posted the latest release in Software-Defined Storage at EMC - ScaleIO 2.0 which is the first major code update we’ve had in some time. I’ve had the opportunity to talk to a LOT of customers about their strategy for success in the new age of Bi-Modal IT. Often I’m asked how ScaleIO can solve “our” ever growing challenge of managing more data with less resources and time; with an expectation of delivering more value back to the business than ever before. Things like data migration, technology refreshes, and high CAPEX, are driving customers to look for a new way to do IT.
When I decided to write a blog about ScaleIO, the first thing I thought about was ScaleIO adoption in the SDS & Hyperconverged market. Many other products in the space tend to focus on varying design points in their architecture such as tight hypervisor integration, simple management, various data services etc. ScaleIO has taken a different approach, with a focus on delivering a highly available, low-TCO, software-defined strategy at Data Center scale. ScaleIO has been created and designed around four distinct areas. Every architecture and feature we add to our product is designed to lower cost, increase elasticity, improve performance, as well as add better automation and management.
Chad Sakac – the fearless leader of EMC Presales and now VCE, wrote a great subject on System Architectures and specifically how things like ScaleIO (Type-3) interact with and manage data. One of his great points on this write-up is that there is not a “one size fits all” approach. When customers choose the right system for the right workload – only great things ensue. Check it out if you haven’t yet:
Now onto some of the recent updates with 2.0 Specifically, we added a number of features that target resiliency. We’ve continued to focus on ensuring data is always highly available – something you should expect from all enterprise storage solutions. In addition, new capabilities provide customers with expanded support options, read caching capabilities as well as efficiencies through maintenance. A quick list below gives you an idea of the things we’ve worked to improve.
5 Node MDM Cluster – This adds an additional two members to the ScaleIO management cluster. We also can now add an additional repository for managing the system metadata – further ensuring continued uptime through failure.
Read Flash Cache – ScaleIO now includes integrated Read caching into the GUI as well as SDS systems. This will work on any flash media compatible with ScaleIO.
ESRS – This one is great for ScaleIO supportability. EMC can now remotely monitor as well as diagnose and repair issues. Huge add for an EMC customer with any number of systems to manage.
Instant Maintenance Mode – Another slick feature for customers who mange and maintain ScaleIO systems. Administrators can now query and place individual hosts in a “maintenance mode” that will prevent against rebuilds for simple operations like patching, reboots, and updates.
NDU Orchestration – Customers can certainly expect to upgrade existing ScaleIO systems from 1.3x versions non-disruptively. We’ve added in a few new features to enhance the upgrade and ongoing cluster maintenance.
There have been a few other cool features added that customers are giving us great feedback on; GUI Front end Management feature has been added to 2.0 – something a number of my customers have been asking for. This one is huge for some of our customers who are planning to use ScaleIO in replacement of some of their traditional storage subsystems. Adding a simple management interface for all the day-to-day tasks was a must have. Many customers will continue to leverage the built in CLI and REST API for highly automated environments.
2.0 brings along Performance Modes, introduced into the framework to include built-in GUI performance tuning.. The goal is to reduce and even likely eliminate many of the manual tasks that were once required for garnering the best performance in ScaleIO clusters. The new functionality will do things like increase I/O buffers, adjust queue depth, memory, as well as some other tunable features (nerd knobs) within the OS environment. What I’ve noticed in my lab environment is about a 20% cluster wide IOPS increase from Default to High Performance with a single click.
Check out some of the tests we recently ran in our Hopkinton Data Center. All workloads are running on 3x Commodity SSD’s across 8 hosts with 2x 10G ports per host (a fairly standard configuration). What you’ll see is that even at relatively small scale (24 drives) ScaleIO performs extremely well.
Test 1 – 8k 100% random read – these are the hero numbers folks
Test 2– 32k 100% random read – notice the linear scaling when we up the block size.
Test 3– 32k 50%Read/50% Write – Key to notice the linear distribution of Writes and Reads across each node & disk in the cluster.
Before joining the ScaleIO team I worked as a presales engineer covering a few large EMC accounts. Something that was abundantly clear to me was the ever growing pressure from their internal business, external competition, as well as forces from the “cloud”. I think ScaleIO is a great fit for customers who have a large amount of block data with a growing challenge in continuing to manage and maintain it. Obviously we’ve added a number of features that are designed to drive more performance, better resilience, and enhance the overall experience with ScaleIO. If you think this is something you’d like to try out go for it! You can download ScaleIO today for non-production environments here (http://www.emc.com/products-solutions/trial-software-download/scaleio.htm) .