ParaScale Cloud Storage Customer overview presentation

Changing The Economics of Bulk File & Content Storage

March 2009

2 Confidential

Powering Cloud Storage

EnterprisesOwn datacenter, own staff

Choose own hardware, re-purpose servers

Start small

MissionDelivering the software platform to address the exploding

requirements to store digital content and archival data

PrivateCloud

Service ProvidersDifferentiated services

Separate clouds – archives vs. video etc.

Large multi-tenant clouds & small hosted clouds

Public Cloud

3 Confidential

ParaScale: commodity servers, standard protocols

Software-only on standard hardware– Standard clients and protocols

– No agents

– Commodity hardware and networks

– Single file performance limited to performance of single node

Benefits– Transparent data migration

– Massive scalability, start small

– Economical

– Your choice of hardware

4 Confidential

Scalable File Storage Technology Landscape

Tightly Coupled / Custom OS Loosely Coupled / Standard OS

High cost

PanasasLustre (Sun)

Custom Clients, Protocols

and Network

High PerformanceComputing

High cost

Standard Clients and Protocols

Custom Network

Clustered NAS

IsilonNetApp GX

Low cost

Standard Clients, Protocols

and Network

Private CloudStorage

GoogleAmazon S3

Powered

Monthly cost

Custom clientsand protocols

Standard Network

Public CloudStorage

5 Confidential

ParaScale Technology

Asymmetric architecture– Massive parallel scaling and

performance

Enterprise features– Load Sharing

– Capacity balancing

– Thin Provisioning

– Replication

Many installations to date– Customers in ongoing trials

without downtime since 2006

– Deep team, 66 patents in storage and clustering

6 Confidential

Two Types of Nodes

Control Nodes– Metadata servers – Max of 2 per cloud

– Dedicated server(s) that maintains the configuration information for your storage cloud

– Orchestrates the replication and migration of files across the cloud

– Provides the central management of your Storage Nodes

– Data does not pass through the control node

Storage Nodes– Data servers – Max of 100 per cloud (this release)

– Dedicated servers that store and deliver data to clients

– Massive amount of flexibility in hardware choices

7 Confidential

Tune the Cloud for Performance or Capacity

Mid-tier nodes with faster disk

Massive I/O Requirements

Inexpensive nodes with lots of disk

Minimal I/O RequirementsCapacity

Scale Out

Performance Scale Out

8 Confidential

Scale Performance and Capacity

Incremental additions to bandwidth and storage– Scale the cloud based on individual

requirements

– No need to compromise on vendor-selected node options

Performance Scaling– Additional nodes provide linear

aggregate throughput• Modest configurations are capable of

70MB/s writes and 100MB/s reads per node

Capacity Scaling– Storage from the node is

automatically added to the cloud

Each node is independent and serves files directly to clients

Cloud Bandwidth

Physical Capacity

Attached disk is pooled

9 Confidential

Simplified Installation and Management

Download and GO– An application that runs in user space

• No custom kernels

• No specialty hardware

• No overpriced appliances

Self healing and configuration– Automatic node discovery

– Continued access upon node failures

– One click retirement of nodes

Exception only Management– No day-to-day configuration or

administration tasks

– Detailed monitoring and reporting

– A single administrator can manage 100’s of nodes

10 Confidential

Policy Engine for Data Redundancy

Automated system administration

– Simplifies management by synchronizing and administering hundreds of volumes and nodes

– Policy based administration of file replication for data availability

• Files are copied to one or more nodes for data protection or improved performance

• Detects failed or evacuated nodes and ensures new replacement copies are created

• Removes the need for expensive SAN data protection schemes

Files are automatically replicated

11 Confidential

Capacity Balancing and Incremental Scaling

System automatically balances capacity across nodes– Files are moved between nodes to address

utilization inefficiencies

– Supports nodes with different amounts of storage or different size disk drives

– Automatically balances when new nodes are added to the cloud

Add or remove storage as needed– Server vendor agnostic– Physical storage can be expanded on

demand• Via another node or added to an existing node

– One click decommission of nodes– Delay purchases to leverage increasing

capacity and reduced costs

12 Confidential

Global Namespace with Thin Provisioning

Consolidated view of network file systems– Each node presents the entire namespace

– Any user or application can access any file from any node

• Contingent to ACLs and file R/W locking

– Removes complexities of localized file administration

Present virtual file system sizes– Presentation not based on physical disk– Can be expanded or contracted without

disruption– Prevents asset over-allocation and

storage silos

13 Confidential

Load Balancing

Client redirection for load sharing– NFS mount redirection

• ParaScale intellectual property

– Clients redirected to nodes based on data locality and current conditions

– Data does not pass through the control node

Hot file system scaling– Policy based replication of files to

multiple nodes– Each node services client

requests independently

14 Confidential

Background

Team– Deep clustering, storage expertise (66 patents amongst the team)

– Team from Veritas, HP, Sun, Locus, Teradata, NetApp

History– Founded 2004,

– Alphas running software since early 2006

– Beta since mid-September 2008

Investors / Funding– Menlo Ventures, Charles River Ventures

• Successful track records within clustered storage (Spinnaker / EqualLogic)

– $11.37 Million to date, Series A - May 2008

For more information:

www.parascale.com

Changing the Economics of Bulk File and Content Storage