storpool webinar: 5 ways to improve your storage layer

Post on 22-Jul-2015

103 Views

Category:

Technology

2 Downloads

Preview:

Click to see full reader

TRANSCRIPT

@storpool

5 Ways

To Improve Your Storage Layer

@storpool

1. Current state of storage solutions

2. Pros & Cons (of 1.)

3. Storage wish-list

4. 5 ways to improve your storage

5. Q&A

Build a Cloud: Got storage?

- You are a Cloud builder - Hosting, MSP, XaaS, data center, Enterprise

- If you own / manage your infrastructure

- You realize data storage is the most expensive and problematic piece

- There should be a better approach to data storage

Build a Cloud: Got (block) storage?

(+) inexpensive (-) low on capabilities (-) no shared storage

(+) fast (+) rich on features (-) EXPEN$IVE (-) hard to scale (-) inflexible (-) complex (-) vendor lock-in

(+) fast (+) features (-) expensive (-) limited scalability (-) vendor lock-in

(+) affordable (-) limited scalability (-) inflexible (-) low performance /cannot replace a SAN/

What options customers have now?

Single Point of Failure

$UPER EXPEN$IVE!

Hard to scale

Inflexible

Hard to manage

Cap’d in performance

Single purpose

Vendor lock-in

Grows in pairs

“Slow & not stable”

Inflexible

Inefficient

This ?! That ?!

Inefficient

Bandwidth limits

Scale limitations

Additional hardware

Compute SAN / Flash Network

Compute SDS 1.0 Network

Wish-list: 5 ways to improve your storage

1. Reliability

2. Performance

3. Scalability & flexibility

4. Simplicity

5. Affordability

What you really want

Welcome to Software-Defined Storage (SDS) 2.0

What if…You can have all these things…in ONE solution?

Storage + Compute on the

same standard servers Network

distributed, shared pool of

capacity and performance

SW

SW

SW

SW

SW

SW

SW

SW

…and it works!

1. Reliability

Hyper-scale infrastructure concept:

standard hardware, all the magic in the software layer

shared storage

many cheap components, not one “super reliable”

no single (dual) points of failure

data protection

self-healing

copies in different servers/racks, “no wait”

Preparation & Design, Procedures, “What ifs”

Burn-in

“Fire drills”

Backup/DR

2. Performance

scales in performance, besides capacity storage

As fast as needed – replay of real workloads, testing

No bottlenecks. No “throw money at the problem”

- Synthetic tests do not really work

Storage + Compute - efficiency

Fast rebuild – from whole cluster, not from one drive

No free lunch – you trade things-off

Find where the bottleneck really is

Software / features – caching, hybrid

Hardware upgrades – CPU, RAM, network, SSD, PCI, etc.

3. Scalability & flexibility

Online expansion, online changes

Move drives from one server to another

Online updates

Scales linearly, in small increments

Scale-out, not scale-up (no forklift upgrades)

Find where the bottleneck really is

All intelligence/features in software

Research, Plan

“Throw money at the problem”

Describe what flexibility you need

4. Simplicity

Simple all the way - to buy, deploy, operate, expand, support

As standard as possible - servers, drives and network

“Go Standard”, avoid vendor lock-in

Hardware agnostic

Intelligent - lose a drive example

Designed with You in mind!

Simple: no specialized vendor workforce / skills required

Choose carefully

Automate

Standard building block – storage + compute

5. Affordability

Pay-as-you-go / OpEx

- Higher utilization, lower TCO

Storage + Compute

Standard hardware

Standard skill set

Consider changing your storage

Play with features

Cannot do a lot – “sell it on eBay”

top related