Apple’s announcement of iCloud was long anticipated but like any Apple news it caused a lot of discussion within the industry. What Apple is planing to deploy not new nor novel. Many companies have been providing similar services for a long time. One of the early pioneers was T-Mobile USA that introduced a peer-to-peer picture sharing application almost a decade ago. At that time it was rather obvious that synchronization capabilities of feature phones (this is preceding the first iPhone by 5 years) was nowhere near to be useful or ubiquitous enough. Similarly devices lacked a WiFi interface. Cost of GPRS bearer (used to deliver those low resolution pictures taken on sub mega-pixel cameras) was exorbitant to allow a user to send the same picture to multiple friends across different times. Instead a network storage (cloud wasn’t a popular term back then) was devised for picture sharing. T-Mobile USA insisted on its handset vendors to support so called “Picture Blogger” service and the result was a varying level of success. Similar to what Apple is attempting, T-Mobile didn’t charge for the service; instead it relied on the level of stickiness or social-networking impact it generated.
Fast-forward to 2011, world is a very different place. 13 Kbit/s GPRS networks have been replaced by 42 Mbit/s HSPA+ capabilities. Instead of a sub mega-pixel camera typical smartphone features a 5+ Mpixel camera. Almost always the same phone has WiFi radio. Furthermore in the last 10 years users have added many devices to their portfolios: MP3/MPEG devices to consume music and casual picture/video viewing, tablets for primarily accessing content using a web-browser, traditional PC/Mac for both accessing and generating/editing content, cameras for generating content.
As always Apple does what it does best: improves upon what others attempted and failed or faced limited success. They take the concept of store-forward architecture to make synchronization and back-up seamless. Total control of the user interface across all Apple devices give them the ability to make something pretty simple and straight-forward to look as if “magic”.
All of this doesn’t come free for Apple. They need to spend capital and operational costs for storage, computational power as well as bandwidth to Internet. Apple’s new data center in North Carolina was designed to store Petabytes of digital media for iCloud users. Apple announced that it will give each user 5 GB of storage for free while charging for additional storage. Just as a comparison, Amazon Cloud service charges $0.14 per GB for the first TB of usage. In other words, if Apple achieves the same level of efficiency of AWS S3 service, 5GB of storage per user would cost less than 3 quarters. Apple provided content such as music, books do not count against this quota since such content is served from a central repository as opposed to being stored individually per user.
The real Achilles’ heel of the iCloud strategy seems to be the bandwidth Apple needs to provide within and primarily in and out of its data center. Let’s try to analyze why:
- Unlike the storage quota, Apple didn’t put any restrictions on the amount of traffic a user can generate to and from its iCloud storage. We believe this is primarily to make the service more appealing and easier to explain to the consumer as opposed to putting meters, throttles and caps as wireless and even some wired service providers do.
- Initially the only serious traffic generator will be pictures. Since the amount of user generated content has no limits, apart from local device storage and possibly bandwidth for Internet access, every taken picture may end up being uploaded to the cloud and downloaded to multiple devices. Today picture sharing, storage sites such as Picasa, Flickr, even Facebook provide very cumbersome methods. Level of integration is limited; even with apps for these services on smartphones “seamlessness” is not there. That is why even when Facebook boasting a rate of 1000 picture uploads per second translates into about 5 pictures per user per month. We expect with iCloud this will change dramatically. Number of pictures taken and shared with other devices will be in the order of 100s per month for average user whereas power users may synchronize 1000s per month.
- Assuming a monthly picture upload traffic of 1000, an average picture size of 1.5 MB and an average number of 2 devices to be synchronized, total monthly traffic per user may reach 3 GB.
- We estimate Apple needs approximately 50 Kb/s bandwidth per user to effectively serve 3 GB/month synchronization traffic. In other words, Apple must have at least a pair (for redundancy) of 50 Gb/s pipes for every 1 Million iCloud user. Even with very competitive pricing (around $1 per Mb/s per month), this would require to spend $100K per month per 1 Million users.
- Considering the fact that iCloud is free and will be available for every iPhone, iPad, Mac, it is conceivable for Apple to reach 100 Million user mark within a couple of years. In our estimation that would require Apple to spend roughly $120M per year just for Internet pipes of approximately 5 Tb/s.
This cost and infrastructure scale requirement are a likely reason why Apple didn’t extend the iCloud service to video. If we consider that iPhone supports HD (720p), even 20-25 minutes of video per day may result in excess of 500 MB. Certainly that would change the scale requirements dramatically, requiring at least one order of magnitude bigger pipes compared to iCloud service limited to picture synchronization.
A better strategy for Apple could be to use its other assets, especially at home such as Airport/Time Capsule to optimize the path between the devices and the storage location. It looks as if iCloud is an excellent solution for iTunes content and an acceptable solution for user-generated low-density media such as pictures, files, messages, contacts, etc. However, for heavy-density media such as video Apple needs a complementary solution. iCloud as it is cannot deliver it at the quality and price level needed.