accelerating application performance and storage savings with … · 2020. 5. 27. · productivity,...

6
www.infinite.io @infiniteio Accelerating Application Performance and Storage Savings with InfiniteIO WHITE PAPER InfiniteIO's metadata-based approach to hybrid cloud data management integrates five technologies that dramatically boost application performance, minimize cloud latency, and reduce storage costs–with zero disruption to IT

Upload: others

Post on 29-Sep-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Accelerating Application Performance and Storage Savings with … · 2020. 5. 27. · productivity, without having to change existing user applications, back-end storage or access

www.infinite.io • @infiniteio

Accelerating Application Performance and Storage Savings with Inf initeIO

WHITE PAPER

InfiniteIO's metadata-based approach to hybrid cloud data management integrates five technologies that dramatically boost application performance, minimize cloud latency, and reduce storage costs–with zero disruption to IT

Page 2: Accelerating Application Performance and Storage Savings with … · 2020. 5. 27. · productivity, without having to change existing user applications, back-end storage or access

1. Accelerating Application Performance and Storage Savings with InfiniteIO | 2

2. ©2020 InfiniteIO, Inc. All rights reserved.

Introduction

The mission success of any organization depends on its applications and data. Analysts forecasting revenue, geneticists

studying DNA sequencing files or engineers compiling source code all require fast access to data. Modern workloads such as

machine learning, DevOps, media and genomics are creating and consuming files at an unimaginable rate. Often, hundreds of

millions or billions of files are utilized, invariably leading to unintended performance and cost constraints.

Modern workflows not only generate massive amounts of data but also require the data to be always available even if never

accessed again. Industry analysts estimate that anywhere from 70 to 80 percent of all data has not been accessed in the

previous 6 to 12 months. Housing infrequently accessed, yet valuable, data is expensive as capacity grows exponentially.

The fundamental challenge for data management–and potential solution–is metadata, as every workflow must first process file

metadata. Metadata is important information about a file, such as where it’s stored and who has access privileges. The business

reality is that all file systems must perform checks on metadata before you can access the associated file. These metadata

requests, though small in size, are absolutely required like airport security, but are a major cause of application latency. Storage

controllers servicing these requests must wait for metadata, much like waiting in the regular airport security line, though it is far

faster and simpler to breeze through TSA Pre-Check.

Enterprise data storage has kept up with the demand for performance through faster media like SSDs or interconnects

like NVMe. However, even fast IT infrastructure must contend with application latency as the on-premises or cloud storage

system grinds through metadata requests. This latency associated with metadata is the last frontier in optimizing application

performance and storage cost.

The InfiniteIO File Metadata Engine (IFME) architecture enables organizations to achieve profound performance for data-

intensive workloads with a simple, scalable platform that eliminates the metadata latency that drags down performance.

InfiniteIO responds to metadata requests directly from the network, before these requests ever get to storage controllers, at

sub-100 microsecond speeds. The apps are no longer waiting for metadata, resulting in dramatically reduced run time and

increased productivity. Metadata intelligence also enables transparent tiering of inactive data from primary file storage to lower-

cost NAS or object storage to dramatically reduce ownership costs, extend IT investments and easily enable hybrid clouds.

Adopting the InfiniteIO architecture is simple and non-disruptive to existing users, workflows and systems. InfiniteIO’s metadata-

first approach, open object storage interoperability and enterprise scale enable organizations to accelerate application

performance, minimize cloud latency and reduce storage costs–without changing existing IT operations, storage or workflows.

Five technologies that turbo-charge applications and storage efficiency

Based on its patent-pending File Metadata Engine architecture, InfiniteIO integrates metadata intelligence and NAS- and

object storage-independent hardware to deliver exceptional application performance, improved productivity and scale while

ensuring continuous access to files. This paper discusses the five must-have InfiniteIO technologies that combine to deliver new

performance levels for data centers and simplify cloud strategies without changing existing IT operations or processes:

1. File metadata engine – Metadata processed independently to increase application performance and reduce cloud latency

2. Metadata analytics – Real-time file metadata performance and attributes are maintained to help optimize the hybrid cloud

3. Hybrid cloud tiering – Cold data dynamically moves to lower-cost NAS or object storage to reduce costs and reclaim capacity

4. Simple deployment – Non-disruptive installation requires no configuration changes to existing applications or infrastructure

5. Enterprise scalability – Platform with high interoperability and support for billions of files to easily accommodate growth

Page 3: Accelerating Application Performance and Storage Savings with … · 2020. 5. 27. · productivity, without having to change existing user applications, back-end storage or access

1. Accelerating Application Performance and Storage Savings with InfiniteIO | 3

2. ©2020 InfiniteIO, Inc. All rights reserved.

