White Papers

iTSherpa Co., Ltd (iTSherpa) provides the social media package
GLAS, live-streaming systems, and e-commerce systems, and also
develops a variety of applications. The company’s strength is its
ability to offer the infrastructure platform for these services as an
all-in-one solution.
In the past, iTSherpa provided its service infrastructure through
physical servers at a data center and IaaS cloud services. However,
recognizing the importance of providing infrastructure through its
own private cloud, the company developed an HCI environment using
Dell EMC PowerEdge-based VMware vSAN™ (a service platform
created by an HCI using VMware vSAN). One year later, iTSherpa
also adopted Dell EMC vSAN Ready Nodes designed with a pre-
tested all-flash Dell EMC PowerEdge R740xd to develop and expand
its HCI environment.
“Using Dell EMC vSAN
Ready Nodes-based HCI for
infrastructure that supports
resource-intensive video delivery
and other services solved the
issues associated with a public
cloud and physical servers, and
allowed us to deploy high-quality
resources swiftly.”
Naoto Sonezaki
Director
Infrastructure Service Department
iTSherpa Co., Ltd.
Investigation for the introduction
of a private cloud
iTSherpa uses its strengths in software development and building
high-availability infrastructure to provide first-class services
showcasing the potential of the Internet.
The company initially offered its service provisioning infrastructure
through physical servers located in a data center; however, after
launching a service to deliver video to PCs 10 years ago, the number
of servers began to increase, eventually reaching several hundred.
Naoto Sonezaki, Director of the Infrastructure Service Department
of iTSherpa, explains, “We attempted to virtualize and aggregate
servers in 2010, but couldn’t achieve the necessary performance.” As
a result, the company relied on a public cloud and physical servers.
However, sharing and watching video on smartphones grew in
popularity, and in 2014, use of the cloud and physical servers
began presenting a variety of issues. “Delivering video required an
immense amount of resources and traffic for coding, conversion,
storage, reading, and delivery. As the volume of video delivered to
smartphones increased, the required storage capacity and other
necessities increased; we couldn’t keep up with demand by using
physical servers and there were cost issues with the cloud. Each
release of a new smartphone OS version required preparation of new
hardware, and even if we had used a high-quality, best-effort cloud
service, congestion from other users would have decreased the
speed. The possibility of maintenance was also an issue,” Sonezaki
continues.
iTSherpa recognized the importance of building its own
infrastructure to provide flexible proposals that met the quality and
cost demanded by users and began investigating the introduction of
a private cloud in 2016.
Building a private cloud platform
with All-Flash vSAN Ready Nodes
iTSherpa focused on flexible, highly scalable HCI for the investigation
and consulted Dell EMC. “After considering multiple technologies,
we decided to introduce VMware vSAN due to its scalability,
convenience, and mobility. VMware has a proven track record and a
strong partnership with Dell EMC. We were also extremely confident
we could receive support whenever necessary,” reveals Sonezaki.
Construction of the private cloud platform started in early 2017 and
was completed in less than three months. The platform consists of 10
Dell EMC PowerEdge R630 servers and has 34TB of usable storage,
with redundancy provided through RAID 1. To guarantee a sufficient
I/O performance, the platform uses NVMe (Non-Volatile Memory
Express) for caching, which is used in a hybrid configuration with
SAS SSD.
iTSherpa initially deployed vSAN. This allowed the system to switch
to a different node if a failure occurred, guaranteeing consistent
service quality for customers. However, the demand for video
delivery services grew beyond expectations and iTSherpa required
new infrastructure immediately. Sonezaki continues, “We anticipated
that the system would eventually need more storage, so we made the
first build Generation One (G1), then decided to build and operate a
new Generation Two (G2) HCI. Due to device compatibility and other