Micron 9100 MAX 2.4TB NVMe Enterprise SSD Review - The New Performance Champion

 

 

ADVERTISEMENT

Review:
Micron 9100 MAX 2.4TB

Reviewed
by:
J.Reynolds

ADVERTISEMENT

Provided
by: Micron

Firmware
version:
0091

 

ADVERTISEMENT

 

Introduction

Welcome to Myce’s review of the Micron 9100 MAX 2.4TB NVMe Enterprise
SSD.

The Micron 9100 is available in two classes: 9100 PRO (for
Read Centric Uses) and 9100 MAX (for Mixed Uses).

We test the high performance 9100 MAX 2.4TB and as our
subtitle suggests it proves to be a scorcher – so please read on to see how the
Micron 9100 sets new high water marks for Enterprise performance.

Revised Reviews

In this review I begin the process of refreshing/revising my
standard approach to presenting test results.

For example –

In the Myce/OakGate 4K Latency Read and Write Latency Tests,
I have added a much closer look at the Quality of Service delivered by a
drive.  This allows readers to compare not just how fast a drive performs but
also how consistently a drive delivers its performance at a specific level of
IOPS.  I feel Quality of Service will increasingly be seen as a key competitive
differentiator.

In the Myce/OakGate Reads and Writes Tests, I am focusing on
presenting the results for Random Reads and Random Writes in the context of IOPS
and I am presenting the results for Sequential Reads and Writes in MB/sec.

In the Myce/Oakgate 4K Mixed Reads/Writes Test, I am also
focusing on presenting the results in the context of IOPS.

The objective is to make the presentation of results more
concise, interesting, and easier to consume (rather than to overload readers
with data). 

I was thinking of dropping the Myce/Oakgate Entropy Test,
which examines how a drive performs with varying levels of compressible and
random data, as seemingly in recent years no SSD controllers compress data (now
that the once ubiquitous SandForce controllers are no longer seen). However,
our standard Entropy Test threw up a strange anomaly in our testing of the
Micron 9100 – please see page 10 to find out what happened.
  

In the Power Consumption Tests, I have dropped the
presentation of ‘Power up to Idle’ and ‘Idle’ power consumption tests as they
are not really relevant to Enterprise drives.

I hope readers approve.  

 


Market Positioning and Specification

Market Positioning

This is how Micron positions the 9100 –

Form
Factor

The Micron
9100 is available in two form factors -

 

 

Specification

Here is Micron's specification for its 9100 range of models
-

 

Here are some pictures of the Micron 9100 2.4TB MAX that I
tested -

The Micron 9100 uses Micron’s 16nm MLC NAND (the same NAND
that we saw used very effectively in the Seagate 1200.2).

The Micron 9100 MAX 2.4TB packages a total of 4TB of NAND
and sets aside 1.6TB as a fixed overprovision. I understand that the 9100 PRO
3.2TB also packages a total of 4TB of NAND but has only 0.8TB as an
overprovision.

Micron states that the 9100 PRO has an endurance level of
about 1 DWPD and the 9100 MAX has an endurance level of about 3 DWPD. The 9100
is warranted for 5 years, although endurance is warranted on a total bytes
written basis; for the Micron 911 MAX 2.4TB this is 9.6 PBW (Petabytes Written)
for sequential writes and 6.57 PBW for random writes.

The 9100 includes RAIN Technology, advanced wear levelling
and a Power Loss Subsystem, which Micron introduces as follows:

RAIN Technology

Redundant array of independent NAND (RAIN) is a technology
designed to extend the lifespan of the SSD. RAIN is similar to redundant array
of independent disks (RAID) technology, but instead of grouping and striping
disks, RAIN groups and stripes storage elements on the drive across multiple
channels, generating and storing parity data along with user data. This data
structure (user data plus parity) enables complete, transparent data recovery
if a single storage element (NAND page, block, or die) fails. If a failure
occurs, the SSD automatically detects it and transparently rebuilds the data.
During this RAIN rebuild process, the drive's performance is reduced
temporarily but will recover after the rebuild process completes.

Wear Levelling

The device uses sophisticated wear-levelling algorithms to
maximize endurance by distributing PROGRAM/ERASE cycles uniformly across all
blocks in the array. Both static and dynamic wear levelling are utilized to
optimize the drive’s lifespan. Both types of wear levelling aim to distribute
“hot” data away from blocks that have experienced relatively heavy wear. Static
wear levelling accomplishes this by moving data that has not been modified for
an extended period of time out of blocks which have seen few P/E cycles and
into more heavily worn blocks. This frees up fresher blocks for new data while
reducing expected wear on tired blocks. Dynamic wear levelling, by contrast,
acts on in-flight data to ensure it is preferentially written to the least-worn
free blocks rather than those closer to the end of their rated life. These
techniques are used together within the controller to optimally balance the
wear profile of the NAND array.

Power Loss Subsystem and Rebuild

The SSD supports an unexpected power loss with a power-backed
write cache. No user data is lost during an unexpected power loss. When power
is subsequently restored, the SSD returns to a ready state within a maximum of
120 seconds.


Now let's head to the next page, to look at Myce’s
Enterprise Testing Methodology.....

 

No posts to display