The Cloudcast (.net) #72 - Cloud-Enabling Memcached and Redis for NoSQL

Hi, thanks Brian. Good morning, and I'm very excited to be here today as we announce the general availability of our NoSQL cloud services. So yeah, it's great to be here.

Sure, Brian. Well, basically we deal with the fastest NoSQLs, which are those NoSQLs entirely served from RAM memory. We are talking about 100,000 transactions per second at sub-millisecond latencies. You need a very specific background and expertise when you are going to do that. So we both come from the high-tech space in Israel. I started a company in the '90s which I took public at NASDAQ in '97, and then I started another company which was sold to one of the large companies in the telecom area. My partner Y started a company by the name of Chendo Networks, which was eventually sold to F5. All our team come from companies who dealt with very high-speed networks, companies such as F5, Crescendo, Cisco, etc. They are used to dealing with those very high throughputs and low latencies. We met when I was a board member at Chisto, and then after we both were freed from our previous engagement, we decided to take this experience and leverage it in this new area of high-speed NoSQL databases.

Well, basically we serve companies who host their applications on clouds. In other words, it is not suitable for companies who use their own data centers and then may wish to store some of the stuff over the cloud. So we offer the service basically at the same data center as where the user would host the application servers in order to avoid any network latencies. So today we offer the service in many data centers of the Amazon Cloud, as well as a few data centers of the Microsoft Azure Cloud. The user base, one of the first things that you will do when hooking to the service would be to tell us to select which data center is used by their application servers, and we will serve from the same data center.

Exactly, exactly. This is what we do. One of the challenges by the way was how to identify the exact availability zone, because your availability zone and my availability zone, even if they are called 1A or 1B, are not necessarily the same physical zone. So we developed a method of identifying the exact location of the user's application servers, and so on and so forth.

Okay, first of all, we are talking here about open-source databases, and the open-source community of Redis, for example, is very, very active. By the way, it's sponsored by VMware. So what we do basically is on the system level. In other words, we solved problems which are not solved by the open-source community. Those are, for example, with Redis you are limited in scalability. You cannot grow your data set beyond the single master server, and this is a real limitation. You also have issues when writing to disk if you would like to back up and have everything in persistent storage. This will degrade the performance of your application, and this is another problem with those two databases. In addition, we are talking about a very unstable performance. The performance of Redis and Memcached, which is very, very fast to begin with, is very sensitive to things such as the load on your application, the load on neighboring applications, the size of the virtual servers you use, and many other things. One of the things that we do is make sure that you have the best performance and a stable performance under all circumstances. Most importantly, we offer what we call true high availability, which means that we offer inter-zone replication. You can replicate your data set within the same data center, so if a node fails, you do not lose any data and the data flow is uninterrupted. Your application actually does not know about this node failure. Also, we have something which is called multi-zone replication, which means that if the entire data center is out, we can serve your application from a replica of your data set in another data center. This is something which is extremely important. In addition, we offer a very simple backup, so you can remotely back up your data set with a mouse click, and also import your data set from remote in case of failure. Practically, this is a very sophisticated way of data portability where you can move your data set between data centers, between clouds, and so on and so forth. So these are the type of things.

Yeah, absolutely. Moreover, all the operational issues today are handled by developers as opposed to the enterprise world or the traditional world where everything was handled by IT departments. These developers or DevOps, as we call them today, there is nothing more they hate than dealing with all those operational issues such as adding nodes, removing nodes, clustering, scaling, upgrading software, recovering from failures, and all that stuff. What we do is basically free them completely from all those activities. Our services are built in a way that once you register and connect your application to our services, there is nothing more you need to do ever, except if you want to look at your dashboard and see metrics, statistics, etc. Other than that, there is nothing else you need to do. The service will automatically take care of all the operational tasks, such as recovering from failures, auto-scaling, and all those things. All that is done fully automatically.

Well, we do both things. In other words, we are great contributors to the open source of Redis and Memcached. So we share some of the things that we developed internally with the community. How to handle all the operational issues, we keep it as our proprietary technology. Today, when it comes to the world of databases, which is database as a service, the meaning of that is there are many databases around, many NoSQL databases, and there are some companies, I wouldn't name those, but there are some companies who offer what is called commercial support for open-source databases, whereby the user will use the open-source database and that commercial company, whenever he gets into issues, problems, etc., they will try to help him and provide him with sometimes additional pieces of complementary software, etc. We took another approach, which is the database as a service approach, which in a way is commercial support because you free the user from all the issues, but it's much more. Besides providing a fully automated service, you also free him from the need to have a commercial support because we are the commercial support.

Absolutely. You also mentioned new business models, etc. Here also we come up with something fresh, which is a pay-as-you-go pricing model. Today, if a user deploys by themselves a database like Redis or Memcached, they will rent a virtual machine from the cloud provider, take the open-source software, upload it, and then when they run out of memory, they will rent another server, and so on and so forth. This means that most times you are over-provisioned, because once you rent a new server, it's still empty and you do not need all that capacity. So on average, you always run with around 50% of the capacity of the server that you don't need. Now with our business model, we charge according to the actual size of the user's data set, which means that you pay only for what you consume, and you avoid this typical over-provisioning of these expensive RAM resources.

Well, basically we will try to convince everyone who uses Redis or Memcached for whatever use case to use our services, because it's better than the do-it-yourself approach. Now, if you look at the market, who is using Redis and Memcached? It started with very specific use cases like caching and timelines, stuff like that, but today you find it all over the place. You find companies, I would say that what we hear in the industry is that most new applications would use a set of three different databases: the traditional MySQL relational databases, which have been around for years, that's number one; then MongoDB, which is a NoSQL database that is also very popular; and Redis. Redis in just three to four years has become one of the most popular databases, obviously one of the top two or three NoSQL databases. So everyone who uses that horizontally is a potential customer of ours.

The easiest way would be to go to www.garantiadata.com, and then you can find all the information and how to reach us and how to talk to us, etc. We're also available on three platform-as-a-service platforms: Heroku, AppFog, and Up Harbor. So there again, you can enter the add-on section of those platform-as-a-service platforms and find our Redis Cloud or Memcached Cloud services.