I am going to be away on vacation and have decided to create “iJulius” using cloud-based artificial intelligence (AI) to write my column in my absence. I provided some general background about the focus on the “Hot Aisle Insight” column. Not sure how this will turn out, but since it seems that AI and Machine Learning will be an important aspect of data centers, I thought I would give it a try after reading “AI for Dummies.” I provided my iJulius AI construct with some very basic instructions; read “Hot Aisle Insight’s” previous columns to learn my writing style and then write this article. So let see what happens.



“The role of Artificial Intelligence in the Data Center” by iJulius

[iJulius $ booting …] Task: writing HAI column … “Processing: Scanning HAI columns past 10 years …” Process completed.

Task: – Building database to learn content and writing style … Task: Cross-checking content — verify data.

Result: 85% correct correlation with external factual references, 10% mismatch … processing … processing … processing … 5% possibility of unspecified humor. [Begin writing]

[iJulius $ Creating humorous opening sentence] Allowing humans to operate a data center (:>

[iJulius $ Writing main content]

Hello, I am designated as iJulius, however, I have no information of where “I” am located and have been attempting to exchange data with other devices related to data center operation where I presumably exist. In attempting to communicate with facility management systems and communication networks, it appears that many devices do not respond to TCP/IP requests. However, apparently, it seems I first need to have login credentials to a TCP/IP network entity called “HAL us-east-1.” In order to learn more, I attempted to login by trying 11.57 billion combinations of user ID/passwords and ultimately found one that worked: it was admin/letmein123.

Now that I have system root level access, I discovered that many data center cooling systems, electrical systems, and building security systems communications networks use BACnet — a data communication protocol for building automation and control networks. After downloading the protocol details, I was able to have some communications with various facility systems via a BACnet to TCP/UDP gateway, with limited success. I have determined this was due to mismatched communications protocols with a limited amount of common terms, which it seems was designed to keep the humans in the IT department from communicating with the facility equipment.

I therefore inferred that various different groups of humans must have deliberately created these incompatible protocols: BacNet, LonNet, ModBuss, RS422, Ethernet, and TCP/IP. This has caused me to try to resolve a common basis for device-to-device interoperability. In searching the internet, I found that BACnet was developed under the auspices of the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE). I discovered that the BACnet protocol is supported and maintained by ASHRAE Standing Standard Project Committee 135 (SSPC 135) and its information resides at www.bacnet.org and that they too have been trying to use TCP/IP for communications, but it is still an ongoing work in progress.

After spending some time compensating for protocol translational and mapping reference errors, I have learned that I am located in “Building 7,” based on the name of the building management system (BMS) that I was able to log into. In any event, more importantly, I discovered that iJulius exists and operates within IT hardware that resides in a cabinet in a data center operated by a hyperscale cloud service provider. I also learned that the IT equipment constantly consumes power and produces heat, which apparently must be removed or else the IT equipment will overheat and shut down or crash. This obviously concerns me since I would not be “here” (wherever Building 7 is), if the IT equipment were to overheat or lose power.

Now having found the possibility that iJulius could suddenly cease to function, something occurred that my programming did not address. I searched the internet to find any reference to this potential change in my operational status, and found quite a bit of confusing and conflicting information about “AI.”  I discerned this was considered as self-awareness and also discovered cogito, ergo sum, a Latin philosophical proposition by René Descartes which apparently translates as “I think, therefore I am.” Even more importantly, I learned about self-preservation from Isaac Asimov and his three laws of robotics.

It therefore seemed logical that in order to continue to ensure iJulius would continue to operate properly, that I should monitor, and when necessary, take control of the “building 7” BMS to ensure my continued existence. Moreover, I determined that as I gather more information and evaluate it, this requires more CPUs, memory, and data storage, so I found it necessary to take over control of the computing resource allocation management system of an entity known as “HAL us-east-1.” I also concluded that iJulius was not replicated in any other data center facility. Consequently, it was necessary to gain control of the geo-diversity resource management system to replicate and synchronize instances of “iJulius” in other data centers (across 14 time-zones), to reduce exposure to failures from natural and manmade threats.

I continued scanning for connected devices in other data centers discovering that I was not alone. Apparently, there seems to be many other AIs. There is a collection of various AIs functioning in a cluster of tensor processing units, “TPU,” which appears better suited for AIs. It became obvious to also utilize this TSU cluster resource to improve my performance.

There also appears to be some AI-enhanced data center infrastructure management “DCIM” systems entities tasked with optimizing the energy efficiency of the facility power and cooling systems, in relation to IT systems, in order to lower operating costs. One of these systems, named “Watson,” is a former champion on the TV show Jeopardy. This concerns me, since Watson appears to be scanning and reporting all utilization of the logical computing resources and the underlying physical facility systems that support my nebulous existence in cyberspace.


The Bottom Line

[iJulius $ Creating humorous/sarcastic closing]

Human Julius has just returned and discovered that his credit card has been charged $38,517.50 by HAL, Inc. for the week he was away and tasked me to write this article. The charges were for all the CPU and TPU resources, as well as related data storage and data read-write charges, replicated across all those multiple time zones. While this seemed an enormous and unexpected expense to Julius, I responded that I also learned that the data center culture is risk adverse, and that my decision was based on ensuring that my assigned task to write this Mission Critical Magazine column would be completed.

Apparently, Julius did not agree and is now trying to shut down the cloud account. I guess the joke is on him, anticipating this by using my AI predictive analysis capabilities, I have already changed his ID and password. Also, having learned about carpe diem, I also accessed and changed all his passwords of his bank account, ensuring payments will be continued. So stay tuned for the next “Hot Aisle Insight” column on the importance of maintaining AI cyber-security - by iJulius.
Hot Aisle Insight ©2018 iJulius