Kemp Keeps ECS Balanced

Disclaimer: I recently attended Dell Technologies World 2019.  My flights, accommodation and conference pass were paid for by Dell Technologies via the Media, Analysts and Influencers program. There is no requirement for me to blog about any of the content presented and I am not compensated in any way for my time at the event.  Some materials presented were discussed under NDA and don’t form part of my blog posts, but could influence future discussions.

As part of my attendance at Dell Technologies World 2019 I had the opportunity to attend Tech Field Day Extra sessions. You can view the videos from the Kemp session here, and download my rough notes from here.

 

Kemp Overview

Established early 2000s, Kemp has around 25000+ customers globally, with 60000+ app deployments in over 115 countries. Their main focus is an ADC (Application Delivery Controller) that you can think of as a “fancy load balancer”. Here’s a photo of Frank Yue telling us more about that.

Application Delivery – Why?

  • Availability – transparent failover when application resources fail
  • Scalability – easily add and remove application resources to meet changing demands
  • Security – authenticate users and protect applications against attack
  • Performance – offload security processing and content optimisation to Load Balancer
  • Control – visibility on application resource availability, health and performance

Product Overview

Kemp offer a

LoadMaster – scalable, secure apps

  • Load balancing
  • Traffic optimisation 
  • Security

There are a few different flavours of the LoadMaster, including cloud-native, virtual, and hardware-based.

360 Central – control, visibility

  • Management
  • Automation
  • Provisioning

360 Vision – Shorter MTTD / MTTR

  • Predictive analytics
  • Automated incident réponse
  • Observability

Yue made the point that “[l]oad balancing is not networking. And it’s not servers either. It’s somehow in between”. Kemp look to “[d]eal with the application from the networking perspective”.

 

Dell EMC ECS

So what’s Dell EMC ECS then? ECS stands for “Elastic Cloud Storage”, and it’s Dell EMC’s software-defined object storage offering. If you’re unfamiliar with it, here are a few points to note:

  • Objects are bundled data with metadata;
  • The object storage application manages the storage;
  • No real file system is needed;
  • Easily scale by just adding disks;
  • Delivers a low TCO.

It’s accessible via an API and offers the following services:

  • S3
  • Atmos
  • Swift
  • NFS

 

Kemp / Dell EMC ECS Solution

So how does a load balancing solution from Kemp help? One of the ideas behind object storage is that you can lower primary storage costs. You can also use it to accelerate cloud native apps. Kemp helps with your ECS deployment by:

  • Maximising value from infrastructure investment
  • Improving service availability and resilience
  • Enabling cloud storage scalability for next generation apps

Load Balancing Use Cases for ECS

High Availability

  • ECS Node redundancy in the event of failure
  • A load balancer is required to allow for automatic failover and event distribution of traffic

Global Balancing

[image courtesy of Kemp]

  • Multiple clusters across different DCs
  • Global Server Load Balancing provides distribution of connections across these clusters based on proximity

Security

  • Offloading encryption from the Dell EMC ECS nodes to Kemp LoadMaster can greatly increase performance and simplify the management of transport layer security certificates
  • IPv6 to IPv4 – Dell EMC ECS does not support IPv6 natively – Kemp will provide that translation to IPv4

 

Thoughts and Further Reading

The first thing that most people ask when seeing this solution is “Won’t the enterprise IT organisation already have a load-balancing solution in place? Why would they go to Kemp to help with their ECS deployment?”. It’s a valid point, but the value here is more that Dell EMC are recommending that customers use the Kemp solution over the built-in load balancer provided with ECS. I’ve witnessed plenty of (potentially frustrating) situations where enterprises deploy multiple load balancing solutions depending on the application requirements or where the project funding was coming from. Remember that things don’t always make sense when it comes to enterprise IT. But putting those issues aside, there are likely plenty of shops looking to deploy ECS in a resilient fashion that haven’t yet had the requirement to deploy a load balancer, and ECS is that first requirement. Kemp are clearly quite good at what they do, and have been in the load balancing game for a while now. The good news is if you adopt their solution for your ECS environment, you can look to leverage their other offerings to provide additional load balancing capabilities for other applications that might require it.

You can read the deployment guide from Dell EMC here, and check out Adam’s preparation post on Kemp here for more background information.

Dell EMC Announces ECS 3.0

ecs_001

Elastic Cloud Storage (ECS to its friends) has been around a little while. Today Dell EMC announced the release of version 3.0. I thought I’d cover off some of the reasons why ECS might be something you’d be interested in. I’ll then go through the new features with ECS 3.0. If you make it that far you’ll be treated to some light opinionalysis to finish off.

 

Why ECS?

Dell EMC provided me with a list of reasons why you might want to consider ECS.

