"Building on Google Cloud Platform is a big investment for the future"
GameWith, Inc. runs Japan’s largest gaming app advice and information site “GameWith”, which made its entrance on the Mothers section of the Tokyo Stock Exchange in 2017. Now gaining attention as one of the fastest growing companies around and not content to remain in Japan, GameWith is expanding overseas focusing on English speaking countries and Taiwan. Not only that, the company is responding to the challenges that drive the gaming industry, such as blockchain technology games and the e-sport teams market.
GameWith also released the WEB Manga Service for smartphones, “MangaWith”, in December 2018. It offers over 15,000 e-comics tagged to large publishing houses and is supported by Google Cloud Platform.
We asked those involved in the launch, Tomoyoshi Murata of the President’s Office, Hiroshi Noguchi, MangaWith business manager, and Ko Taguchi of the Services Development Division, to talk about introducing GCP and working with Cloud Ace.
Hiroshi: We are developing and expanding two web media services, “GameWith” and “MangaWith” on AWS and GCP respectively.
We decided to use GCP when launching our new 2018 service because we had gained enough internal knowhow on AWS with “GameWith”. We could now familiarize ourselves with GCP in addition to AWS, and began building MangaWith on it from scratch.
Ko: We needed to choose a computing service adapted to the services we were developing in-house. If we just continued with AWS because that’s all we knew how to use, it wouldn’t be good for us as a company. Our aim this time was to challenge ourselves by opting for GCP.
Tomoyoshi: We are now adding new functions to the services we provide. We are hedging our bets by widening our options as a company, and “MangaWith” is the litmus test.
MangaWith’s GCP configuration
Hiroshi: We launched “MangaWith” in December 2018 and began development in June of the same year. The worry was that there was only 6 months to develop it, but our experience with AWS helped us to set an approximate completion date. Still, it was difficult to estimate and we were confused in the beginning.
Tomoyoshi：We needed 2.5 engineer man-months for development. I also worked on the business planning side. So in total, we tackled the project in 4.5 man-months.
Hiroshi: We wondered whether to use GAE or GKE after deciding on GCP for the architecture. One characteristic of the Manga Service is that peak times for access are at night or during morning commutes, therefore we chose GKE after considering the concentrated speed of scaling.
Also, we have to prepare an environment where there is a a lot of external cooperation and collecting of externally encrypted source codes because the manga ebooks need to be encrypted. One of GKE’s strengths is creating a loosely coupled system for MangaWith to receive external source codes.
Tomoyoshi: We need to cooperate with intermediaries, not only the publishing houses, to gather encrypted data, and we also need to collaborate with companies who provide encryption solutions when sending the manga. Though we didn’t think about it at the time, GKE’s capabilities are a great help with this.
Hiroshi: That was a good point about it because there’s a chance of similar situations in the future, but it was quite hard to actually realize (smiles). We didn’t have much reference material, so we searched for sources on GitHub and built it by going through ones that looked like they matched one after the other.
Also, as many parts needed a static IP for service quality, we made an effort to take the scale into account while building. We divided things with GKE into two - things that needed a static IP and things that didn’t, and because the scale of parts that needed a static IP was fixed, we created a service that wouldn’t crash even when it encountered a fault or error. There is an upper limit to the parts which don’t, but they are relatively easy to scale.
Ko: We use GCS solution service to store the cover images and manga images. Also, we use about 10 GCP product services including Cloud Memorystore, BigQuery, CloudBuild, etc.
Effect of GCP’s introduction
Hiroshi: When we initially released “MangaWith”, there was an instant spike in hits due to publicity from press releases and Youtuber videos. Even so, GCP’s scaling and stability of operation was fantastic.
Also, as I mentioned earlier, we link to a lot of external servers, so one great benefit is being able to develop the site while on a loosely coupled system. Another good thing is that we can do coding quickly and easily because GCP services basically gain access using API.
Ko: Although we prepared numerous test environments, until now we couldn’t build source codes freely in those environments.
Hiroshi: We can quickly deploy our preferred branch in the development environment. It’s also very beneficial that the functions provided by API or SDK make resources easy to operate and control.
Tomoyoshi: Similar to Hiroshi and Ko, I also had concerns that people would not know how to handle the development schedule or development workload with GCP, but I think you can make swift progress if you build up knowhow. With that in mind, I think we accomplished our initial aim to accumulate new knowledge and turn new product ideas into reality. It has been a nice challenge.
Rapid support from Cloud Ace to build GCP
Hiroshi: We chose Cloud Ace because we’d heard from an infrastructure engineer that they had a fast response rate. We considered one other company for the introduction phase, but Cloud Ace responded incredibly quickly when we contacted them. This speed of response was a huge factor for us due to our extremely tight development schedule. Also, they set up a development chatroom and gave prompt and thorough answers to our questions in just a few minutes or hours. This gave us a good sense of their reliability and technical capabilities.
Cloud Ace’s server-side engineers and front-end engineers, as well as their infrastructure engineers, supported us with AI development, staging and building a service environment, in addition to the security side of the Manga service - all within a very tight time-frame. It was particularly reassuring to have them build our staging infrastructure. They also supported me personally with development on Kubernetes.
It was, of course, also very helpful that we could consult with Cloud Ace in detail about actual systems development in addition to use of GCP. They promptly raised the upper limit value when we consulted them about, for example, a problem which often occurs with Manifest files, which is “What to do when an error page appears the moment they are deployed?” In any case, we were very grateful for their support.
How are you going to use Google Cloud Platform from now on?
Hiroshi: We want to improve the CD part. The automation is going well but I’d like to install Spinnaker so as to reflect a more stable service. Also, we couldn’t use GCP Cloud NAT as it was released only a short while before we released MangaWith, therefore I’d like to try it out.
At the moment, we are only using the basic package, which gives us about 50% functionality. I’d like to try various other features in the near future.
Ko: I’m currently using firebase to make it easier to develop “MangaWith”, but due to IP address restrictions I had to employ various workarounds, so I’d like that to be addressed.
Tomoyoshi: “MangaWith” was born from a desire to exploit “GameWith”’s main asset, which is its many game users, and to create a new IP. Manga and gaming have a close affinity with adaptations of manga into games. There are currently about 15,000 manga available online through the site’s collaboration with large publishing houses. We also want to strengthen joint promotional activities with games companies. This service has only just been released and has quite a lot of teething problems, but we’re going to work hard at bringing it on.
This is a translation of an article published by Cloud Ace, Inc.
Available online: https://www.cloud-ace.jp/case/detail29/