IBM Cloud Foundry

Visual recognition. We put together a number of very effective demonstrations, over a short period (7-10 days) to show our management team as well as customers how it would be done, how the Bluemix applications could be integrated with other services (eg. Amazon Web Services, and Microsoft's OneDrive) to create one very capable, automated, integrated application that could solve one of our customer's knotty problems.

The visual recognition application was the only thing we tried during this 7-10 days period. Significant issues with technical and account support (there was absolutely none in any time of what would be considered a 'timely' manner, led us to realize that it was not a service we could use for production, and we therefore did not continue to try out other services.

Although we didn't get to the point of using several services, the services for storage and visual recognition, as well as those for integrating with other cloud provider services (e.g., AWS) were smooth and effortless.

ALMOST EVERYTHING. Account support was virtually non-existent. We had set up a trial account on the basis of how many training images we could use (they advertised that it was something like 100 per day--but it turned out that the actual number was for the entire trial period. Thus, we almost immediately ran into our limit and could not proceed. We set up a second trial account, but couldn't get that one to work at all (with no meaningful error messages to why it wasn't working). Finally in desperation, we set up a third account tied to a personal credit card. (It was the weekend and we had to have a demonstration ready for Monday afternoon.) That third account, although setup, also would not work. (Some of the problems could have been the result of not having three separate, easily available emails, phone numbers, and billing addresses to use in setting up the accounts, which certainly exacerbated the issues.) On Monday morning we were finally able to reach someone that helped get the 'billed' account turned up--but even they admitted they could not figure out why it had not worked on Sunday.

Over the period of about three days we made multiple attempts to reach both technical and account support. Generally the wait was something close to 24 hours, far, far beyond what we would be able to use in a production environment.

Technical support was generally quite a bit more helpful that the account support team. The technical folks were able to get 'stuck' and non-working account working again, but the account support team was completely unable to provide any billing information. This included how much had been billed to the 'personal' credit card (the one we had to use because it was setup during the weekend when our own internal finance folks couldn't provide a company credit card), and they were completely unable to provide an actual statement, either then or for months later.

One thing of particular note--the 'on-line help' feature for account services was extremely disappointing. There was absolutely no way to get any kind of billing or accounting information using the on-line services. You would think that you could, but you simply can not!

We were finally able to get an accounting almost 6 months later and ONLY because it had gone to some form of 'collection' department within IBM and they were trying to get the final $24 paid. (The biggest part of the amount due from the demonstration period, which was about 7-10 days and had amounted to about $350 in charges, at least 10 times what we had expected) had been automatically charged to the 'personal' credit card we used in desperation to get an account set up so we could get a very important demonstration for our out-of-town customers who were expecting a demonstration the following Monday afternoon. However, the billing fell across two accounting months, with only about $24 in the second month, but during that account, the bank had changed their association with MasterCard to Visa and had issued new cards and rejected any automatic billing to the old card. This left a balance of $24 to be paid, which we had no clue of. We noticed that we had only received the one bill, but not the second, but again, calls to IBM were either not returned or the people we reached told us they could not find out how much was still owed or how we could pay the balance by phone. Finally, almost 6 months after our demo, we received an email from their 'collections' department. We told them that we were not going to provide every 1 cent of funds UNTIL they provided us with an accounting of the services used, etc. That took a few days and we were finally able to get the $24 resolved.

We explained most of this to at least five different technical/account support people along the way. At NO time, did ANY IBM representative ever offer ANY kind of 'consideration' for the trouble we had had, even after learning that the company was unable to reimburse us for 'personal' expenses BECAUSE we could not receive a standard billing statement.

Flexible development environments available, all interoperative, from Docker-based to apiconnect-based. We can use several repo-sites and keep code versions well tracked and reclaimable on any of them. The networked nature of the systems means we can develop from a world wide basis of engineers and programmers, although right now we have one Senior Software Engineer and a couple of coders, in different countries.

Datasources can be connected from anywhere.

Mobile Endpoint Security, and Server Security (meeting or exceeding 27001 and 27002) with IBM, represent resellable value to us.

We are a fledgling company, but as soon as we are able to afford to use the Blockchains offered by IBM, we will do so, because we can eliminate one entire class of financial (or any trust/transaction-based) risk this way.

With the use of Cordova we can code our front ends once and cover the web, Android and iOS platforms together with minimal fuss to tailor the code.

Sometimes the API Connect GUIs don't cleanly disengage after attaching models or updating schema and it is hard to know what has been written successfully and which (if any) models or tables were missed. I shouldn't have to manually check through a list of 377 models to find the ones in and out of a list on either models, folder or database tables. Printing a summary even in logs which did a "diff" sort of thing between 'task-set' and 'task-completed' (referring to attaching models or updating schema as tasks here as 'tasks').

Provide access to Postgres Database in Sydney datacentre for Australia.

Clearer documentation around setting up a secure (referring to SSL and certificate setup here) server on eg, chubby1.au-sydney.mybluemix.net.

Allow a ramp in pricing onto the Blockchains. We will not be able to afford it until quite a few years into production, even if we launch successfully.

PaaS

Watson (did not use in prd, just saw good demos)

Bare metal servers

At the time we used there was no direct Docker offering (had to use containers via CloudFoundry api, which is another layer to learn)

No hosted Cassandra database offering (or similar DB, like Amazon Dynamo)

I was unable to use VPN link with another provider: Both sides had incompatible configurations and it was impossible to instantiate working VPN connection. Support was only able to point that the other party uses settings that are incompatible with IBM.

Easy to use

Simple UX

Good design

Expensive

Makes a number of integrated modules available seamlessly (e.g., mobile, database, authentication, etc.

The node js platform for hosting IoT components is extremely easy to use.

The node-red graphical wiring tool is makes it especially easy to integrate 3rd party services.

Excellent customer support.

More tutorials/education materials for Cloudant would be nice.

More flexible notifications for outages would be helpful.

More integrations to third party services within node-red would be useful.

Speed - with a few regions we can always load balance for maximum performance.

Availabilty - anywhere, anytime.

Usability - everything in one place. Usage metrics and services.

Maybe a better user management interface. Every region and all users in one view.

Bluemix is great at providing a custom environment that offers everything I need to develop and run my software, rather than just providing a blank virtual server that I need to configure

Bluemix is very good at getting a "ready to use" environment deployed very quickly

Offers a wide variety of very unique, and cutting edge services. For example, I recently came across a Bluemix service that analyses any documents I want, such as email, and can tell me what emotions I am conveying. Is it a hateful sounding email, or a happy sounding one.

Documentation is lacking unfortunately, and while the new version of BlueMix is very quick and easy to use, it still requires a heavy development background to get features out of it that would end up rivaling AWS.

Stability isn't where AWS is, and while being a perfect development system, it isn't where you want to run a production site where downtime causes financial losses.

It is very hard to move your app off of BlueMix to another cloud service, as you just can't get the same level of platform customization, and you are going to absolutely be stuck to other PAAS providers.