Explaining packet delays under virtualization

Abstract

This paper performs controlled experiments with two popular virtualization techniques, Linux-VServer and Xen, to examine the effects of virtualization on packet sending and receiving delays. Using a controlled setting allows us to independently investigate the influence on delay measurements when competing virtual machines (VMs) perform tasks that consume CPU, memory, I/O, hard disk, and network bandwidth. Our results indicate that heavy network usage from competing VMs can introduce delays as high as 100 ms to round-trip times. Furthermore, virtualization adds most of this delay when sending packets, whereas packet reception introduces little extra delay. Based on our findings, we discuss guidelines and propose a feedback mechanism to avoid measurement bias under virtualization.

The nearest replica can be farther than you think

“…The first category [9,35,37,40,42] concentrates on studying the impact of virtualization on the network and application performance in a cloud environment. These works show that sharing of hardware resources can have a negative effect on latency, throughput, and bandwidth of applications running on these virtual machines.…”

Performance Modelling of Partially Replicated In-Memory Transactional Stores

“…First, in this type of infrastructures, little or no knowledge is available about the underlying network topology, hardware infrastructure and virtualization software overhead: this affects the possibility of measuring resource demands accurately, and makes the analytical derivation of response times cumbersome [34]. Moreover, complex applications' software stack typically lies on top of group communication toolkits that provide several inter-process synchronization services (like failure detection, group membership, remote procedure calls) the configuration and the internal design of this layer also affect performance in a way that is hard to predict [35].…”

Network Path Capacity Comparison without Accurate Packet Time Information

“…Figure 1 shows the measured arrival times of 200 packets from a 10Gbps NIC with IC enabled, and we can see that these packets are handled by 6 interrupts. VM scheduling [21], [4] is commonly used in cloud computing, and it enables multiple VMs to share the same pool of CPUs on a physical machine. However, it interferes with packet timestamping of VMs.…”

“…However, it is hard to measure such short times due to the limited system capability [9], [17]. Second, various software and hardware factors at the receiver of packets, such as interrupt moderation [19], [9] (commonly used in highspeed network cards) and virtual machine (VM) scheduling [21], [4] (commonly used in cloud computing), greatly change the original packet time information. As a result, the packet time information measured by the packet receiver is not correct.…”