For applications like a machine learning workflow, with hundreds of thousands of pictures used for model training, the associated file metadata operations can overwhelm traditional NAS systems, adding latency to application run times and slowing overall workload performance (Figure 2). In retrospect, this is not surprising as NAS systems are designed to serve reads and writes fast. However, as we know, 90 percent of file operations are not reads or writes.

Leveraging the IFME architecture, the InfiniteIO Application Accelerator offloads and responds to all file metadata requests at the speed of memory, which is, at sub-100 microseconds, some 5 to 20 times faster than today’s storage systems (Figure 3). By accelerating 90 percent of file operations, InfiniteIO removes the file metadata latency bottleneck that impedes workload performance. IT teams can deliver faster responses to global data requests, shorter application run time and increased productivity, without having to change existing user applications, back-end storage or access to data. Consequently, the NAS continues to serve file I/O and can dedicate all CPU, memory and I/O resources to reads and writes.

Figure 1. Metadata requests cause application latency; a typical file read generates a metadata offload ratio of 7 metadata requests for 1 read/write

1. File Metadata EngineAll storage systems create and process metadata, and because metadata is foundational to data access, requests for metadata (although small in size) are highly sensitive to latency. To optimize performance and efficiency for the underlying storage or cloud systems without complexity, risk and cost constraints, you need to manage metadata separate from the file system.

InfiniteIO solutions are powered by the InfiniteIO File Metadata Engine (IFME) to speed up application performance, reduce cloud latency and efficiently assign files to the most performance- or cost-optimized storage without disrupting the existing IT environment. The IFME uses deep packet inspection to extract and process file metadata separately in real-time; it keeps the metadata current by actively learning, storing and updating all file metadata in DRAM.

A simple file read isn't so simpleAbout 90 percent of storage system requests are for file metadata (Figure 1). In this example, a typical file read generates seven metadata requests—looking for directories and checking for permissions—before it actually gets to the read. In other words, 7 out of 8 requests are for metadata, representing a metadata offload ratio of 7:1. The higher the ratio the higher the latency caused by metadata, which slows responses for the file. Even the fastest storage systems experience this latency as they all must wait for the metadata. The metadata offload ratio, therefore, represents the opportunity to improve application performance by reducing the latency caused by filesystems responding to metadata requests.

Figure 3. Processing metadata separately from file I/O reduces latency by up to 50X

Figure 2. Metadata is a major contributor to latency; reducing this latency dramatically improves application performance

Read /mnt/homedirs/Bob/test/debug.txt

Page 4: Accelerating Application Performance and Storage Savings with … · 2020. 5. 27. · productivity, without having to change existing user applications, back-end storage or access

1. Accelerating Application Performance and Storage Savings with InfiniteIO | 4

2. ©2020 InfiniteIO, Inc. All rights reserved.

Figure 4. Built-in dashboards show important performance, utilization, ROI and other metrics across the InfiniteIO-powered hybrid cloud

2. Metadata Analytics

Leveraging the File Metadata Engine architecture, InfiniteIO provides real-time insights that automate and improve data management across the

hybrid cloud (Figure 4). The analytics enable IT administrators to take steps to maximize performance and cost savings.

Empowered by this metadata intelligence, you can gain:

• Insight on metadata offload performance on all NFS operations

• Visibility into all files that have been migrated with cloud storage consumption per user for charge-back purposes

• Tools that identify opportunities to optimize capacity and performance at no cost (Figure 5).

By providing IT managers with a complete picture of all file attributes, InfiniteIO not only speeds up applications but also ensures files are automatically assigned to the most logical, high-performance and cost-efficient areas of the hybrid cloud infrastructure.

Figure 5. Infinite Insight identifies cold data in a real customer environment and potential savings after tiering it from primary NAS to cloud storage

Page 5: Accelerating Application Performance and Storage Savings with … · 2020. 5. 27. · productivity, without having to change existing user applications, back-end storage or access

1. Accelerating Application Performance and Storage Savings with InfiniteIO | 5

2. ©2020 InfiniteIO, Inc. All rights reserved.

3. Hybrid cloud tiering Typical data management approaches require IT managers to use various disparate tools to identify and move cold data. These tools can be complicated, time-consuming and costly, especially as data capacities grow. Some require administrators to deploy many virtual machines or leave behind symbolic links or stubs that multiply and add risk. Others might lock users to a specific storage vendor’s data tiering solution, limit cloud storage options, or require configuration changes to clients and applications. Incurring performance penalties on metadata access compound their limitations.

InfiniteIO enables continuous data tiering with native file access through a simple software upgrade to the InfiniteIO Application Accelerator or a standalone product called Hybrid Cloud Tiering. Granular policies quickly identify inactive data and automatically place this infrequently-accessed data on lower-cost NAS or any object storage (public or private cloud). Native file access makes it easier to leverage migrated data for cloud-based applications such as serverless computing, bioinformatics, analytics and machine learning. This open architecture allows customers and partners to not only maximize the value of content but also reduce the cost of data storage, on average, from $0.50/GB of tier 1 NAS capacity to $0.05/GB, depending on the object storage platform. User-defined parameters include file attributes, such as file size, frequency of access and user/group IDs, as well as project-based options (Figure 6).

