The Terminator Skynet…it’s here! – Artificial Intelligence in the Data Centre

We’ve all been there. Just updated some drivers, applied some firmware, installed some piece of software and BANG…issues! How long it takes to resolve will depend on the issue, but either way, it takes up precious time.

What is the first thing most of us do when we hit an issue? Open up Google, find someone else who has hit the same issue, then hope that the resolution fixes your problem.

Now imagine Morgan Freeman reading the next few sentences like he did at the end of Shawshank Redemption:

What if…

…you didn’t actually need to hit the issue in the first place.

…when just one person has hit this issue, no one else would.

…when your hardware has failed or is about to fail, you automatically get sent a replacement.

…there was a product out there that could do all of this for free.

One of the current buzzwords around the IT industry is Artificial Intelligence (AI) and while it sounds cool and makes us all think of the Terminator movies, what does this actually mean?

Has Skynet finally arrived to wipe out all human life? No, not yet, but there is one product out there which is trying to make some of your hopes become a reality…HPE InfoSight.

HPE InfoSight – what is it?

Over a decade ago, Hewlett Packard Enterprise (HPE) started designing systems with sensors across the infrastructure stack and HPE InfoSight is now collecting trillions of these data points. This data has now grown into multiple petabytes of information and provides visibility into your Infrastructure.

While it all sounds very impressive collecting and storing this amount of data, it really doesn’t mean much if you can’t use it. This is where the AI kicks in.

The AI will track this data (along with input from the support engineers) so when client XYZ down the road:

• has the same hardware,
• connected with this firmware,
• and has run across this particular problem,

The AI will kick in and block clients from installing the offending software to begin with. If the same issue has already been identified, an alert with the fix will be sent to the client.

The future vision will be to expand this database to cover most, if not all HPE hardware, providing even more in-depth knowledge across the entire data centre. This future vision will also include the “Skynet” clause where InfoSight can (or will) implement automatic fixes so there is no human interaction required at all.

The sensor and telemetry data is sent from the client’s site to the InfoSight database via an outbound only secure tunnel. It does not contain personally identifiable information or any user data stored on the infrastructure…well at least if the AI ever did become self-aware, it thankfully won’t be able to track you down. Once the AI has processed the data, it is then presented back to the client via a secure, easy to read, web portal.

Real world application

HPE InfoSight has full stack visibility. It is unique in that its analytics look across the entire environment. This will cut down on the finger pointing scenarios you can sometimes get into when dealing with multiple components within a data centre. This visibility covers storage, network, host CPU/memory resources, all the way up to the operating system/applications. This can be especially useful when dealing with VMware performance issues.

I can’t tell you how many times I’ve heard users complain about the “system” being slow. How nice would it be to have something to look at everything involved in the system (i.e. full stack). It’s those “needle in a hay stack” type issues, which all administrators dread, but hopefully will have an answer to now.

Examples of how the AI works in the real world

When InfoSight predicted and prevented a catastrophic “All-Paths-Down” situation for storage customers due to a potential issue with a network card in the host. Leveraging InfoSight, support engineers determined that the fibre channel recovery mechanism might fail due to a double abort issue within the card. HPE InfoSight used a signature pattern-matching algorithm to identify 100 customers susceptible to this issue and applied a workaround that prevented the issue.

Another example is when a storage system began to run out of memory, disrupting the file services. Thanks to all the data being collected, engineers were able to pinpoint the issue to an exact time, which was then pasted to the development team. Within the same day, development created a patch and sent it out to the client. Once the fix was verified by support, the HPE InfoSight AI searched its database for clients with the same configuration and alerted them all. Support applied the fix to these sites, which resulted in zero downtime.

Visibility of all these levels has provided Nimble Storage with these impressive support numbers:

• – all level 1/2 support calls are dealt with automatically through HPE InfoSight, this accounts for 86% of all calls raised
• – only level 3 support engineers exist for Nimble storage
• – 46% of these support issues are actually storage related, the other 54% of issues are found and fixed elsewhere within the environment.

HPE InfoSight history

In the beginning, Nimble Storage created The Pachinko Machine, a simple database engine based on rules for auto support. Auto support tasks included simple RMA activities such as when a disk drive failed, support would automatically send a new replacement disk.

This engine was the basis for InfoSight. All arrays sold from then on, would use this engine going forward. The AI was born and began processing this data internally for their support engineers. It was soon after they realised how powerful this tool was so they made it available to the end user for root cause analysis. From there, any other clients that matched the same criteria would either be:

• notified of the issue,
• have the issue automatically fixed before they were aware of it,
• denied the ability to install the offending software.

In 2017, HPE acquired Nimble Storage. One of the key benefits of HPE acquiring Nimble Storage was InfoSight. HPE InfoSight is now known throughout the storage industry as one of the smartest interfaces out there due to its AI backend.

HPE have embraced InfoSight and have now added their 3PAR storage, ProLiant, Apollo and Synergy servers into the mix to provide end users one of the best support tools available on the market today.

Who can use HPE InfoSight?

One of the best things about HPE InfoSight is that is it free to use with supported products (certain features are only available to end users with a current HPE maintenance agreement).

The products that currently support HPE InfoSight:

1. HPE Nimble Storage Arrays

All Nimble arrays

2. HPE 3PAR Storage Arrays

For global visibility, you have to have a minimum of 3PAR OS version 2.2.x and Service Processor (SP) version 2.4.2. I’ll have a bit more about global visibility in a second.

For the cross-stack analytics, there’s a bit more you need. Today, you have to have 3PAR OS version 3.3.1 and 3PAR SP version of 5.0.3 (aka 5.0 MU3). You must also enable the RDA transport.

3. HPE Servers

The following servers with iLO 4 and iLO 5 are supported:

• HPE ProLiant Gen8 and later servers
• HPE ProLiant BladeSystem Gen8 and later servers
• HPE Synergy Gen9 or later compute modules
• HPE Apollo Gen8 and later servers

iLO Firmware

• iLO 4 firmware 2.54 or later
• iLO 5 firmware 1.15 or later

NOTE: HPE recommends iLO 5 firmware 1.37

• iLO Amplifier Pack 1.30 or later

How can Data^#3 help?

If you have any of the supported HPE hardware, you can take advantage of HPE InfoSight. Whether you require support with setup, configuration, firmware upgrades or would just like to know how HPE InfoSight can help your business, reach out to your local Data^#3 account representative or contact us today.