Highly Efficient Data Protection

One of the problems we have is protecting unstructured data at scale. To this end, ECS uses a hybrid protection scheme comprised of triple mirroring, erasure coding and XOR algorithms. The key benefits of this approach are:

  • Lower storage overhead option for cold data scenarios
  • Enhanced data durability without the overhead of storing multiple copies

 

Efficient Large and Small File Storage

  • Small files stored in cache and written to a single disk through box-carting
  • Large files over 128MB in size are erasure coded immediately vs triple mirroring and erasure coded later
  • Provides up to 20% higher throughput for larger files

 

Fully Geo-distributed High Availability & Protection

  • A geographically distributed environment that acts as single logical resource
  • Active/Active platform with access to content through a single global namespace
  • Provides geo-caching to improve operational performance and reduces latency
  • Read/write access from any location globally

 

Comprehensive Data Access

Simultaneous access to underlying data through multiple interfaces

  • Object, File, HDFS
  • Support for S3, Swift, Atmos, Centera CAS, and NFS v3
  • HDFS compatible with Cloudera, Pivotal, Hortonworks, etc.

What does this mean?

  • Native Upgrade path for Centera/Atmos
  • Enables S3 like offering in-house
  • Eliminates storage gateways
  • Breaks down storage silos

 

Native Multi-tenant Architecture

  • Shared storage resources amongst multiple applications and tenants
  • System securely and automatically separates Namespaces, object buckets and users
  • Integration with LDAP and AD environments
  • Ensures the integrity of customers’ stored data

 

Built-in Metadata search

Integrated Metadata storage – store metadata using the same constructs as objects eliminating the need for a separate database and infrastructure to run it.

Metadata search via SQL construct

  • Enables applications and users to query metadata using SQL constructs. Supports several attribute and sort functions.
  • Global metadata search
  • Enables applications and users to search across the global namespace.

 

So what’s really new in 3.0?

So this ECS stuff is great, but what’s exciting about 3.0?

Advanced Retention Management

  • Event Based Retention – Enables application to specify retention period that will start when a specified event occurs
  • Litigation Hold – Enables application to temporarily prevent deletion of an object that is subject to an investigation or legal action
  • Min/Max Governor – Enables system administrator to specify a min and max value for the default retention period

 

This unblocks Centera customers using ARM from migration to ECS. I’m actually really excited about this, mainly because I was a big Centera fanboy and have found it difficult to put forward other EMC solutions to replace it for customers heavily leveraging ARM.

 

SNMP Traps Support

  • ECS 3.0 will support for SNMP Traps for ECS critical events
  • SNMP Traps is an optional feature, based on whether system admin configures SNMP information via UI/API
  • When configured, ECS sends a SNMP Trap to the configured server for any event that causes an alert on the management API
  • ECS supports the ability to configure up to 10 SNMP Trap Destination targets
  • SNMPv2 and SNMPv3 (USM mode) support
  • SNMP Query Service support (CPU & Memory)

 

Remote Syslog Support

  • Shipping ECS Monitoring & Diagnostics logs to a remote syslog server
  • Ability to forward all ECS Audit Logs and ECS Alerts to a centralized Syslog server
  • Forward OS syslog messages
  • Support for UDP and TCP based communication with syslog servers
  • Support for multiple redundant syslog servers , all active
  • Distributed service, resilient to node failures
  • Only System Admin can perform syslog management operations
  • Specify a severity threshold of logs to be forwarded
  • Ability to Add, Edit and Delete Syslog server configuration from the portal and REST API
  • Logs can be seen on ECS nodes in /var/log/<node IP>/syslog.log

 

Platform Lockdown

ECS will support a ability to do the following via the ECS RESTful management API

  • Lockdown an entire cluster
  • Lockdown a specific node
  • Unlock a locked node

A new management user role, the lock admin user, for locking is defined that will have the privilege of locking/unlocking the cluster.

In ECS 3.0 this will be a pre-provisioned local user ‘emcsecurity’.

The lock admin user i.e. ‘emcsecurity’ will have the ability to

  • Modify their password (forced during first login)
  • Lock the cluster
  • Lock a node
  • Unlock a node

System admin/monitor has the privilege to view the lock status of a node but NOT modify it

 

Thoughts and Further Reading

I’ve been talking to a lot of scale-out object storage folks lately. Given the amount of EMC stuff I’ve covered here previously, it’s a little surprising that this is the first time I’ve posted about ECS. That doesn’t mean it’s the first time I’ve looked, and I’ve had customers looking at it fairly seriously. In my opinion, the advanced retention management in 3.0 is really going to put a few customers over the line and finally give them the confidence to throw their Centera grids in the river (figuratively speaking).

The cool thing about ECS, like a lot of these types of solutions, is that you can consume it on your terms, via

  • Appliance;
  • Software defined;
  • Dedicated cloud; and
  • Multi-tenant Storage Cloud.

If you’re feeling keen on ECS, you can take it for a spin here. You can also download a version for free, non-production use here. Grab the datasheet from here.