Figure 7. Transparent and continuous file placement to any lower-cost NAS or cloud storage saves up to 80% in costs

Figure 6. User-defined migration options include file characteristics, frequency of access, user/group IDs and project parameters

For example, an IT manager might set a policy to move all files greater than 10MB that have not been accessed in the last 180 days, or create a

policy for files associated with a specific directory or set of directories. Since the IFME always has a current view of all metadata, tiering can begin

immediately without requiring rescans of storage, special software or stubs. Regardless of the policy, access to tiered files remains unchanged

as they always appear to clients and applications as though they are local. If a client accesses an archived file, InfiniteIO’s Hybrid Cloud Tiering

recalls the file to its original location and the NAS responds back to the client (Figure 7).

IT teams can also take advantage of complimentary software tools that can be installed on Mac or Linux environments to quickly identify inactive or cold files and visualize opportunities to migrate this data to lower-cost storage targets. More information is available at www.infinite.io/ROI.

Page 6: Accelerating Application Performance and Storage Savings with … · 2020. 5. 27. · productivity, without having to change existing user applications, back-end storage or access

1. Accelerating Application Performance and Storage Savings with InfiniteIO | 6

2. ©2020 InfiniteIO, Inc. All rights reserved.

4. Simple deploymentTraditional performance acceleration and data life-cycle solutions often, in the end, disrupt IT and end users. Typical solutions force you to remount clients, install proprietary software or rescan file systems looking for updated metadata.

With an intelligent metadata management approach, InfiniteIO avoids unnecessary downtime and complexity. Customers can install the InfiniteIO products like a network switch. As the installation is live, applications continue to run unaffected throughout the deployment. IT teams can avoid changes to storage and clients as well as rescanning of file systems. They can simply drop InfiniteIO products into the network and can immediately benefit from faster application performance and the ability to identify and transparently tier inactive files off primary storage to lower-cost infrastructure.

5. Enterprise scalability on a single systemInfiniteIO provides a single scalable platform designed to support enterprises as they grow or as application demands change over time. Both the Application Accelerator and Hybrid Cloud Tiering products support industry-standard network interfaces. InfiniteIO solutions are inter-operable with any enterprise NAS platform supporting the NFSv3 protocol and any object storage based on S3 or Swift. In addition, IT managers can implement software updates on demand without disruption. Deployed as an appliance, InfiniteIO solutions can scale to 1 billion files on a single node or 2 billion files in a clustered configuration, while avoiding capacity-based licensing that imposes unnecessary data management fees. Future updates will add increased scale and performance.

Metadata-first hybrid cloud fuels workload performance and cost savings The impact of metadata on modern workloads from machine learning to DevOps and genomics is forcing organizations to embrace new approaches to accelerate performance without incurring holistic changes in their infrastructure. Based on InfiniteIO File Metadata Engine architecture, the InfiniteIO Application Accelerator provides the world’s fastest metadata engine to reduce application latency with minimal to zero disruption to the important projects, workflows and operations. Metadata offloading helps remove latency bottlenecks and reduces application run time by up to 50%, while increasing productivity.

With InfiniteIO Hybrid Cloud Tiering, IT managers can leverage an open approach to object storage usage to cut data management costs down to pennies per gigabyte.

Metadata management is a breakthrough technology enabling research labs, engineering firms, media agencies, software development teams, financial institutions and others to deliver the performance gains and cost efficiency required of modern data centers and hybrid cloud deployments for today’s workloads.

• Watch the 2-minute video at https://youtu.be/mots0pJ_-1I to learn how InfiniteIO accelerates storage performance and cost savings

• Visit www.infinite.io to learn more about InfiniteIO Application Accelerator and Hybrid Cloud Tiering products

• Download the Infinite Insight hybrid cloud storage assessment tool from www.infinite.io/ROI to quickly identify hidden, cold data in your real world environment and visualize opportunities to save up to 80% in storage costs.

Next Steps

InfiniteIO provides the lowest possible latency for file metadata, enabling applications to run faster, reduce development cycles, and

increase data productivity. Based in Austin, Texas, InfiniteIO independently processes file metadata to simultaneously accelerate

application performance, minimize cloud latency and reduce storage costs for global enterprises. Learn more at www.infinite.io, @infiniteio

and LinkedIn.

©2020 InfiniteIO, Inc. All rights reserved. InfiniteIO is a trademark of InfiniteIO, Inc., registered in the United States and other countries.

All other brand names mentioned herein are for identification purposes only and may be the trademarks of their respective holder(s).

11305 Four Points Drive, Bldg 1, Ste 250, Austin TX 78726 • 512.375.4171 • [email protected] • www.infinite.io • @infiniteio