Hello everyone.

We have began the first step of exploring around, in order to find a topic for our thesis. Although I have to say I am pretty indecisive, my preferred field would be Information Retrieval, Network Science, Knowledge Extraction. I'm struggling with picking where to apply these exactly and I was wondering if anyone has any advice? I have a deadline soon to prepare the literature review and it's being stressful already.

Thank you in advance

 

Hello,

Are you mainly interested in software or hardware? Is the topic related to high level business aspects or more operational levels such as disaster recovery, failover or maintenance? I may have some ideas, depending on your input.

2 minutes ago, Ghideon said:

Hello,

Are you mainly interested in software or hardware? Is the topic related to high level business aspects or more operational levels such as disaster recovery, failover or maintenance? I may have some ideas, depending on your input.

I'd say software. As for the second question, I don't quite have any preference so any ideas are welcomed!  

Ok! Let's try some reasoning:

1: Network Science: for instance a system with geo redundancy.
2: Information Retrieval: gather information from various components of the system above
3: Knowledge Extraction: extract "knowledge" from the "information" retrieved above.

Combining 1, 2 and 3: 

Lets say some system, or set of systems, are mission critical and have automatic failover and real time replication of data between several sites. The systems have some means of sending logs in real time to a central destination. At this central location the information is analysed to try to extract knowledge. Now we have a setup where several possible thesis topics can apply. Speculative examples and ideas:
- What software is currently available? 
- What algorithms are required in this setup? 
- How can for instance machine learning or AI be applied?

Example; trying to include 1-3 and also both operations and business: Low level logs from one or more sites states that response times increase. How long will it take before customers will be affected and maybe head elsewhere? Can possible problems at a low level be automatically transformed into an estimate of loss of revenue? Can maintenance be prioritised immediately from a loss of revenue perspective rather than technical severity? 

One or more of these aspects may give some ideas about thesis topic.
Note: Of course there's software that can, or at least is claiming, to be able to do this. But as far as I know it is not necessarily a simple task to actually do it in reality.  

1 hour ago, inuski said:

I'd say software. As for the second question, I don't quite have any preference so any ideas are welcomed!  

Hard disk data recovery, when they were not physically damaged, is probably the easiest thing you can do. Start from learning file-systems used in Windows and Unix systems. How to raw read/write data from/to hard-disks. Actually you can end up making your own file recovery software which will bring you millions of dollars at the end..

11 hours ago, Ghideon said:

Ok! Let's try some reasoning:

1: Network Science: for instance a system with geo redundancy.
2: Information Retrieval: gather information from various components of the system above
3: Knowledge Extraction: extract "knowledge" from the "information" retrieved above.

Combining 1, 2 and 3: 

Lets say some system, or set of systems, are mission critical and have automatic failover and real time replication of data between several sites. The systems have some means of sending logs in real time to a central destination. At this central location the information is analysed to try to extract knowledge. Now we have a setup where several possible thesis topics can apply. Speculative examples and ideas:
- What software is currently available? 
- What algorithms are required in this setup? 
- How can for instance machine learning or AI be applied?

Example; trying to include 1-3 and also both operations and business: Low level logs from one or more sites states that response times increase. How long will it take before customers will be affected and maybe head elsewhere? Can possible problems at a low level be automatically transformed into an estimate of loss of revenue? Can maintenance be prioritised immediately from a loss of revenue perspective rather than technical severity? 

One or more of these aspects may give some ideas about thesis topic.
Note: Of course there's software that can, or at least is claiming, to be able to do this. But as far as I know it is not necessarily a simple task to actually do it in reality.  

I wouldn't want a combination of all three though as that really sounds like too much work to get into. I was hoping I did something simple as some algorithm benchmark, or some type of prediction using text mining. I'm  not sure if I make sense as I'm still in my first year (I have other courses to attend to be fully ready for the thesis) but we still have to pick a topic and work pretend on our thesis for the Research Methods course. I'll fully work on it in 2020.

10 hours ago, Sensei said:

Hard disk data recovery, when they were not physically damaged, is probably the easiest thing you can do. Start from learning file-systems used in Windows and Unix systems. How to raw read/write data from/to hard-disks. Actually you can end up making your own file recovery software which will bring you millions of dollars at the end..

That sounds good but I'm not quite confident with my skills 

 

On 2018-11-15 at 11:52 AM, inuski said:

something simple as some algorithm benchmark

 

On 2018-11-15 at 11:52 AM, inuski said:

That sounds good but I'm not quite confident with my skills 

New idea based on your input so far; combining a variant of the suggestion from @Sensei with your interest in algorithms. Are you familiar with the various versions of RAID* ? You could look into questions like "What algorithms are used? How do systems recover from disk failures in different RAID levels?" 

*) https://en.wikipedia.org/wiki/Standard_RAID_levels